[Klaviyo](https://www.klaviyo.com/) made a big splash last year, bringing entry-[level predictive scoring](https://faraday.ai/blog/why-you-need-predictive-lead-scoring-today) to their massive audience. And we love it—the industry is catching up to what we’ve been doing for years. With AI-powered tools like [predictive analytics](https://faraday.ai/templates/lead-prioritization), [personalized product recommendations](https://faraday.ai/templates/next-best-offer), and advanced [sentiment analysis](https://faraday.ai/blog/seamus-bigquery-interview) for their emails lists, Klaviyo is giving marketers powerful capabilities to forecast customer behavior, personalize messaging, and automate campaigns. But if you’re looking to take it to the next level, there’s a critical missing piece: [deep consumer insights powered](https://faraday.ai/features/faraday-identity-graph) by [third-party data](https://faraday.ai/docs/abstractions/traits).

## Klaviyo’s predictive tools, and their limitations

[Klaviyo’s AI tools](https://www.klaviyo.com/features/ai) offer a solid foundation for email marketers, helping to predict engagement and automate messaging. But recent changes—including increasing costs and [stricter policies around active contacts and list sizes](https://help.klaviyo.com/hc/en-us/articles/24263920096027)—are making efficiency more important than ever.

Though Klaviyo’s AI is powerful, their analysis is also stymied by its lack of inclusion of third-party data. This type of rich demographic and behavioral data inherently improve the depth, nuance, and critically, speed-to-deployment of predictive models. It also radically improves the accuracy and value that they can be expected to produce.

Only with this type of rich third-party data can marketers expect to achieve the highest levels of  speed and precision with their predictive marketing, and that’s where Faraday’s built-in integrations comes in.

## Optimizing Klaviyo lists with Faraday personas

Integrating Faraday’s [personas](https://faraday.ai/docs/abstractions/persona_sets) and [customer segmentation](https://faraday.ai/docs/how-to/connections/segment) capabilities into your existing Klaviyo stack takes your email lists to the next level, allowing you to gain deep insights that Klaviyo alone can’t provide.

With Faraday, you can segment your email lists not just by basic engagement metrics but by [demographics, life events, property details, and other behavioral signals](https://faraday.ai/features/faraday-identity-graph). This level of insight empowers you to tailor content, messaging, and campaigns that are perfectly suited to your audience. Whether it’s understanding the impact of a new home purchase or identifying consumers in a particular life stage, Faraday’s data-driven approach helps you reach the right people at the right time with the most relevant messaging.

By adding these third-party insights to your Klaviyo platform, you unlock smarter segmentation, faster insights, and a deeper understanding of your customers—making your email marketing more efficient, relevant, and impactful.

### The inclusion of this this data improves your Klaviyo lists by:

* **Building smarter segments** based on real-world demographics, life stages, and property data.  
* **Refining automated journeys** to align with consumers’ actual behaviors and preferences.  
* **Improve conversion rates** by sending the right message to the right people at the right time.

## The cold start problem—and how to overcome it

An additional benefit of integrating third-party data into your Klaviyo stack is the ability to avoid the [cold start problem](https://www.sciencedirect.com/topics/computer-science/cold-start-problem), a classic challenge faced by many marketing tools, including Klaviyo’s out of the box predictive scoring. While these tools are able to generate generally accurate predictions on long time customers using only first-party data, they quickly run into issues when targeting leads or recently acquired customers with very little observed behavior. 

In these cases, the model struggles to quickly generate meaningful insights, as it lacks the foundation of prior data to build from. This delay in data accumulation can significantly limit the effectiveness of campaigns until enough data has been collected.

Faraday inclusion of our third-party data eliminates this barrier by enabling you to build highly targeted customer segments right from the start. Instead of waiting months for data to accumulate, Faraday’s built-in data provides immediate insights, empowering you to craft campaigns optimized for engagement and conversion from day one. Without waiting for slow data accumulation, you can launch campaigns quickly with confidence.

## So who will actually convert? Let’s find out.

Your Klaviyo lists might still be growing, but are they optimized for conversion? Are you maximizing their value-generating potential? With Faraday, you can build a free persona model and see for yourself. Think of all the ways you could use this—better segmentation, more precise messaging, and automated journeys that actually resonate.

Klaviyo is a great start for any email marketing, but Klaviyo \+ Faraday together can take your email marketing to the next level\!

Enhance your Klaviyo lists with Faraday’s third-party data and personas to unlock deeper insights, smarter segmentation, and higher conversion rates—solving the cold start problem and optimizing your email marketing from day one.

Leveling up your Klaviyo lists with Faraday personas

If you are reading this, you are probably talking to Faraday about a potential contract and have a healthy amount of skepticism as to whether we can actually do what we say we are going to do.

And here’s the thing: you’re right to be skeptical. With tight budgets and increasing pressure to show ROI, you can't afford to bet on unproven solutions or spend countless cycles figuring out if a tool is really going to live up to the promise. You need a way to separate legitimate predictive tools from the noise.

Every week, we get to talk to dozens of CMOs, CTOs, & CEOs across industries who are trying to discern the same thing: which tools are legitimate and which ones are not. As a leader of a high profile brand, you are inundated with promises of predictive insights and transformation, but you need more than just promises – you need guarantees.

That's where our holdout evaluation comes in. It's not a proprietary methodology we invented – it's an industry-standard approach used by data scientists to validate predictive models.

**What makes us different is simple: we run this validation before you sign any contract.** Because frankly, if we can't prove we'll deliver real results for your business, we don't want to waste your time.

