In our last blog in this series, we tackled a common question: [*When should I update my personas*](/blog/when-and-why-to-update-personas)*?* The answer, in short: almost never. Personas are a foundation—not a real-time dashboard—and you don’t pour new concrete every week.

But if that post answered the *when*, this one’s about the *what*. As in: How many [persona sets](/docs/abstractions/persona_sets) do I actually need? 

But it’s 2025 and we’re all very busy, so here’s the TLDR: **You almost certainly only want ONE persona set.** 

But for those of us who want a deeper dive, read on. 👇

## What is a persona set?

A [persona set](/docs/abstractions/persona_sets) is a small group of customer personas—each one representing a distinct cluster of your real customers. These personas are archetypes based on shared traits and behaviors, built from the people who have actually converted. They're the foundation for structured personalization across your brand.

So let’s clarify terms:

* A **Persona** is a single, representative profile within a set. Think "Urban Millennial Max" or "Suburban Saver Susan."  
* A **Persona set** is the full group of those profiles—usually 3 to 7 personas total—that collectively describe your customer base.

Personas aren't based on assumptions or hypotheticals. And they’re definitely not built around leads or website visitors. The reason is that personas need to be grounded in actual converted customers, you know the ones who said *yes*, not the ones who ghosted you. These people have the traits you want to seek out in new leads who you also want to convert.

The purpose of a persona set is to give your brand a coherent, actionable structure for engaging your audience. You use them to vary messaging, align creative, and personalize campaigns across the funnel.

## How do we build them?

At Faraday, we build persona sets using [enriched data](/docs/abstractions/datasets) and [k-means clustering](/blog/k-means-clustering-customer-personas). We start with your customer file—ideally recent customers—and enhance it with third-party traits from the [Faraday Identity Graph](/features/faraday-identity-graph), which includes demographic signals (like age, household size, income range), geographic markers, lifestyle and behavior data, and more. This rich, consented data allows our AI to surface patterns that wouldn’t be obvious from first-party data alone.

After we enrich your data, we use a method called *k-means clustering*, which automatically groups all customers into 3–7 distinct clusters. That range is deliberate: too few, and you lose nuance; too many, and the personas stop being practical—after all, personas power personalization and you don’t really want to customize 37 emails each time you make a send\! Remember, our AI is always powerful but is only value-driving when tied to a concrete business function.

The result: a persona set that reflects your business as it is today—and that you can use to personalize everything from top-of-funnel ads to post-purchase journeys.

## So how many persona sets do I need?

**In a nutshell, one is enough (we promise).**

Here’s the temptation (and we’ve heard it many times): you start to see interesting subgroups, and you think, “Should I just make a whole new persona set for my high-value customers? Or for leads? Or for gift buyers?”

But this is important: resist the urge. One persona set, built on your customer base (ideally recent customers), can answer all of those questions *if* you use it well. And what’s more, every additional persona set you create adds complexity—fragmented messaging, duplicated creative, messy reporting—with very little return.

Think about it practically. Let’s say you’re using personas to personalize emails. If you have 35 personas, you need 35 distinct emails each time you make a send, tailored to each persona. The same would be true for a direct mail campaign, or any other use case built on personas. 

And if you really want to dig into those subgroups, that is still a great option with one persona set, you just use **secondary analysis** instead of additional personas.

## Use secondary analysis for deeper insights

Let’s say you’re curious about your high-LTV buyers. Or people who tend to buy during promotions. Or those who give gifts.

The instinct might be to build a new persona set just for those customers—but that’s the wrong move.

Instead, use **secondary analysis** on the persona set you already have. This means identifying which of your existing personas are most likely to exhibit those behaviors. Does persona B have more high-value buyers? Does persona D respond more often to promotions? 

This lets you tailor messaging, promotions, or next-best-action logic *within* your current framework—without rebuilding the architecture from scratch.

You’re analyzing, not segmenting. That distinction matters. If your persona set is the architectural structure, secondary analysis is like putting sensors in each room to see who uses what, when, and why. It adds intelligence without requiring you to redraw the floorplan.

That’s why secondary analysis is the right tool for exploring behavior. It lets you get more specific without fragmenting your strategy.

## So would I ever use a second persona set?

We get this question a lot—and it’s not a bad one. In theory, there *are* situations where a second persona set could make sense. But in practice, they’re vanishingly rare.

Let’s say your business has two truly distinct models—like Netflix in the DVD-by-mail era vs. its streaming service. Maybe the customer base for each product barely overlaps. In a case like that, sure, there’s a logic to considering separate persona sets.

But even then, we don’t recommend starting there. First, build a single persona set from your full customer base and run secondary analysis. You may find that your personas naturally reflect the split without needing to manufacture two totally separate structures.

The risk of jumping too soon into multiple sets is high: fragmented messaging, duplicate effort, and confusing strategy. You’ve gone from one coherent framework to parallel tracks that require twice the upkeep—and often deliver half the clarity. 

**As Andy put it in a recent meeting:** 

**“If you really, really, really, really think you need more than one… you still don’t.”**

## How to use personas

Once you have a persona set built from your customers, the value comes from how you apply it.

You don’t just file it away as “interesting.” You use it as a structure to guide how you engage people across the entire funnel.

### Start with leads

Even though your personas were built on customer data, you can (and should) apply them to leads and prospects. Every lead can be mapped to the closest-matching persona based on their traits.

That lets you do smart things early—like showing persona-aligned Facebook ads, driving people to persona-specific landing pages, or customizing messaging tone based on emotional drivers. 

**As Andy says:**

**“You want continuity between how you handle leads and how you handle customers. That means the architecture has to start early.”** 

### Personalize campaigns

This is personalization 101\. Instead of blasting one-size-fits-all emails or ads, you vary your content based on what you know about each persona. Not to the point of infinite complexity—but enough to make the message feel more relevant, more human.

And you don’t need 37 personas to get there. Three to seven is plenty.

### Operationalize and scale

Personas are a strategic lens. They help you organize your audience, inform creative, and structure internal decision-making. In a large org, you could even assign brand managers to own each persona—getting deep into what makes that group tick.

Just don’t mistake them for a silver bullet. You’re not using personas to predict exact behaviors (like who’ll buy Product X on Thursday). For that, you’d use specific outcomes or predictive models. But for building audience structure and making smarter content decisions, personas are the starting point.

## Conclusion

If you take nothing else from this, remember this: your persona set is not a one-off project. It’s the structural framework—the architectural blueprint—that supports your entire customer engagement strategy.

When you get it right, everything else becomes easier. Ads align. Messaging resonates. Creative connects. And the decisions you make, from acquisition to retention, rest on a solid foundation.

Don’t overcomplicate it. Build once. Use often. And always start with the people who said yes.

Nearly every brand needs just one carefully built persona set—use secondary analysis, not additional sets, when you want more detail.

So how many persona sets do you need anyway?

Sharing data for testing is a big step, and we know that sending anyone your closely guarded trove of customer information can be a little scary. That’s why, at Faraday, we take clear, deliberate steps to eliminate any risk to your records.

Here’s what to know about how we handle your data, in plain terms.

## NDA: At a glance

To start, we make sure the rules are clear on both sides: we’re bound by a mutual NDA, meaning confidentiality goes both ways. 

Here are some key points that keep you secure:

* As a data processor under data privacy laws, any data a client or prospect shares with us is used solely for providing our services or running agreed-upon tests. So when you send us your data, you can feel secure in knowing exactly what it will be used for and when.  
* Data security is our top priority. We’re [SOC 2 Type II audited](https://www.imperva.com/learn/data-security/soc-2-compliance/), read the next section to learn more.  
* Our mutual NDA ensures that any data you share is protected by strict confidentiality obligations.

## Security: We've got the receipts

At Faraday, we take data security seriously. That’s not just talk—we’re independently audited, fully encrypted, and built to comply with modern privacy laws. Whether you're testing data or running production campaigns, your information stays safe, confidential, and under your control.

Here’s how we do it:

* **SOC 2 Type II audited:** We undergo regular third-party audits to ensure our security controls are working as designed. This certification covers data protection, availability, and confidentiality across our systems.  
* **Compliant with major privacy laws:** Faraday operates as a data processor/service provider under regulations like GDPR and CCPA. That means we only use your data for the agreed-upon purposes—and never for our own gain.  
* **Fully encrypted:** All customer data is encrypted both at rest and in transit using industry-standard protocols. No shortcuts.  
* **Minimum internal access:** Access to data is tightly restricted, even within your own brand’s workspace. Only the team members directly responsible for running your project can view your data. (For example, the person writing this post—a marketer—has never seen a client record.)  
* **Siloed accounts:** Your data is never pooled with other clients’ data for modeling or analysis. Each account is kept fully separate, and customer-level data is only ever used to train models for your specific use case.

Want to dig deeper? Take a look at [our security page](/security).

## Validation: See it before you believe it

And why do we ask for your data before you buy? 

Transparency and evidence matter. At Faraday, we don’t believe in taking on clients we can’t generate value for. That’s why we conduct a complimentary holdout test by default, before you ever commit. This test only shows how well we can model on your data, with your data, so it requires an upload prior to signing up.

* We don’t just tell you our agents are working, we show you. To accomplish this, we run secure, real-data tests to validate Faraday’s performance before any commitment is made.  
* With these initial models, you’ll see the impact of our predictions on your specific use case, with your individual data (enriched with the Faraday Identity Graph), not just a generic demo.  
* No [black box](/blog/the-ai-blackbox-is-a-liability), no sales fluff, just results you can evaluate firsthand.  
* The results are yours to explore—you’ll see performance before making any commitment.  

Want to know how that works? Take a look at this blog about how [Faraday validates our models using your unique data](/blog/holdout-testing-with-faraday).

## In conclusion

Protecting your data is one of the most important things we do at Faraday. Our platform doesn’t work unless your data is safe—and we take that responsibility seriously. Hundreds of brands have trusted us with their data, and we’re proud to say this is one area where we consistently shine.

If you still have questions or concerns, don’t hesitate to reach out. We’re here to help.

Faraday takes a variety of measures to protect your data during pre-purchase model validation, here's the TLDR

How Faraday protects your data during testing

If you read our last blog, [What’s in a deployment? How to read your Faraday outputs](how-to-interpret-your-outputs), you’ve already figured out what comes out of Faraday, now it’s time to understand how to shape it. While that post broke down the different types of payloads you can get from Faraday, this time, we’re going one level deeper: into the deployment settings themselves. 

Think of this as a user-friendly guide to all the options you might see when setting up a deployment—from filters to formats, geography to structure, and beyond. These settings help ensure that your predictions don’t just look good in a spreadsheet—they’re ready to plug into your actual tools, from CRMs to ad platforms.

If you've ever wondered what settings like "representation format," "add geometry," or “deployment analysis” actually do—or how they impact your final output—this blog is for you.

## But first, what’s a deployment again?

[Deployments](../docs/abstractions/deployments) (called “targets” in the API) are how you tell Faraday where to send your predictions and what those predictions should look like when they arrive. They’re created within a pipeline and allow for detailed customization to make sure the output fits your intended use case—whether you’re running a digital ad campaign, syncing data back into a CRM, or creating a list for direct mail.

### 1\. Where your predictions go

The first decision you make when creating a deployment is choosing its destination. This could be a Faraday-managed connection (like your CRM, ESP, or ad platform), a location in your own infrastructure (like a cloud bucket or database), or just a downloadable CSV. This is the foundation—everything else shapes what gets delivered. If you’d like to see a CSV broken down, check out that [last blog](how-to-interpret-your-outputs) we mentioned\!

### 2\. What your predictions look like

#### Representation format

In the app, we define this as the following: “Choose how you want the pipeline's output to be formatted for this target."

Think of representation format as the style or privacy level of your output—whether it’s raw, anonymized, or summarized. Deployments can be configured in a few different formats depending on how you plan to use the data:

* **Hashed**: This is the default format, typically used for ad platforms. PII is hashed to protect identity while allowing for matching.  
* **Referenced**: This format lets you use your own internal identifier to merge the data back into your system.  
* **Aggregated**: Useful for geotargeted ads or planning, canvassing, direct mail and more,  this format summarizes how many individuals fall into key groups (like a persona or outcome) per geographic area. You’ll get a row per geographic area, with counts of people per persona or outcome. In this row, data can be aggregated into the following categories: ZIP code, census tract, carrier route, county, metro area, state, etc.  
* **Identified:** This format includes direct personal identifiers like name and address, allowing you to activate immediately through mailers, call centers, or CRM integrations. Because this format exposes PII, it’s only available in secure workflows with appropriate safeguards and permissions in place.

#### Column headers

You also get to choose how the column headers are labeled:

* **Machine-friendly** headers are concise and consistent, ideal for automated ingestion. Useful especially for our API users.   
* **Human-friendly** headers spell things out a bit more, which can make it easier for marketers and other stakeholders to work with the data directly, especially in tools like CRMs or email platforms. For example, “`fdy_persona_set_{UUID}_persona_name`” could appear as “`fdy_persona_set_customers_persona_name`” in human-readable mode.

### 3\. Filters, limits, and structure tweaks

#### Filters

Filters allow you to narrow your deployment to just the segments of interest. You can filter by personas, outcomes (via percentile or probability), or even recommender results (such as top product matches). This is how you target only the individuals who matter most for a given campaign. For example, give me the top 10% of scorers for my outcome. 

#### Limits

Need to limit the number of people you're deploying to? You can set a top or bottom row limit—useful for campaign budgeting,  prioritization, or suppression. 

**📘Note:** in preview mode, deployments are capped at 1,000 rows. We break down the difference between preview mode and a deployment later in this blog\!

#### Structure

You can also rename or reorder columns to better match the formatting requirements of your destination platform. For example, if you're sending a list to LinkedIn or Facebook, you can use one of our pre-formatted options for easy compatibility.

### 4\. Geography & “Add geometry”

When building a geotargeted campaign, geographic options and the “add geometry” setting come into play. You can define the geographic granularity—ZIP code, metro area, census tract, etc.—and include shape data for use in map-based planning or delivery platforms. This setting only applies to aggregated deployments. 

If you’d like more information about how to read these outputs, just look at [our recent blog](how-to-interpret-your-outputs) which broke down how to read the labels\!

[![Example of what a add geometry output might look like](/images/blog/add-geometry-output.png)](/images/blog/add-geometry-output.png)

### 5\. Deployment analysis

Deployment analysis gives you a snapshot of what’s inside your deployment before you send it off. It includes summary stats on key traits and geographic distributions, delivered as a downloadable PDF or accessible via the API. This can help you validate that the list aligns with your goals—for example, checking that you’re really hitting the right income bands or urban areas before launching a campaign.

📘 **Note:** Access to traits in deployment analysis depends on your account permissions. If you’re not seeing the traits you need, reach out to your account manager to learn more about available options.

### 6\. Score explainability

Finally, if you've selected prediction explanations during pipeline setup, your deployment can also include score explainability. This details which traits had the greatest influence on an individual’s predicted score. It’s useful for internal transparency, model evaluation, or even surfacing to customer-facing teams.

If you're not sure whether you've enabled or don’t know how to read it, reach out to your Faraday team, we’ll confirm or help turn it on.