Below we’ve outlined what we are doing “behind the scenes” and how you can evaluate if we will add the value you need. 

## What is a holdout evaluation?

A holdout evaluation is a method for validating predictive models by comparing their predictions to real-world outcomes. We **set aside a portion of recent conversion data**—data the model has never seen before—and use it as an independent benchmark, against which predictions can be compared.

This “holdout set” acts as a real-world proving ground. It is untouched until after the predictive model has been trained. 

Meanwhile, the rest of your historical data is combined with rich third-party insights from the [Faraday Identity Graph](https://faraday.ai/docs/abstractions/traits) ([which includes 1,500 attributes on 240M+ U.S. consumers and their households](https://faraday.ai/features/faraday-identity-graph)). We use this combined data to train the new predictive model.

Training a predictive model simply means finding patterns in conversion behavior. In other words, what can we learn from your previous customers to identify future customer behavior? Once trained, the model generates predictions—scoring leads based on their likelihood to convert and only after these predictions are made do we compare them to the **real conversion rates** in the holdout set.

If the highest-scoring leads actually convert at higher rates than lower-scoring ones, the model is validated as accurate and reliable. 

Holdout evaluations ensure predictive models aren’t just theoretical—they work in the real world, using real data, to deliver real results.

## How holdout validation works at Faraday

We've streamlined this validation process to be transparent and straightforward:

1. ![customer conversion data](/images/blog/holdout_testing_diag/holdout_testing_diag-01.png)

   #### Prepare your data and set aside a holdout group

   We start with your recent [lead conversion data](https://faraday.ai/blog/first-party-vs-third-party-data-marketing). Before any processing occurs, we reserve a portion of your most recent leads as a "holdout set"—data that the model will never see during training. This ensures we have an independent benchmark for evaluating predictions.

2. ![merged with FIG](/images/blog/holdout_testing_diag/holdout_testing_diag-02.png)

   #### Combine data with Faraday Identity Graph

   With the holdout set reserved, we then combine remaining data with Faraday’s [built-in, rich  consumer data](https://faraday.ai/features/faraday-identity-graph) sourced from the Faraday Identity Graph. This expanded dataset provides a deeper view into key predictive signals, like property and demographic information, life events, buying habits, and more.

3. ![model is trained and predictions begin](/images/blog/holdout_testing_diag/holdout_testing_diag-03.png)

   #### Train the model and begin predictions

   Using this enriched dataset (excluding the holdout set), we train a predictive model to identify patterns in conversion behavior—learning from traits, behaviors, and other signals.

   Once the model is trained, it scores each lead based on its likelihood to convert, generating a ranked list of opportunities—from high-value leads to those less likely to convert.

4. ![model is validated](/images/blog/holdout_testing_diag/holdout_testing_diag-04.png)

   #### Compare predictions to reality to validate model

   Finally, we put the model to the test: we compare its predictions against the real outcomes in that untouched holdout set. If the highest-scoring leads don't convert at significantly higher rates, we go back to the drawing board.


## How to know if a holdout validation was successful

A holdout validation is considered successful when the following criteria are met:

1. **Clear correlation between predicted probability and actual conversion rates**  
   The model’s predicted likelihood to convert should directly align with the actual conversion rates. If the predictions are accurate, high-probability leads should convert at a noticeably higher rate than lower-probability ones.

2. **High-scoring leads convert at significantly higher rates**  
   The top-scoring leads should consistently perform better than those ranked lower. This demonstrates that the model is identifying truly valuable leads.

3. **Consistent performance across different customer segments**  
   The model should show reliable results across various segments—whether you’re targeting different demographics, regions, or other customer profiles. If the model works well across a broad set of data, it proves its reliability and scalability.

When a model meets these criteria, it’s ready for deployment—whether that means optimizing [lead buying](https://faraday.ai/docs/how-to/use-cases/leads), refining your [direct mail strategy](https://faraday.ai/templates/lead-suppression), or enhancing your email [personalization](https://faraday.ai/templates/personalization) approach.

## Questions you should ask us

We want you to push us on what matters most for your business. Here are key questions to evaluate if we're the right fit:

1. **How does this model's performance compare to the lift you are seeing with  similar businesses in our industry?**

   While we can’t share client specifics, we can provide insight into broader trends and expected performance benchmarks.

   → **Why this matters:** Understanding industry benchmarks helps you gauge whether predictive modeling is truly a competitive advantage for your business.

2. **How much historical data do we need for reliable predictions?**

   The answer varies by business model and conversion volume. We’ll help determine whether you have enough data—or how to supplement it if needed.

   → **Why this matters:** Predictive models work best when trained on strong data. Knowing the threshold for reliability ensures you're making data-driven decisions with confidence.

3. **What happens if the model doesn't perform as expected?**

   We have a structured iteration process to refine models when necessary, ensuring they reach meaningful predictive power.

   → **Why this matters:** No model is perfect on the first run. Understanding the process for continuous improvement ensures long-term success.

4. **How will this integrate with our existing marketing stack?**

   We’ll provide technical specs, discuss implementation timelines, and outline the resources needed to get up and running.

   → **Why this matters:** Seamless integration means faster time-to-value and less friction in your workflow.


## Why holdout evaluations work

The businesses who win this year will be those who successfully incorporate predictive modeling into their core workflows. It's no longer a nice-to-have – it's a need-to-have. But we know you can't afford to make decisions based on hunches or unproven tools. That's why we're eager to validate our value up front. 