### 7\. Preview mode vs. full deployment

Before launching a full deployment, it’s helpful to run a preview—and it’s important to understand how previews behave differently. In preview mode, your deployment is capped at 1,000 rows, no matter what filters or limits you’ve set. This cap exists so you can sanity-check your configuration without triggering a full export. Once you’re ready to go live, the full deployment will reflect your entire population and apply all settings as specified—no artificial row limits included.

## Wrap-up

Deployments are more than just the final step in your predictive workflow—they’re the bridge between insight and action. With the right configuration, you ensure your models aren’t just accurate, but actually useful—delivering the right predictions, in the right format, to the right destination.

A practical guide to customizing Faraday deployments—from filters and formats to geography and output structure—so your predictions arrive ready for action.

How to shape your deployments: a guide to filters, formats, and more

“So what do we actually get from Faraday? Is there some kind of Excel file I can open, with all the data points laid out in plain English?”  \- Your CEO in your 1 on 1\.

Sound familiar? In this post we want to help you communicate to your leadership what they can expect the Faraday system to provide them, enabling the marketing and product workflows of their dreams.  

Faraday is often used to optimize outreach (through use cases like [lead scoring](https://faraday.ai/templates/lead-rejection) and [personalization](https://faraday.ai/templates/thematic-personalization)), helping teams identify which people are most likely to take meaningful action and tailor their communication to them. But the output you get includes much more than just a score. It’s a full snapshot of each person or household in your audience: scores, traits, behavioral segments, and cohort tags.

This post breaks down exactly what you'll find in a Faraday [deployment](https://faraday.ai/docs/abstractions/deployments), and how to use each piece to drive smarter strategy and communication across the funnel.

## Outcome scores: the two numbers that tell the story

This section refers specifically to [outcome models](https://faraday.ai/docs/abstractions/outcomes)—predictions like “which of my leads are most likely to purchase,” “who will respond to outreach,” or “who’s at risk of churn.” These are the most common prediction types in Faraday, and they return two scores per person.

Every outcome prediction in Faraday includes:

* [Propensity percentile](https://faraday.ai/blog/propensity-vs-probability-predictions): how this person ranks compared to everyone else in your prediction population. For example, a 95th percentile means they’re more likely to convert than 95% of people in the group.  
* [Propensity probability](https://faraday.ai/blog/propensity-vs-probability-predictions): the actual chance this person will take the action you care about. For instance, a 0.72 means a 72% likelihood of conversion.

If the percentile helps you prioritize, the probability helps you forecast. Together, they give you both a ranked list and a sense of real-world likelihood.

In this post, we’re using a CSV file as the example format for clarity, but keep in mind, the structure and content of a Faraday deployment is the same no matter how you receive it—whether via data warehouse, API, or direct integrations. 

But regardless of where you get this information, we know that all this information can feel a bit overwhelming at first. While our support team is always here to help you make sense of your results, we’ve broken things down below to give you a head start.

Below is an example of an output that you could expect to see when you request a probability or propensity outcome. These returns can look a little intimidating if you haven’t read one before so lets break down what you’re looking at.

[![Sample CSV output showing outcome score](/images/blog/how-to-interpret-outputs/prob-vs-prop-returncsv.png)](/images/blog/how-to-interpret-outputs/prob-vs-prop-returncsv.png)

A much more detailed breakdown of propensity vs. probability is explained in our previous blog, [Propensity vs probability: Understanding the difference between raw scores and probabilities](https://faraday.ai/blog/propensity-vs-probability-predictions), written by one of our incredible data scientists, [Thibault Dody](https://www.linkedin.com/in/thibault-dody/). If you'd like a more detailed explanation of how these scores are calculated and when to use them, check that one out\!

## Recommender outputs: ranked predictions across many options

While outcome models predict whether someone will take a specific action, [recommender models](https://faraday.ai/docs/abstractions/recommenders) answer a different kind of question: “Which of these options is best for this person?” Rather than producing percentiles or probabilities, they focus on relative ranking, helping you surface the best match for each person from a set of options.

For example, a recommender might tell you which product a customer is most likely to purchase, or which offer will resonate most. Instead of returning a single score, these models return a ranked list of options for each person, based on predicted likelihood.

In a typical recommender output, you’ll see:

* A **list of recommendations** (e.g., products, offers, messages)  
* A **rank** indicating preference order  
* A **relevance score** (optional), showing how strong the model’s confidence is

You can use these outputs to power personalized product recommendations in email or on your website, route individuals into different messaging tracks based on their attributes or behavior, and dynamically customize the content or offers they see to create a more relevant and engaging experience. This kind of real-time personalization can significantly boost conversions and improve lifetime value\!

Here, we get to see an example of a recommender model return:

[![Sample CSV output showing recommenders](/images/blog/how-to-interpret-outputs/rec-returncsv.png)](/images/blog/how-to-interpret-outputs/rec-returncsv.png)

## Personas: behavioral context you can act on

Faraday [personas](https://faraday.ai/docs/abstractions/persona_sets) are behavioral segments created by clustering your customers based on hundreds of data points. They help you go beyond just knowing who’s likely to act—you get a sense of why they might act, and how to engage them. We build these personas using a k-means clustering process, which groups similar individuals together based on shared traits and behaviors, you can learn more about how it works [here](https://faraday.ai/blog/k-means-clustering-customer-personas).

You can think of a persona as archetypes. An “Eco Enthusiast” might respond well to sustainability-focused messaging, while a “Status Seeker” may prefer premium positioning.

Here’s how you might expect a persona return to look.

[![Sample CSV output showing personas](/images/blog/how-to-interpret-outputs/persona-returncsv.png)](/images/blog/how-to-interpret-outputs/persona-returncsv.png)

Personas also help you:

* **Tailor messaging and offers** to match motivations  
* **Pick the right channels**, whether that’s social, email, or direct mail  
* **Inform product and UX decisions** based on lifestyle or priorities (e.g. “Busy Parents” value speed and simplicity)

Want to go deeper? Our CEO, [Andy](https://www.linkedin.com/in/rossmeissl/), just wrote a whole [blog](https://faraday.ai/blog/when-and-why-to-update-personas) about how personas work, and why they’re the foundation of various other predictive tools.

## Traits: the signals behind the scenes

[Traits](https://faraday.ai/docs/abstractions/traits) are the demographic, behavioral, and lifestyle characteristics that describe each individual in your prediction results—attributes like age range, income bracket, homeownership status, household size, and more. At Faraday, these traits are powered by our proprietary [Faraday Identity Graph](https://faraday.ai/features/faraday-identity-graph) (FIG)—a comprehensive, privacy-conscious dataset that includes rich, consent-based attributes on over 240 million U.S. adults and their households. FIG combines data from hundreds of public and commercial sources into a unified identity framework, allowing us to surface actionable insights without requiring you to collect or manage sensitive personal data yourself.

[![Sample CSV output showing traits](/images/blog/how-to-interpret-outputs/trait-returncsv.png)](/images/blog/how-to-interpret-outputs/trait-returncsv.png)

These are the features the model uses to make its predictions. They also help you:

* Filter or sort your results  
* Build secondary audiences  
* Spot unusual patterns (e.g., "why are so many top scorers renters?")

In short, traits let you peek under the hood. They can explain why someone scored highly—or give you clues about where to dig deeper. 

If you’d like to dig a little deeper into the 1500+ rich consumer traits included by default with every Faraday  agent, take a look [here](https://faraday.ai/features/faraday-identity-graph)\!

## Cohort membership: groups you define

Faraday also returns [cohort](http://i) indicators that show whether someone belongs to one of the custom segments you’ve defined—like "Past purchasers," "Subscribers," or "Churn risks."

You can also create cohorts using the data already available within Faraday. For example, you could define a cohort for men between the ages of 21–45 who live in California, Utah, or Nevada. Once created, each person in your population is tagged with a simple true/false value—do they match this cohort or not? That makes these cohorts incredibly useful as [boolean filters](https://www.scribbr.com/working-with-sources/boolean-operators/): you can use them to include or exclude groups from predictions, targeting, or reporting.

Cohort membership helps you:

* Exclude irrelevant groups (like existing customers)  
* Compare segments side-by-side  
* Layer in your own business logic  
* Understand whether a person fits within an important audience segment

If that sounds similar to Personas, there are some critical differences worth calling out:

* **Personas are discovered.** Faraday generates them by identifying patterns in behavior and demographics across your customer base.  
* **Cohorts are defined.** You create them based on your business rules and goals—like engagement history, purchase frequency, or lifecycle stage.

In other words, Personas reveal who people are; cohorts reflect what they’ve done, who they are, or how your business thinks about them.

## Prediction explanations: why someone scored the way they did

Sometimes, knowing what someone scored isn’t enough—[you want to understand why](https://faraday.ai/blog/what-is-explainable-ai-why-is-it-important). That’s where prediction explanations come in.

When enabled, explanations provide a breakdown of which traits had the biggest impact on a person’s prediction. You might see that homeownership, household size, or a specific behavioral indicator pushed a score up or down. 

[![Sample CSV output showing explainability](/images/blog/how-to-interpret-outputs/explainability-returncsv.png)](/images/blog/how-to-interpret-outputs/explainability-returncsv.png)

These explanations can help your team:

* Sense-check the model's logic  
* Find patterns across high or low scorers  
* Build trust and transparency into outreach decisions

In short, explanations turn the model from a black box into something you can interrogate and understand. For more detail on how to interpret this output, see the documentation on [deployments](https://faraday.ai/docs/abstractions/deployments#understanding-score-explainability). 

## Putting it all together

A raw score can only tell you so much.

Let’s say you know Fred is **40% likely to convert**. That’s useful. But it’s limited. Without context, you don’t know who Fred is, what kind of message might resonate with him, or whether he even fits your audience strategy.

Now imagine you have a full output:

* Fred is in the **40th percentile**, which means he ranks higher than 40% of your prediction population—and he has a **0.40 probability** of converting.  
* He belongs to your **Newsletter Signups** cohort  
* He’s assigned to the **"Value Shoppers"** persona  
* He lives in a **three-person household** earning **$50-75k**

Suddenly, you’re not just guessing—you’re seeing the full picture. Maybe he’s not your top priority, but his persona suggests he’s responsive to discounts, and he’s already shown interest by signing up for your emails. That’s enough to make a targeted, budget-conscious offer worth testing.

Understanding your outputs turns a single prediction into a story. And stories are what make strategy work. Want to see a real example? 

## In conclusion

Prediction results from Faraday aren’t meant to sit in a spreadsheet. They’re designed to drive action: better segmentation, sharper messaging, and smarter growth. When you understand what each piece means—scores, personas, traits, cohorts—you can put the full power of predictive AI to work.

Want help interpreting your outputs? Reach out to our team or explore more in the Faraday docs.

This guide walks through every component of a Faraday deployment—scores, personas, traits, cohorts, and more—so you can turn raw predictive outputs into strategic, actionable insights.

What’s in a deployment? How to read your Faraday outputs

[![Faraday personas dashboard](/images/blog/when-to-update-personas/persona-set.png)](/images/blog/when-to-update-personas/persona-set.png)

Not long ago, a client asked us a simple question: *"When do we need to update our personas?"* They had just launched a new campaign, were seeing some shifts in engagement, and started wondering if maybe their whole persona set needed a refresh.

It’s a fair question—and a tempting one. When you’re working with data-driven segments, it’s easy to assume they should constantly change. But our answer surprised them.

**“Probably not.”**

Faraday Personas serve as the foundation a range of predictions can be built on. They help you organize your customer base in a way that’s strategic, insightful, and easy to act on. But like any good foundation, you want to get it right before you start building on top of it.

Sure, things can change. Maybe you want to add an extension, knock out a wall, or open up the kitchen. But those changes are a lot easier—and less disruptive—if the foundation is solid and built to last.

So when do you actually need to revisit your personas?

The short answer: rarely. Once you’ve landed on an architecture that works, stick with it.

Let’s unpack why—because the logic behind that “rarely” is more interesting than it sounds.

[![Image of foundation](/images/blog/when-to-update-personas/foundation.jpg)](/images/blog/when-to-update-personas/foundation.jpg)

## How personas are generated: a quick look at k-means

Just a second there—we are going to tell you when to update your personas, but first, in the spirit of explainability, we wanted to give you a little more context about how these personas are computed to begin with. Under the hood, Faraday uses a variant of a decades-old algorithm called “k-means clustering” to create personas.

Here’s how it works:

* **Fig 1:** The algorithm starts by placing random center points across your dataset—think of these as placeholders for where clusters might end up. It’s easiest to visualize this in two dimensions—like income and age (as shown below)—but in practice, our models actually cluster across seven dimensions at once.  
* **Fig 2:** Each customer record is assigned to the center point it’s closest to, using straight-line (Euclidean) distance. This step groups people based on proximity to those initial centers.  
* **Fig 3:** Once everyone is grouped, the center points “move” to the average location of their assigned members—this is the re-centering step.  
* **Fig 4:** After the centers update, each customer is reassessed and may be reassigned to a new group based on updated distances. The process repeats—reassign, recenter—until the centers stop moving.

[![K-means diagram](/images/blog/when-to-update-personas/k-means_diagram.png)](/images/blog/when-to-update-personas/k-means_diagram.png)

That final clustering defines your persona set.

The result is a set of personas that’s grounded in data, tuned to your strategy, and stable enough to support real decision-making.

## Personas are a framework, not a forecast

A [Persona set](https://faraday.ai/docs/abstractions/persona_sets) in Faraday’s platform is a structured way to segment your customer base. It’s not a supervised prediction model in itself—but it is still predictive. Built using unsupervised machine learning, Persona sets identify natural clusters in your customer data based on shared behaviors and traits. These clusters don’t predict a specific future outcome, but they do reveal the underlying structure in your audience—patterns you can act on.

In that way, Personas become a foundation for other predictions. By dividing your audience into meaningful, behavior-based clusters, you create a consistent lens through which you can view performance, target messaging, and even optimize creative strategies.

That’s why it’s important to get it right. And sometimes, getting there takes a few iterations:

* You may need to refine *who* you're segmenting (e.g. maybe you only want to include actual buyers, not people who just requested a sample).  
* You might experiment with *which traits* produce the most useful clusters (like dropping gender if it isn’t meaningful for your business).  
* You’ll need to choose how many Personas your team can support operationally—four might be optimal, but maybe three is more manageable.  
* You might decide to limit the data to only recent customers if your business has changed significantly over time.

These early decisions shape your architecture. But once you're confident in the structure—and it's aligned with your business goals—you shouldn't need to change it often.

In other words, once that foundation is poured and leveled, you’re ready to build. And just like with a house, it's much easier to make layout decisions or room additions early—before the drywall goes up and the furniture’s moved in. After that, you're not just shifting strategy; you're tearing up the floorboards. So it's worth taking the time to get it right.

## Stability matters

So now that we understand where the Personas come from and what they do, we can start to talk about why or why not to consider updating them. 

And here’s the upside of *not* tinkering: consistency gives you clarity.

If your Persona structure stays stable, you can actually see how your audience shifts over time. Maybe one segment’s getting smaller. Maybe another’s growing thanks to a campaign you launched six months ago. Maybe a new customer type is starting to show up in a meaningful way.

None of that is obvious if you keep changing the structure. You need that solid foundation if you want to track change.

Think of it like this: if you renovate your house every year, you’ll never really know whether the neighborhood’s changing—or if you’re just seeing different views through different windows.

## When you *should* consider a change

All that said, sometimes it’s the right move to rebuild.