At the end of the day, three things matter to us:

1. **Confidence before commitment:**  
   You see proof before you commit

2. **Higher marketing efficiency:** 
   Your marketing dollars go to the leads most likely to convert

3. **Stronger sales performance:** 
   Your sales team focuses on opportunities worth their time

If we can't demonstrate value, we'll be the first to tell you we're not the right fit. But if we can – and we usually do – we want you to know the exact savings or additional revenue you can expect to see over the course of the next year with Faraday. Interesting in generating value you can trust? Reach out to learn more about what we can do for your business\!

Faraday uses holdout testing to validate predictive models before you commit, ensuring they deliver real, measurable results.

Predictive modeling generates tangible value, and we can prove it

A [ping post](https://activeprospect.com/glossary/ping-post/) is a real time lead distribution method where lead sellers send basic lead details (the "**ping**") to potential buyers, who then evaluate the lead based on predefined criteria. If a buyer decides to bid on the lead, they receive the full details (the "**post**") and complete the purchase. This system allows for fast, competitive bidding, commonly used in industries like home services, financial services, and insurance.

While this method offers efficiency, it also introduces challenges. Many businesses end up paying for leads that don’t convert, leading to wasted budget and ineffective outreach efforts. The ability to filter and score leads before purchasing them is crucial to maintaining a high return on investment (ROI) and optimizing marketing spend.

## How we use Ping Posts at Faraday

The challenge with ping-post lead buying is that companies often spend money on leads that don’t convert. At Faraday, we integrate with ActiveProspect’s [LeadConduit](https://activeprospect.com/leadconduit/) to enhance this process with predictive scoring. By applying machine learning at the ping stage, brands can bid only on high-converting prospects, filtering out low-quality leads in real time. This optimization reduces wasted spend and improves overall campaign efficiency.

Predictive scoring works by analyzing historical data, behavioral signals, and demographic attributes to assess a lead’s likelihood of converting. Thanks to the built-in  [Faraday Identity Graph](https://app.faraday.ai/traits) ([which comes preloaded with 1500+ attributes on nearly every US consumer and their households](https://faraday.ai/features/faraday-identity-graph)), our models score each lead against rich third-party data before a bid is placed instead of relying on limited seller-provided information. This creates the opportunity for businesses to make data-driven decisions in real time, ensuring they focus only on high-quality opportunities.

For example, an insurance provider purchasing thousands of leads daily might bid on any lead that meets basic criteria, but many of those leads never convert—driving up acquisition costs. If each lead costs $10 and converts at a rate of 1 in 5, their effective customer acquisition cost (CAC) would be $50. But not all leads are equal. By rejecting the bottom 40%—the least likely to convert—they could improve their conversion rate significantly, lowering their CAC to just $30.

In this scenario, instead of wasting budget on low-quality leads, they could acquire more customers for the same spend. With predictive lead scoring, they could make these decisions in real time, ensuring they only invest in leads with real potential.

## Even without TCPA 1:1 consent rule, better targeting still matters

[While the expected TCPA 1:1 consent rule did not go into effect earlier this year](https://faraday.ai/blog/tcpa_1-1_vacated), that doesn’t mean businesses can ignore lead quality or continue to enact spray and pray targeting in 2025\. Just because you’re not being forced to change doesn’t mean you shouldn’t. Rising acquisition costs and increasing competition make it critical to invest in leads that are both high-quality and likely to convert. Faraday’s real-time predictive scoring ensures that brands maximize their ROI while staying ahead of the curve. By leveraging AI-driven insights, lead buyers can refine their targeting strategies and remain competitive in an evolving marketplace.

Additionally, compliance concerns aren’t going away. Even without the immediate pressure of TCPA amendments, consumer privacy laws and regulations continue to evolve, and businesses that proactively refine their lead acquisition strategies will be better prepared for future changes. More rigorous lead assessment doesn’t just improve efficiency—it also reduces the risk of compliance issues and enhances the overall customer experience by ensuring outreach efforts are more relevant and targeted.

## Looking ahead

The lead generation landscape will continue to evolve, and companies that prioritize precision over volume will have the advantage. Rather than buying as many leads as possible and hoping some might convert, businesses need to shift toward smarter, data-driven strategies that optimize every dollar spent.

Faraday’s predictive lead scoring is a powerful tool for staying ahead, allowing businesses to purchase leads with confidence and efficiency. Whether regulatory shifts happen or not, the need for better targeting and smarter decision-making isn’t going away.

Ping posts are a real time lead distribution method where basic lead details are sent to potential buyers, who then decide whether to bid before receiving the full lead information.

So what is a Ping Post anyway?

Using machine learning to identify customer personas in your data can help improve your understanding of your customer base, and with the right infrastructure, deliver personalized experiences at a greater scale.

In a nutshell, the process entails using an algorithm to segment your customer data into distinct groups, or personas. Faraday uses a time-tested method called k-means clustering.

## What is k-means clustering?

K-means clustering is an unsupervised clustering algorithm that was first introduced in 1957. An unsupervised algorithm does not require the data to be labeled in order to train the model. This is an important feature of the k-means algorithm because it allows for the discovery of subgroups within the data without any previous assumptions about possible groupings — in this case, personas.

## How does k-means clustering work?

When presented with a data set, the k-means algorithm first introduces random center points around which each data point will eventually be clustered (circles, Fig. 1). Then each data point is assigned to the group of the center to which they are closest.
This happens by calculating the distance from a data point to each cluster center using an equation known as the Euclidean distance measure. For two dimensions, the Euclidean distance is represented by the diagonal line connecting the point to each of the cluster centers. The point is then assigned to the cluster center with the shortest diagonal line (Fig. 2).