Here are the situations where it makes sense:

* **Gradual obsolescence.** One of your Personas slowly fades out over time. They’re no longer showing up in your data, and that segment just isn’t relevant anymore. It might be time to re-cluster and rebalance the set.  
* **Big business changes.** Think mergers, acquisitions, or divestments. If your business fundamentally changes—like Hasbro acquiring Wizards of the Coast—you’re suddenly dealing with a whole new audience (hello, D\&D players). That’s not a tweak. That’s a structural shift.

These are rare, but when they happen, it’s worth stepping back and reassessing.

## Build it to last

At the end of the day, your Persona set isn’t just a report—it’s the foundation your customer strategy is built on. It supports everything: creative, targeting, product design, lifecycle messaging, and more.

That’s why it’s built to be stable, and consistent. You don’t want to tear up the foundation unless you have a very good reason.

And when that moment does come—whether it’s time for a renovation or a full rebuild—Faraday’s team is here to help you do it the right way.

But most of the time? You’re better off getting it right the first time and building on what’s already solid.

A solid persona set should be built to last—because stability, not constant revision, is what makes it most valuable for long-term strategy.

When to update your personas - and why you probably shouldn’t

**TL;DR:** This guide shows you how to plug Faraday's predictive lead scoring directly into your LeadConduit flow. It takes about 10–15 minutes if your [Lookup API](https://faraday.ai/docs/features/lookup-api) pipeline is already set up, and you’ll get real-time intelligence on every incoming lead.

— — — 

Already an [ActiveProspect](https://activeprospect.com/) user? Great call 👏

But did you know you could be getting even *more* value out of [LeadConduit](https://activeprospect.com/leadconduit/)?

By integrating Faraday with ActiveProspect, you can automatically score incoming leads in real time, allowing LeadConduit to reject low-quality ones before they hit your CRM. That means more time for your team to focus on the right people—and less budget wasted on the wrong ones.

In this post, we’ll walk you through exactly how to connect Faraday to your LeadConduit flow using a custom step. It’s fast, flexible, and a whole lot easier than building an internal model from scratch.

## Why integrate Faraday into LeadConduit?

This integration gives you real-time access to Faraday’s AI-powered lead insights—including conversion propensities, persona classifications, and custom attributes. That means you can stop low-propensity leads at the source, route high-quality prospects based on persona-fit or probability thresholds, and enrich the rest for downstream personalization.

It's fast (millisecond response time), flexible, and fully customizable to your lead acquisition goals.

## Step 1: Get your Faraday setup ready

Before anything can flow through LeadConduit, you’ll want to make sure your Faraday pipeline is in place and ready to respond. This is your starting point: a deployed Lookup API endpoint that’s ready to score leads in real time.

To make the connection work, you’ll need a few pieces of info from your Faraday account:

* Your completed Lookup API deployment—this is where all your scores are stored before retrieval  
* Your target ID  
* Your endpoint URL: `https://api.faraday.ai/v1/targets/<YOUR_TARGET_ID>/lookup`  
* Your production API key (a.k.a. your token)  
* Optional: A list of payload response fields you want to retrieve (e.g. \- probability scores, persona set clusters, attributes, etc.)

If you’ve got those five pieces, you’re ready to build. And if you're not sure where to find some of these, don't stress—your Faraday point of contact can walk you through it. 

## Step 2: Set up your custom step in LeadConduit

Now that Faraday is ready to respond, let’s wire it into your LeadConduit flow. This step is where identity data flows in, predictions come out, and your routing logic can make smarter decisions in real time.

Inside LeadConduit, go to the flow you want to edit and add a custom integration step. Here’s how to configure it:

* Format: JSON  
* Method: `POST`  
* URL: `https://api.faraday.ai/v1/targets/<YOUR_TARGET_ID>/lookup`  
* Headers:  
  * `Content-Type: application/json`  
  * `Authorization: Bearer <YOUR_API_KEY>`

[![New Integration](/images/blog/how-to-set-up-activeprospect/new-integration.png)](/images/blog/how-to-set-up-activeprospect/new-integration.png)

[![Add Faraday Integration](/images/blog/how-to-set-up-activeprospect/select-fdy-integration.png)](/images/blog/how-to-set-up-activeprospect/select-fdy-integration.png)

### Field mappings

Next, you’ll need to map your LeadConduit identity fields (all as JSON properties) to Faraday’s inputs. Here’s what you’ll want to map.

* `person_first_name`  
* `person_last_name`  
* `house_number_and_street`  
* `city`  
* `state`  
* `postcode`  
* `email`  
* `phone`

The more identifiers you include, the better the match rate. A good rule of thumb: aim to send at least email or phone, plus name and address when possible. 

Once that’s set up, Faraday will respond with predictive scores in real time. But here’s the part that matters: what you do with that score.

[![Add Faraday Integration](/images/blog/how-to-set-up-activeprospect/field-mapping.png)](/images/blog/how-to-set-up-activeprospect/field-mapping.png)

## What do you do with the score?

Once Faraday returns a predictive score (and any other fields you've configured), you can use LeadConduit’s flexible logic tools—like filters or branching steps—to take action based on that information.

💡 For example, many LeadConduit users set up filters based on returned fields from third-party integrations. According to ActiveProspect’s [API documentation](https://activeprospect.redoc.ly), it's common to create conditions that gate or route leads depending on values returned from an external service.

Your team might choose to:

* Pause or reject leads below a certain score threshold  
* Route high-scoring leads directly to sales or a priority CRM path  
* Nurture mid-tier leads via a different flow  
* Append persona data for downstream personalization

Just note: exactly how you act on the data depends on how your LeadConduit flow is configured. You’ll want to coordinate with your ops team or your ActiveProspect contact to ensure the field names and logic steps match your goals.

## Understanding the data exchange

👀 Visual learner? Here’s a simplified view of how data moves through this integration:

[![Diagram of leadflow through Faraday and LeadConduit](/images/blog/integrate-leadsconduit-blog/lead-conduit-diagram.png)](/images/blog/integrate-leadsconduit-blog/lead-conduit-diagram.png)

### Input requirements

You don’t need to send everything, but certain combinations are more effective:

* Email (or SHA256-hashed)  
* Full address (street \+ city \+ state)  
* Phone \+ last name

The more data you include, the more accurate your results.

### Output fields you can expect

When Faraday processes the identity, it returns structured output to your flow. These include:

* The identity information submitted  
* Outcomes like `percentile` and `probability`  
* Persona details like `cluster_id` and `cluster_name`  
* Traits based on your use case (custom model attributes)  
* Cohort memberships  
* Metadata like `match_type` and `error`

Note: `match_type` and `error` are static key names (they don’t change). The others depend on your pipeline config and what you’ve set up with your Faraday team.

## Example request & response

An example call:

[![An example call](/images/blog/integrate-leadsconduit-blog/example-call.png)](/images/blog/integrate-leadsconduit-blog/example-call.png)

An example response:

[![An example response](/images/blog/integrate-leadsconduit-blog/example-response.png)](/images/blog/integrate-leadsconduit-blog/example-response.png)

## How this supports your lead acquisition strategy

This isn’t just a tech integration—it’s a strategy upgrade. With Faraday connected to LeadConduit, you’re applying predictive intelligence the moment a lead enters your flow.

That means:

* Lower acquisition costs, since poor-fit leads are rejected earlier  
* Better sales efficiency, with lead prioritization based on conversion likelihood  
* More personalized outreach, using appended persona and trait data

You’re making smarter decisions earlier—and giving your team a cleaner, more focused pipeline to work with.

## Final thoughts

This setup gives your lead flow an intelligence upgrade. Whether you're filtering out poor-fit leads, scoring for immediate action, or enriching records for downstream personalization, Faraday’s integration with LeadConduit upgrades your entire lead flow—from first contact to final conversion.