Once each point has been assigned to the randomly chosen cluster centers, the centers then "update," moving to the center of the group by recalculating the mean (average) distances to all the points in the group (Fig. 3).

After the cluster centers have updated to their new positions, each point is reassigned to a new group using the Euclidean distance measure again. Most of the points will remain in their original groups, but some points will be assigned to a new group (Fig. 4). So, then the centers must be updated again to be at the center of the new group. This pattern repeats until the centers no longer move.

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

## Using k-means clustering for customer personas

All of the above is how k-means clustering worked back in 1957. Since then there have been upgrades allowing us to use something called categorical features. At Faraday, we create persona groups using a mix of these features.

As far as algorithms go, k-means is rather straightforward, but has one major downfall: you have to tell it how many groups to look for, and it will always succeed in finding that many groups. That means if you tell the algorithm to look for 500 groups, it will find them — even if there is no reason for there to be 500 distinct groups within the data.

If you're a marketer, you probably don't want to create 500 unique ads for every campaign. But should you make four? Or six? Or ten?

## How to find the optimal number of personas using k-means clustering

We target the optimal number of groups hidden within the data, rather than guessing and hoping we are right. We’ll share a few methods we use to find the optimal number of personas for our users.

### The elbow method: finding diminishing returns

The most common metric for measuring the success of k-means clustering is called the ratio of sum of squares (rss). At Faraday, we determine the optimal number of personas by comparing the rss for k-means models with two to fourteen personas. We employ the elbow method, which shows the cutoff after which we see diminishing returns for increasing the number of personas. For most clients, the optimal number of distinct personas is usually between four and seven.

**Going beyond the elbow method**

The elbow method is simple, but it has a downside: sometimes, there just isn't a clear single point that is optimal. Maybe three personas could be optimal, maybe six, and we have to intervene based on our client’s business needs to choose one or the other as the optimal number of personas.

Additionally, while reducing variability within each group is an extremely useful way to measure k-means clustering success, it doesn't tell the whole story. We believe that an important aspect of clustered groups is stability, which isn't considered with the elbow method selection process. Our clients expect that their persona groups won't shift every time they convert a lead. This is why we are currently researching a new method for determining the optimal number of personas, which focuses on the stability of the persona group.

### The bootstrap method: focusing on stability

Our first step to assess the stability of k-means clustering involves resampling the client’s data using the bootstrap method. Like k-means clustering, bootstrapping has been around for a while – it was first [introduced in 1979 by Bradley Efron](https://projecteuclid.org/journals/annals-of-statistics/volume-7/issue-1/Bootstrap-Methods-Another-Look-at-the-Jackknife/10.1214/aos/1176344552.full). The basic idea behind bootstrapping is to take data from the pool we have, with the ability to pick the same data point multiple times. This is a method known as sampling with replacement.

The bootstrap method provides us with a powerful tool to measure how much our desired clustering outcome could change with different data. We can get a good sense of how much the personas will change with the addition of new customers. Our goal is to minimize how much influence single customers, or small groups of customers, have on the resulting personas. We want personas that remain relatively unchanged by single customers. In other words, we want more stability.

## Pulling it all together

Once we have the results from bootstrapping, we use the ratio of sum of squares again to measure just how stable the personas are. What we’re looking for is a ratio close to 1, which would indicate that there is no variability within the personas.

The stability metric for determining the optimal number of personas shows promising results for the basic examples in this blog post. At Faraday, we’re actively working on applying this methodology to the more complex persona groupings our clients rely on. Stay tuned for more developments with respect to improving our personas methodology.

Learn how machine learning is used to identify distinct clusters—or personas—in your customer data.

How k-means clustering is used to identify customer personas

Faraday has been pioneering the use of [Google Cloud's](https://cloud.google.com/gcp?utm_source=google&utm_medium=cpc&utm_campaign=na-US-all-en-dr-bkws-all-all-trial-e-dr-1710134&utm_content=text-ad-none-any-DEV_c-CRE_665735450627-ADGP_Hybrid+%7C+BKWS+-+EXA+%7C+Txt-Core-Google+Cloud-KWID_43700081237254438-kwd-6458750523&utm_term=KW_google%20cloud-ST_google+cloud&gad_source=1&gclid=CjwKCAiA2JG9BhAuEiwAH_zf3ipLZUgcmqUgvAG4mSazGgzYHHuc-Sxw_1kXaAED5OkUnKhoJ5VNGxoCrRIQAvD_BwE&gclsrc=aw.ds) BigQuery and with its recent addition of LLMs, we’ve unlocked even more powerful use-cases. In this video, Seamus dives into how LLMs are improving sentiment analysis, content generation, and streamlining our data processes—all within [BigQuery](https://cloud.google.com/bigquery?utm_source=google&utm_medium=cpc&utm_campaign=na-US-all-en-dr-bkws-all-all-trial-e-dr-1710134&utm_content=text-ad-none-any-DEV_c-CRE_724346015211-ADGP_Hybrid+%7C+BKWS+-+MIX+%7C+Txt-Data+Analytics-BigQuery-KWID_43700081227767773-kwd-47616965283&utm_term=KW_bigquery-ST_bigquery&gad_source=1&gclid=CjwKCAiA2JG9BhAuEiwAH_zf3rR-UNnTVF20r8oRT-f2T4EOwmKb8fymYRf1cOrh2cj3ZH1M7JjJThoCkDYQAvD_BwE&gclsrc=aw.ds), without needing complex infrastructure.