Have questions or want help configuring your deployment? Reach out—we’re happy to walk through it together.

Integrate Faraday’s real-time predictive lead scoring into your LeadConduit flow in under 15 minutes—so you can filter, route, and enrich leads based on high-propensity signals before they hit your CRM.

Integrating Faraday and LeadConduit - a technical guide

You’ve got data—most likely lots of it. And if you’re the type of person who ends up on a Faraday blog, you’ve probably already considered the value predictive modeling could add to your brand. But not all predictions are created equally.

## When does your data require enrichment?

For some tasks, modeling on first-party data alone—like transactions or web events tied to known users—is enough. This works well for optimizing things like cart abandonment campaigns, existing customer retention, or upsells within your CRM. If you determine that’s all you need at your company, Faraday can absolutely model on just your first-party data.

However, first-party-only modeling comes with a few major limitations:

* It can’t help you reach new customers who haven’t engaged with you yet  
* It struggles with cold start problems when launching new products or campaigns  
* It’s limited by the scale and diversity of your existing audience  
* It can miss broader behavioral patterns visible only in third-party data

If your goals include reaching new audiences or launching into unknown territory, you’ll quickly find the cracks in a first-party-only strategy. That’s where Faraday’s data advantage comes into play.

We don’t just model on your existing data—we help you *expand* it. Through identity resolution, we enrich your records with insights from the Faraday Identity Graph (FIG), a privacy-safe dataset covering over 240 million U.S. adults. This allows us to bridge the gap between what you already know and the wider world of high-propensity customers you haven’t met yet.

Enrichment lets us go beyond who a person is, and start modeling what they’re likely to do.

But attentive readers will have noticed a word that might stand out: “**enriched**.” No, we’re not talking about flour, uranium, or even beer. In this context, enrichment refers to our ability to match the identities in your dataset to real individuals or households in FIG—so we can enhance your records with high-value predictive traits.

[![Image of "That's definitely enriched beer," meme](/images/blog/enriched-beer.gif)](/images/blog/enriched-beer.gif)


## So what is identity matching, anyway?

Identity matching is the process of taking the first party records you give us—names, emails, phone numbers, addresses—and matching them to real people and places in FIG. It’s how we go from “some person” to “this person, at this residence, with these behavioral signals.”

Faraday supports two levels of matches:

* **Person-level match**: We recognize the individual based on identifiers like name \+ address, email, or phone, and enrich with person-level traits (e.g. age, gender, purchase propensity) and property-level traits (e.g. home value, square footage).  
* **Residence-level match**: We recognize the household address, but not the individual. In this case, we can still provide property-level traits and aggregate averages for that address.

## How the Faraday Identity Graph fits in

As noted, at the core of this process is FIG—our proprietary dataset, which we’ve been building over the last 12 years. This set covers thousands of behavioral, demographic, and psychographic attributes on over 240 million U.S. adults and their households. When we get a match, FIG is what powers the predictive features we feed into your model.

The inclusion of this set—and third-party modeling more broadly—solves some critical problems that first-party data alone can’t address:

* **Cold start**: FIG helps predict outcomes even when you have little to no historical data on a segment  
* **Audience expansion**: You’re not limited to your CRM—you can model and target high-propensity lookalikes across the U.S.  
* **Signal density**: FIG adds new, behaviorally-relevant features that enrich your model’s inputs  
* **Long-term accuracy**: More diverse and representative data leads to more generalizable predictions

No match? No enrichment. And that means fewer signals to work with.

## Why match rate matters

Match rate—the percentage of records we’re able to enrich—is a direct lever on how well Faraday can perform. It governs:

* The **accuracy** of your predictive model  
* The **reach** of your campaigns  
* The **value** of our insights to your team

Here’s what typical match rates look like, depending on the identifiers in your dataset:

| Input type | Typical match rate |
| ----- | ----- |
| Name \+ address | 60–80% |
| Email only | 30–50% |
| Phone only | 5–10% |

If your list is heavy on phones or anonymized emails, that’s likely dragging things down. And if you’re seeing \<50% match overall, it’s time to investigate.

## Troubleshooting a low match rate?

One of the most common issues we encounter, and accordingly one of the biggest impediments we face when it comes to a good model is a low match rate. But sometimes diagnosing an issue isn’t as easy as you might expect. So here are a few common culprits to look for:

* Sparse or malformed identity data   
* Hard-to-match populations (e.g. younger adults, recent immigrants, underbanked)  
* Anonymized or junk emails (e.g. `email@email.com`, Apple relays, Doordash+ aliases)  
* Business names and commercial address (rather than personal records)  
* Missing names, flipped fields, or swapped first/last names  
* PO Boxes and apartment units with formatting issues

Sometimes it’s just formatting. Sometimes it’s about who’s in your list. But if your list looks like this, you probably have a guess where to start.

| First name | Last name | Email  | Phone number | Last outreach date |
| :---- | :---- | :---- | :---- | :---- |
| Rose, Benjamin | ben.rose@… | Faraday.ai  | 434 54🥕4 | Went great 👍 |

## What we do when match rate is low

When match rates dip below expectations, we kick off a diagnostic process. We look at the identifiers you provided, check for issues like junk emails or missing fields, and investigate how much of the dataset contains the high-value combo of name and physical address.

Depending on what we find, we might:

* Fix dataset mapping or formatting issues  
* Ask for additional identifiers (e.g. billing vs. shipping address)  
* Run a Match Boost through upstream vendors to attach new emails, phones, or addresses to help improve match

**Heads-up:** Match Boost isn’t always included in contracts, so we’ll confirm that before we proceed.

And if we’re getting solid residence-level matches but not person-level ones, it may just be that this population isn’t present in FIG. That’s a data coverage limitation, not a system failure.

## Why identity resolution is crucial for predictive modeling

Your predictive model learns from the signals in your data. The richer the signals, the more precise the model. If we can’t resolve identities, we can’t generate signals. And that means the model is essentially working with blind spots.

Match rate isn’t just a technical stat—it’s the foundation for how much value you can drive with AI. So if your match rate looks low, let’s dig in together. Because the better we know your audience, the better your predictions will be.

## Wrapping it all up

At the end of the day, your predictive model is only as strong as the signals it’s trained on. And assuming we’re not talking about limited, first-party only modeling, those signals depend entirely on whether we can recognize the people in your data—and enrich those records with meaningful traits.

That’s why identity resolution isn’t just a behind-the-scenes technical step. It’s the bridge between your raw data and the actionable insights that drive real outcomes. From expanding your audience and overcoming cold start challenges to increasing model precision and campaign ROI, enrichment through the Faraday Identity Graph is what unlocks the full power of predictive AI.

So if you’re seeing lower-than-expected match rates, or if you’re unsure whether your data is giving you the best possible results, let’s take a look together. Fixing a match rate issue might be as simple as a formatting tweak—or it might point to a strategic opportunity to boost your reach and improve your outcomes.

Because predictive modeling shouldn’t just be about what has worked before—it should be about what’s going to work next. And that starts with knowing who you’re modeling.

Your predictive model is only as strong as your data’s match rate—because if you can’t resolve who’s in your list, you can’t predict what they’ll do.

Identity matching makes or breaks your predictive strategy

You’re likely here because you want to get more out of your lead acquisition—and if you’re using [ActiveProspect’s](https://activeprospect.com/) [LeadConduit](https://activeprospect.com/leadconduit/), you’re already ahead of the game. But that is only the first step. Imagine combining that real-time lead filtering power with the [predictive intelligence and lead scoring](https://faraday.ai/templates) that Faraday has on offer. That’s where things get interesting.