**In this video, you’ll learn:**
- How Faraday improves sentiment analysis accuracy with BigQuery LLM feature
- The benefits of generating personalized content directly within BigQuery
- How BigQuery’s LLM feature simplifies our workflows and data processes

<iframe width="560" height="315" src="https://www.youtube.com/embed/luRR3cyMqrM?si=IIbOHJLK5nwU6jaO" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>

Join Faraday’s CTO Seamus Abshere as he shares how we’re using Google Cloud’s BigQuery to transform our workflows with the new Large Language Model (LLM) capabilities.

How Faraday simplifies sentiment analysis and generates personalized content using BigQuery

## Difference between first and third party data

Nowadays, the most successful brands are data-driven organizations. For a long time, brands harnessed both first and third party data to strengthen their understanding of their customers, improve decision-making, create more personalized experiences, and more. Let’s zoom out: what’s the difference between first-party data and third-party data?

**First party data** is data that your brand collects on your customers and leads. Page visits, form submissions, purchases, and email engagement are all examples of first party data, and each one provides unique insights and context into how your customers and leads interact with your brand. Depending on how and where your brand collects data, first party data can come in a variety of forms: engagement, demographic, transactional, financial, lifestyle, and more.

**Third party data** is data that’s been collected by a third party, or not your brand. It comes in the form of demographics such as age or gender, psychographics such as hobbies and interests, and financial data such as household income. Third party data providers like Acxiom allow brands to purchase the data for use in marketing.

Until recently, third-party cookies and social scraping were common sources for data enrichment meaning that browsers would creepily follow consumers across the internet to track their interests, behaviors, and other personal information that was never willingly shared for commercial use. It enabled personalization at the expense of consumer privacy, but responsible brands have been preparing for the [death of third-party cookies](/blog/death-of-third-party-cookies-strategies).

## Benefits of first and third party data for consumer marketing

All types of data have their own individual merits when it comes to consumer marketing, and in order to create a cohesive, impactful [customer insight strategy](/blog/how-to-create-a-customer-insight-strategy), you need more than just your first party engagement data, otherwise you run the risk of sending the wrong message at the wrong time. A good combination of first and third party data plays a key role in two pillars of a strong marketing strategy: customer insights and predictions.

### Customer insights

A strong foundation of data enables marketers to segment customers and leads into distinct subgroups so that they can be targeted with relevant engagement, and with both first and third party data fueling your customer analytics tools, you’ll find that your customers and leads are more than just purchases and clickstreams–they’re real people, with real interactions outside of your brand.

By unifying first and third party data, your customer insights upgrade from “person that abandoned their cart” to “John Doe, married father of two, golf enthusiast, cat owner, and brick-and-mortar shopper.” And the best part is that this can be done at a scale that applies to entire segments of your customer base. The patterns are there in your data, you just need to find them.

### Predicting customer behavior

The value of pure customer insights through data is clear, but you can take your strategy a step further by using that data to [predict virtually any customer behavior](/blog/propensity-models-predict-customer-behavior) your brand is interested in. Likelihood to churn, convert, make a purchase, purchase again–whatever strategy you want to reinforce or employ, customer predictions using propensity models can get you there quickly and accurately.

Propensity models pore over your customer and lead data to find historical patterns, and then use those patterns to make the customer predictions you need. Plus, by using both first and third party data, your predictive models become even more accurate because they have more data to work with. Being reactive just doesn’t cut it anymore, and by predicting customer behavior, you can be proactive to ensure that lead converts, or that customer decides to stay on board.

## First party data collection techniques

Perhaps the most important benefit of first party data is that your brand is in complete control of how you collect it. Some strategies are more effective than others, but it’s important to implement a variety of strategies to ensure healthy engagement with both customers and leads.

### Web forms & surveys

Web forms can be a great way to capture basic information like name, email, mailing address, and possibly some financial or lifestyle information as well. In a similar vein, albeit more resource intensive, surveys are a fantastic way to poll your customers and leads to learn more about what they’re looking for, what they like about your brand (or even what they don’t!). Consider offering a small incentive, such as a free coffee, for any who fill out a survey.

### Loyalty programs

One of the best ways to engage with and retain your current customers is to implement a loyalty program. Loyalty programs enable brands to consistently engage with their customers through features like allowing customers to rack up points per purchase that can be redeemed for discounts or freebies. On top of the rewards you can offer your customers, it gives you another avenue through which you can engage with them–like sending automated, personalized emails when they reach a certain point or purchase threshold.

### Online engagement

Users will engage with your website in different ways, and the more data you collect, the more you’ll understand what it is that your visitors are looking for. Digital ads on a platform like Instagram can help you find the platform that fits your ideal customers. Visitor behavior on your site, such as scroll depth or visits on a certain page can similarly help you find what pages resonate most with them, which you can then use to guide engagement copy & creative.

## First party data enrichment

Modern brands are adopting a first party data enrichment strategy through third party data providers to gain a true understanding of who their customers are so that every engagement is a delight. It’s possible to both respect your customers’ privacy and use third party data to personalize engagement–you just need to be smart about it. Faraday helps brands do just that.

With cookies in the rear-view window, it’s more important than ever to ensure you’re being [responsible](/features/responsible-ai) and privacy-centric with any data you use. Avoid data that involves shady practices like social scraping, and look for [opt-in, permissioned data](/features/data)–it’s the way of the future, and not utilizing it is leaving your brand at a disadvantage.