This integration gives you the best of both worlds: Faraday’s ability to assess how likely a lead is to convert, and LeadConduit’s ability to make an instant decision on whether that lead is worth bidding on through their ping-post functionality. The result? Smarter decisions, faster processing, higher conversion rates, and optimal use of your marketing budget.

## What is LeadConduit?

ActiveProspect is a leader in optimizing lead acquisition, offering tools that help businesses validate and qualify leads in real-time. Their product, LeadConduit automates the decision-making process of bidding on affiliate leads, including rejecting leads that don’t meet your criteria. If you're buying or routing a lot of leads, this kind of automation is critical.

## So why integrate with Faraday?

While LeadConduit handles the immediate pass/fail/bid criteria, Faraday adds predictive depth to your decision-making through one of our [AI agents](https://faraday.ai/templates/lead-rejection). Our models assess how likely a lead is to convert based on how closely they resemble your customers. Instead of working from a fixed rule set, you’re working from the probability of a lead converting.

By connecting the two platforms, you can automatically filter out low-potential leads before you even bid on them. That means your conversion rates improve, your budget goes further, your sales team spends less time chasing dead ends, and your funnel stays clean and efficient.

## How this works for Faraday customers

When a new lead enters your system, LeadConduit sends that lead’s basic identifying information—things like name, email, phone number, and address—to Faraday for evaluation. From there, the [Lookup API](https://faraday.ai/docs/features/lookup-api) handles two core tasks: [identity matching](https://faraday.ai/blog/identity-resolution) and [predictive scoring](https://faraday.ai/blog/predictive-lead-scoring-b2c).

First, we match the lead to a real individual in our consumer data universe. The more identifiers you include, the more likely we are to find a high-confidence match. Even partial submissions—like just an email and zipcode—can often yield results, though additional fields like full address or phone help improve match accuracy and reduce false positives.

Once a match is made, we instantly return a payload of predictive insights. This can include a [conversion likelihood score](https://faraday.ai/templates/lead-prioritization), [persona classification](https://faraday.ai/templates/thematic-personalization), and any other configured outcomes you’ve built in your Faraday target. These models are customized to your business goals—whether that’s subscription signups, purchases, loan applications, or any other high-value conversion.

All of this happens in milliseconds, a requirement for operating within ping-post. As soon as the lead hits LeadConduit, the Lookup API evaluates it, and LeadConduit uses that response to decide whether to accept or reject the lead in real time. That means you’re not waiting for enrichment or reviewing leads after the fact—you’re getting a go/no-go signal before any budget is committed.

It’s like embedding a predictive intelligence checkpoint directly into your acquisition pipeline. Instead of just filtering on static criteria like age or location, you’re making decisions based on real behavioral and demographic alignment with your highest-value customers.

For additional details on inputs, outputs, and setup options, check out our [Lookup API documentation](https://faraday.ai/docs/features/lookup-api).

## Why this integration matters for your broader strategy

When Faraday and LeadConduit work together, you're not just filtering more effectively—you're building a smarter, faster, and more resilient lead acquisition system that adapts as you grow.

This integration powers:

* **More automation** – Less manual lead review, more consistency at scale  
* **Smarter prioritization** – High-potential leads move forward, low-value leads don’t  
* **Scalable efficiency** – Whether you process hundreds or thousands of leads, the system just works  
* **More efficient use of your marketing budget** – Stops wasted spend on bad leads.

Because it runs on live data and your custom models, it evolves alongside your business. As your audience evolves, your predictions evolve with them. It also adapts in real time to new lead sources, new personas, or entirely new markets.

Perhaps most importantly, this setup takes the grunt work off your team’s plate. With automated evaluation and rejection, your team can focus on what really matters: building relationships, closing deals, and driving growth.

## Ready to put your lead acquisition on autopilot?

Faraday \+ LeadConduit makes it easy to scale smarter, spend better, and stay focused on what actually drives revenue. If you're ready to see what predictive, real-time lead filtering can do for your funnel, let’s talk.

LeadConduit is an extremely powerful tool on its own, but when combined with Faraday’s predictive AI agents, it becomes a real-time decision engine that helps you prioritize high-value leads before you ever place a bid.

Supercharge your lead acquisition with Faraday + LeadConduit

At Faraday, we talk a lot about [Forward-Deployed Data Science (FDS)](https://faraday.ai/blog/FDS-changes-the-game)—our team of data scientists who work directly with clients to build predictive models, validate results, and ensure you get the most out of AI agents. This service (although new) has already proven invaluable to our clients. It means that they get not only accurate predictions, but comprehensive strategic plans to ensure that our agents actually drive business value.

But data science is only half of the equation. The best predictive AI agents also require another type of data specialist to operate at maximum efficiency: **Data Engineers.** 

And just like our data science team, Faraday’s data engineers don’t disappear behind the curtain—as is often the case at our competitors. Instead they work directly with clients, in a “forward-deployed” capacity. They’re the ones helping you format and structure your data correctly, making sure your onboarding process is as smooth as possible, and reducing or resolving errors  to ensure your pipeline stays live and healthy. If you’ve ever submitted a dataset and gotten feedback on how to improve it, or had someone help troubleshoot an ingestion issue, you’ve already worked with our data engineering team.

Take [Chris Thomas](https://www.linkedin.com/in/krys-tomis/), one of our incredible data engineers, [who coincidentally just wrote a whole blog about how to best avoid pitfalls while importing your data into Faraday](https://faraday.ai/blog/ingesting-data-pt2). His job—and the job of all our data engineers—is to make sure your data is clean, structured, and ready for predictive modeling. And in just the last weeks he’s helped numerous clients avoid pitfalls that could have set their timelines back by months—thanks Chris\!

So what’s the difference between data science and data engineering, and how do both help you get the most out of Faraday? Let’s break it down like Raygun\!

[![Raygun break dancing](/images/blog/raygun.gif)](/images/blog/raygun.gif)

### Data science: turning data into predictions

Forward-Deployed Data Scientists (FDS) focus on understanding your business goals and applying AI to help you achieve them. They understand the nuances and behaviors of the predictive AI agents that power our business, but also can provide strategic advice on the implementation of agents that ensures that your predictions drive value. 

Their job is to help you:

* **Define and refine your use case**—what do you want to predict? How can AI drive better decisions for your team?
* **Data discovery and solicitation**–what data is needed for the use case.
* **Build predictive models** that identify key audiences, behaviors, and trends.
* **Validate model performance** to ensure accuracy and impact before deployment.
* **Interpret and apply insights** so you can confidently take action based on data-driven predictions.

For example, if you’re a solar company looking to expand into new markets, an FDS won’t just give you a model—they’ll help you determine the right questions to ask in the first place. How do we ensure that the data used for training isn’t biased by work in existing markets? What is the logical cut off for differentiating promising new markets from a likely low performing market? Which of our [powerful AI agents](https://faraday.ai/templates) can increase your conversion rates the fastest? By framing the problem strategically, they ensure that your predictions are aligned with real business objectives—helping you make informed decisions that drive sales, optimize marketing spend, and accelerate growth.

### Data engineering: optimizing your data and models

Data Engineers on the other hand focus on the infrastructure that makes data science possible, while working directly with you to ensure your data is in top shape. 

Their job is to help you:

* **Structure your data correctly** before ingestion so it matches well against the [Faraday Identity Graph](https://faraday.ai/features/faraday-identity-graph) (FIG).  

* **Troubleshoot data issues** so you don’t lose time on formatting errors.  

* **Automate through properly configured integration** so your data flows seamlessly into and out of Faraday without constant manual updates.

Where data science can be viewed as more strategic and high-level—focusing on where you want to go, why, and how it will affect your business—data engineering is tactical, focused on how to get there efficiently. Engineers ask questions like, “How can we best implement this idea?” or “What changes will make this process more efficient?” 