So what are your options for first party data enrichment? The obvious answer is licensing directly from third-party data vendors like Acxiom or Epsilon. Something to note, however, is that you’ll need to sift through all that data to find what’s most relevant to your brand at any given time–your ideal customer profile will likely change over time, after all. On the other hand, many brands are adopting customer prediction platforms like Faraday so that they can access all the third party data they need, with machine learning algorithms that tell them what’s most important, saving time & resources that are better put elsewhere, like launching campaigns.

## How to leverage first and third party data with Faraday

Faraday enables your brand to generate actionable insights and customer predictions using both first and third party data–or first party data enrichment. No matter what your stack looks like, Faraday has [hundreds of integrations](/integrations) to ensure your predictions are where you need them, when you need them, and without a single line of code. On the flip side, if your brand is developer-minded, Faraday offers a [robust developer API](/docs/reference) so that you can implement predictions your own way.

### Predict personas to make personalization a breeze through rich insights

Like I mentioned earlier, first party data enrichment completely transforms what you know about your customers, and allows you to visualize clear indicators for who’s represented in any segment of people your brand is interested in. Turn that newly-discovered John Doe, married father of two, golf enthusiast, cat owner, and brick-and-mortar shopper–part of the Middle-aged Married Mike persona–into internal talk track alignment, guided personalization on any platform, and more.

[![Faraday customer personas](/images/blog/first_party_vs_third_party/first-party-vs-third-party-personas.png)](/images/blog/first_party_vs_third_party/first-party-vs-third-party-personas.png)

### Predict propensity to take virtually any action

With your personalization in-hand, use propensity models to predict who’s (or where’s) most [likely to buy](/predict/category/likely_buyers) your product, who’s worth [retaining](/predict/category/churn_scores), how to [prioritize your leads](/predict/category/lead_scores), & more. Customer predictions with machine learning turn your reactive marketing proactive, so that you’re always delivering the right message at the right time, and personalized to boot.

[![Faraday propensity to buy outcome](/images/blog/first_party_vs_third_party/first-party-vs-third-party-outcome.png)](/images/blog/first_party_vs_third_party/first-party-vs-third-party-outcome.png)

**Ready to try it out? [Create a free account](https://faraday.ai/signup).**

An effective marketing strategy is data-driven, and to create one, you need a strong mix of first and third party data. Learn the difference between first and third party data in marketing, and how smart brands are using first party data enrichment to drive growth.

First party vs third party data: What's the difference?

[Propensity predictions](https://faraday.ai/features/predictions) indicate the likelihood of an individual to perform a defined action (buy, buy again, churn, etc.), which enable businesses to grow and retain revenue by engaging with the right customers, leads, or audiences at the right time. To get accurate propensity predictions, you need a good propensity model. Here’s how Faraday approaches propensity modeling.

## Selecting the right propensity model

Our prediction system classifies against outcomes using an array of predictive algorithms, including individual decision trees, random decision forests, logistic regressions, and neural networks. At run time, we are attempting, in each case, to figure out which algorithm works best. Usually it’s random decision forests.

## How random decision forests work

Random decision forests are made up of individual decision trees, which are classifier algorithms that look like flow charts, showing the choices made to reach a certain outcome.

Predictions aren’t based on a single tree, though; we use dozens of trees to improve the accuracy of the algorithm, hence the random decision forest. We look at the universe of trees, their splits and branches, and choose the ones that are the most promising when it comes to creating a path towards the desired outcome. This path leads from seemingly arbitrary data to a classification: the “yes” or “no” to the outcome. In other words, we're defining someone as a “good potential customer.” (In the image above, the “yes” is “mostly converted” and the “no” is the “mostly did not convert.”)

While the classification seems binary, the reality is that often there's not a clear “yes” or “no.” We are dealing with the likelihood that someone performs an action, and that is never definitively 100% or 0%. During the prediction process, we employ propensity scores, which are a way of saying, “This person is more likely to be a good potential customer than not.” Every individual in the group we’re looking at is scored, and those in the top percentiles are recommended to the client as the best fits for the outcome they want to realize.

_Now, if there are so many predictive algorithms we can use, why do we choose random decision forests?_

## Benefits of predicting propensity using random decision forests

### Building explainable models

Well, for one, they’re the most explainable; with so many decision trees at play, we have a strong sense of the different importances at each node split in a tree. These importances show us which attributes contribute most toward the success of an outcome, or the most toward failure.

### Handling missing data

The random decision forest technique also handles missing data very well, especially compared to a logistic regression or neural network, which both tend to require values to make predictions. In a decision tree, if there is a missing value, it moves onto the next node because there are other decision paths in the case of missing data. When you scale that to a whole random decision forest, the prediction becomes very accurate and isn’t perturbed by missing data.

### Identifying collinearity

A third reason we use random decision forests is collinearity. Random decision forests may see that some data are correlated — say, mortgage value and income. A lot of the attributes the algorithm deals with are linearly related but don’t necessarily directly drive a certain customer behavior.

A random decision forest algorithm finds areas of greatest information gain, meaning that it may realize there is interdependence between attributes, but it will choose the feature that gives the algorithm the greatest leverage over the prediction. This is particularly helpful for us, as we deal with tremendous amounts of customer data, and data on humans is generally noisy and often subverts expected distributions and assumptions.