At Faraday, a Forward-Deployed Engineer (FDE) like Chris undertakes tasks like supporting integrations with your data warehouse, reviewing your data during ingress, flagging potential issues, and suggesting optimizations to improve matching and model performance. 

### At Faraday, you benefit from both types of data experts

We don’t offer both Forward-Deployed Data Science and Forward-Deployed Data Engineering just because it sounds nice (although doesn’t it?)—we do it because it’s the fastest way to get actionable insights into our clients’ hands.

Making data engineering a client-facing function is a huge step that moves the process forward, ensuring that costly holdups and mistakes don’t happen before your predictive models even have a chance to work and that they keep working at peak efficiency after deployment. 

Here’s what that means for you:

🤔 **No more guessing if your data is good enough.** FDEs provide hands-on support before and during ingestion, catching issues early so they don’t slow you down later. After deployment they’re the ones making sure your models stay healthy.

⏳ **Faster time to value.** Instead of back-and-forth troubleshooting that delays model deployment, our FDEs work with you upfront to get data right the first time.

📊 **Stronger, more accurate predictions.** The best AI models in the world can’t fix bad data. By optimizing your data at the start, we ensure your models perform at their peak.

💰 **Fewer costly mistakes.** If your data isn’t structured correctly, you could end up with mismatched records, poor model performance, or wasted budget. FDEs help prevent that before it happens.

🤝 **A seamless experience.** Whether you’re working with an FDS to interpret predictions or an FDE to streamline integrations, you have direct support from an expert who understands your business.

In short, data science helps you make smarter decisions, and data engineering ensures your data is ready to support them. While forward-deploying both roles isn’t yet standard practice in the industry, we believe transparency, explainability, and speed to deployment aren’t just nice to have—they’re essential. And we’re not in the habit of waiting for the industry to catch up.

Our clients see the impact firsthand. Numerous businesses have leveraged our forward-deployed approach to streamline their data, improve model performance, and drive real results. You can explore a few of their [success stories here](https://faraday.ai/stories).

If you’re already working with Faraday, you know what this means for your business and bottom line. And if you’re not—shouldn’t you be?

Let’s chat about how we can help you put your data to work.

Faraday’s forward-deployed data scientists and engineers work directly with clients to ensure clean, structured data and precise predictive models—maximizing accuracy, efficiency, and business impact.

Data science + data engineering: How Faraday’s forward-deployed team fulfills all your data needs

Before committing to any predictive modeling platform, you need to know if their services actually work for you. That’s where a [validation test](https://faraday.ai/blog/cross-validation-technical-blog) comes in. Here at Faraday, a validation test (such as a holdout or cross validation) lets you see how well our AI agents can predict real outcomes—before you even sign up. By using your historical data, we can prove the accuracy of our agents and models without ever seeing your future results. Essentially, if the model is predictive, that means we're a good fit for helping you decide who to send direct mail to, how to accept and organize leads with lead scoring, and more. It’s the best way to validate AI-driven predictions—before you invest, and without any risk.

## How does it work?

A holdout test works by splitting your historical data into two sets: **one we see** (pulled from your recent conversion data, this is known as the test set) and **one we don’t** (pulled from the **most** recent conversion data, this is known as the holdout). First, we use only the training set to build a predictive agent. Once it is built, we generate predictions and compare them against the conversions that were ‘held out.’ If our agent accurately predicts those hidden outcomes, you can be confident in its ability to drive real results.

Cross-validation works similarly, but instead of holding out a single set for testing, it splits the data multiple times into different training and validation sets. Each subset takes a turn as the validation set while the rest are used for training, ensuring a more comprehensive assessment of model performance.

If you’d like a more detailed description of how this process works, take a look at this [blog](https://faraday.ai/blog/holdout-testing-with-faraday) from our COO, Robin Spencer, or for more detailed setup instructions, see this [blog](https://faraday.ai/blog/holdout-testing-evaluate-prediction-accuracy) from our CEO, Andy Rossmeissl.

Validation testing gives you a real-world view of the model’s accuracy, providing the reassurance that our predictions will stand the test of time—by making sure the AI isn't just memorizing past data, but identifying trends that matter going forward. But for the test to work, we need the right data—uploaded securely. Here’s what you need to know to get results as quickly as possible.

## What data do you need?

To set up a validation test, you must provide **names and dates** for both **leads and conversions**.

### Leads

We need to know when someone became a lead. Use the [Leads template](https://faraday.ai/data_solicitation/leads_template.xlsx) to provide this information, making sure to include the date for each lead. A leads file should ideally include the following fields:

* **First name**  
* **Last name**  
* **Address** (Street, City, State, Zip)  
* **Optional:** Email, Email Hash, Phone  
* **Date** (e.g., the date a lead was added to your system or when they received a mailer)

These personally identifiable details (PII) allow us to match leads to our demographic dataset. Without PII, we won’t be able to validate predictions, and the validation test cannot be run.

🚧 **Note on date formatting:** You can format the date in any standard way (e.g., YYYYMMDD, MM/DD/YYYY) and our system will automatically handle it.

### Conversion event data (transactions)

We also need the dates of individual transactions—along with optional details like product type and amount. Use the [Orders/Purchases/Transactions template](https://faraday.ai/data_solicitation/orders_purchases_transactions_template.xlsx) to submit this data. The conversion file should have the same structure as the leads file (first name, last name, address, etc.), but instead of lead dates, it should contain conversion dates.

### Conversions (customer-level data)

Customer-level conversion data (e.g., just the first or last transaction date) **is not ideal** because we need each individual transaction date. Although you can provide customer-level data using the [Customers template](https://faraday.ai/data_solicitation/customers_template.xlsx), we typically do not use it for a validation test.

## Uploading your data securely

To keep your data safe, always upload it through the **Faraday dashboard** in the Datasets section. You can also connect your platform to your existing data warehouse, but as this integration is more involved, this step would likely come after your successful validation test has been completed and you’ve decided to sign up with Faraday\!

🚧 **Do not email us any data.** The Faraday dashboard is the only secure place to upload files.

### Steps to upload:

1. Go to [app.faraday.ai](https://app.faraday.ai) and sign in.  

[![Sign in page](/images/blog/are-you-ready-holdout/sign-up.png)](/images/blog/are-you-ready-holdout/sign-up.png)

2. If you see a test mode warning, disable it before uploading your data.  

[![Make sure to leave test mode](/images/blog/are-you-ready-holdout/leave-test-mode.png)](/images/blog/are-you-ready-holdout/leave-test-mode.png)

3. Navigate to the **Datasets** section. 

[![Navigate to the Datasets section](/images/blog/are-you-ready-holdout/datasets.png)](/images/blog/are-you-ready-holdout/datasets.png)

4. Click **New Dataset** and select **CSV**. 

[![Select new dataset](/images/blog/are-you-ready-holdout/new-dataset.png)](/images/blog/are-you-ready-holdout/new-dataset.png)

5. Upload your **leads.csv** file and your **conversions.csv** file.

[![Choose CSV upload](/images/blog/are-you-ready-holdout/upload-csv.png)](/images/blog/are-you-ready-holdout/upload-csv.png)

## Want a little more help?

If you're still unsure how to set up your validation test, we've got you covered. [Watch this video walkthrough](https://www.loom.com/share/ab3750fdb50d47c2b6a0fc05e96e10dd?sid=aefe17c6-ba5f-458f-986f-73db3cf43877) with Faraday CTO Seamus to see the process step by step.

And if you still have questions, reach out to us at **support@faraday.ai**—we're happy to help\!

And that’s it\! Once uploaded, we’ll take care of the rest. Ready to get started? Upload your data today and see how Faraday’s AI agents can drive real results for your business.

Validate Faraday’s predictive accuracy risk-free with a holdout test—just upload your data, and we’ll prove our models work before you commit.

Getting started with Faraday