## How propensity predictions help improve marketing efforts

Propensity predictions are especially helpful in improving campaign targeting efforts across a number of marketing channels. By identifying individuals ranked in the top percentile of the analyzed group—those that have the highest likelihood of performing the desired action—businesses typically see improvements in efficiency metrics (cost per acquisition, return on ad spend, etc.). Here are some popular business use cases for propensity predictions:

- Reaching likely buyers
- Driving repeat purchases
- Scoring and prioritizing leads

Learn how propensity models help predict the likelihood of individuals to exhibit a defined behavior.

How propensity models help predict customer behavior

If you're in marketing or sales, you've probably already heard of the [Telephone Consumer Protection Act (TCPA)](https://www.twilio.com/docs/glossary/what-is-telephone-consumer-protection-act-tcpa), which regulates how businesses can communicate with consumers through telemarketing calls and texts. You also probably knew that in 2023, they introduced a new [1:1 Consent rule](https://docs.fcc.gov/public/attachments/DOC-408396A1.pdf), which aimed to close the [lead generator loophole](https://www.orrick.com/en/Insights/2023/12/FCC-Closes-TCPA-Lead-Generator-Loophole-Requires-One-to-One-Consent?utm), tightening consent requirements and limiting how consumer phone numbers are distributed. This would have required businesses to solicit individual consent from each consumer they planned to contact by phone.


## So, what happened to the rule?

Although originally scheduled to take effect today (January 27, 2025), at the last minute, a postponement was announced, pushing the effective date to January 2026\. However, just hours after this delay, the [Eleventh Circuit Court vacated the rule entirely](https://natlawreview.com/article/eleventh-circuit-vacates-tcpa-one-one-consent-rule-immediately-after-fcc-postpones), eliminating its implementation altogether. So for now, the existing TCPA consent requirements remain unchanged.

**With the TCPA 1:1 Consent rule:**

For companies that rely on purchasing large lists of leads to grow, this rule would have required a significant shift in strategy. By curbing the "spray and pray" approach of casting a wide net with unqualified prospects, businesses would need to pivot toward precision targeting and engagement with smaller, more qualified audiences. While this might initially seem like a challenge, it offers a tremendous opportunity to refine marketing campaigns and focus resources on higher-value prospects.

The benefits of this shift would be multifold. Businesses would see an increase in ROI from more intentional and personalized campaigns, as every interaction would be designed to resonate with the recipient. Consumer trust would also improve, as individuals would have a clearer understanding of who is contacting them and why. The era of vague, broad-based consent would give way to meaningful, 1:1 engagement that prioritizes quality over quantity.

This approach would not only reduce waste but also empower companies to connect with the right consumers at the right time. While these changes might seem daunting, they would pave the way for more effective, trust-driven marketing—a strategy that businesses could turn into a competitive advantage with the right tools and expertise.

**Without the TCPA 1:1 Consent rule:**

We’re back to the world of massive, unqualified lists. Without the additional compliance requirements, businesses can continue to cast a wide net in their telemarketing efforts. However, this approach risks higher opt-out rates, lower engagement, and potential reputational harm if messages are perceived as spam. 

## You don’t need to wait for the government to tell you try smarter targeting, here’s how Faraday can help today

Whether the 1:1 Consent rule returns or not, targeting remains a critical factor in marketing success. At Faraday, we prioritize quality over quantity, helping businesses focus on the most qualified prospects instead of chasing unqualified leads. With our predictive tools, companies can:

- **Target the right consumers**: Faraday can analyze market trends and [customer data](https://faraday.ai/features/faraday-identity-graph) to identify the [consumer segments](https://faraday.ai/templates/customer-targeting) most likely to show interest in engagement. This ensures companies can focus their efforts on earning explicit consent from likely-to-convert individuals.  
- **Use the right tools**: With fewer leads available, focusing on the right prospects is more crucial than ever. For instance, instead of focusing on [lead rejection](https://faraday.ai/templates/lead-rejection), since the lowest common denominator of leads won’t be there already, use that budget instead on tools like [lead routing](https://faraday.ai/blog/how-to-use-ai-to-deliver-a-better-quality-pipeline) to make sure the new leads make it to the ideal sales rep and [adaptive discounting](https://faraday.ai/templates/adaptive-discounting) to pique their interest.  
- **Maximize value:** After you win your best-fit customers, leverage the various predictive tools available through Faraday like [churn scores](https://faraday.ai/blog/churn-prevention-machine-learning) and [product recommendation](https://faraday.ai/templates/next-best-offer) to ensure that you maximize their lifetime value.

Regulatory changes may come and go, but Faraday’s data-driven approach ensures that businesses stay ahead, delivering effective and efficient marketing strategies no matter the landscape.

## Need help navigating this change?

The world is full of unexpected outcomes, like the 1:1 Consent rule getting canceled hours before it went into effect\! At Faraday, we’re ready to help you adjust your strategy to target the highest-quality leads and stay ahead of industry trends regardless of how regulations change. Reach out today to ensure your lead generation stays ahead of the curve\!

The TCPA's new 1:1 rule would have required individualized consent to contact a lead by phone but was vacated at the last minute.

TCPA 1:1 Consent rule vacated, but smarter targeting with Faraday is already here

Restaurants love Toast because it focuses on _their_ problems:

<a href="https://pos.toasttab.com/blog/il-primo-pizzeria" target="_blank"><img src="/images/blog/new-imperative-for-vertical-saas-built-in-ai/toast-pizzeria.png" width="372px" /></a>

Toast is a VSaaS (vertical software-as-a-service) - software **purpose built for unique industry workflows**. This means faster time to value compared to generic "horizontal" solutions. For example, Toast has a dedicated "Pizza POS" system with [easy topping swaps!](https://pos.toasttab.com/restaurant-pos/pizza-pos) A pizzeria wouldn't want to reinvent that wheel - but they also wouldn't want to pick a POS system that doesn't support it.

For the roofing industry, you _could_ customize Salesforce to track "Roof Area" - or you can use a [roofing CRM like Acculynx](https://acculynx.com/). There is a wonderful explosion of examples like this - companies that build exactly what their clients need, because they ONLY build for those clients:

- [bloomerang](https://bloomerang.co/) - for non-profits
- [mindbody](https://www.mindbodyonline.com/) - for wellness centers and spas
- [ALINE](https://alineops.com/) - for senior living
- [archy](https://www.archy.com/) - for dental offices
- and hundreds more.

With growth comes competition, and the question is: what's next for Vertical SaaS?

I would argue it is **incorporating AI features** that are specific to their industries. For VSaaS companies that haven't yet incorporated an AI feature, or are [looking to upgrade beyond first party data-powered features](/blog/why-first-party-data-and-rfm-is-not-enough), there are many options on the market. Here, we organize them into 3 levels - purpose-built, data warehouse-native, and low-level API. Each is useful for different use cases and organizations.

## Traditional focus on control points

Best-in-class VSaaS players have been focusing on [capturing "control points"](https://www.tidemarkcap.com/vskp-chapter/control-points-2-0). The control points are "back office" and "front office." Back office is the system of record, and VSaaS often starts by trying to become this - the "last system you turn off". Acculynx tracks roofing jobs and leads, while another Xero, an accounting VSaaS, [keeps the books](https://www.xero.com/us/security/data-protection/). From there, they can focus on integrating with or directly providing front-office technology - whatever the employee or customer interacts with. Every adult in the United States has probably held a [Toast POS](https://pos.toasttab.com/hardware/toast-go) in their hand.

## The new imperative - built-in AI features

What's next for VSaaS? How can you extend further into the value chain - or compete with an incumbent who has captured a control point? How can you [avoid ChatGPT-based extinction](https://www.tidemarkcap.com/post/ai-extinction-or-evolution)? One possibility is adding AI features specifically tuned for an industry, which only VSaaS can provide.

There are at least 2 reasons VSaaS is well positioned for AI.

- **Data**. As a system of record, a VSaaS might be the _only_ place a business can get the data it needs to make a critical decision or prediction.
- **Semantic layer**. Since VSaaS understands the semantics of its vertical - in other words, it knows what pizza toppings or roof sizes _mean_ - it can automatically build and use AI models.

As [Tidemark](https://www.tidemarkcap.com) puts it, **"flip your product from being reactive to proactive."** For example, Acculynx rates every roof lead 🟢 🟡 🔴 and [doubled conversion rates for clients who focused on green-light leads](https://www.prweb.com/releases/2023/4/prweb19276673.htm).

## Taking the first step

Assuming building AI into their product is critical path, VSaaS vendors must decide where to start. It can be helpful to think of the options as organized into levels from most out-of-the-box to lowest level. Different VSaaS companies will find different levels to be the most appropriate.

![](/images/blog/new-imperative-for-vertical-saas-built-in-ai/level-advantages.png)

### Level 2 - easiest - purpose built AI

The most accessible way for VSaaS to incorporate AI into its products is to adopt an existing lower-level component whose unit of prediction is the same - one that is purpose built for predicting against people, products, or search. For example, any system that deals with people can build predictions about their behavior by using Faraday's [business outcomes](https://faraday.ai/docs/product/outcomes). A system in charge of inventory can add intuitive AI-based search by incorporating Bloomreach's [product search](https://documentation.bloomreach.com/discovery/reference/product-search-category-api). Systems for managing creative assets can build next generation functionality by incorporating AI features like Cloudinary's [automated image editing and generative background replacement](https://cloudinary.com/products/cloudinary_ai).

### Level 1 - data warehouse native

Going a level down, VSaaS looking to develop its own AI capacity outside of what's already on the market in level 2 should evaluate the machine learning components [built into data warehouses](https://cloud.google.com/bigquery/docs/bqml-introduction). These are not just for one-off analyses; entire applications can be developed on top of them. This level sits at the sweet spot between specific units of prediction in level 2 and raw HTTP APIs in level 0. Generally controlled by SQL commands, they treat structured, unstructured, and even [file/object data](https://cloud.google.com/bigquery/docs/annotate-image) as tables and allow feature engineering, modeling, evaluation, and prediction as SQL queries. They are designed to process data at any scale, automatically shuffling data and managing memory without requiring manual partitioning or configuration by users. So, if you build a solution on top of them, you get scaling essentially built in.

Almost all data warehouses now allow you to do [generative AI](https://faraday.ai/blog/quickstart-use-bard-in-bigquery). This is much more attractive for batch use cases than creating error-prone HTTP connections to LLM APIs across many rows of data (you wouldn't use this for building chatbots that require low latency - see level 0). Don't forget about predictive use cases too - [time series modeling](https://towardsdatascience.com/time-series-forecasting-in-snowflake-using-sql-6ce7cedce862), for example, is available on all of the major data warehouses and allows sophisticated projections with built-in and custom holidays. [SQL-based classification and regression](https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-boosted-tree) using XGBOOST and random forest are bread-and-butter use cases that fit perfectly with tabular input and output data; once you've done feature engineering with standard SQL functions, it's hard to imagine a simpler way. [Unsupervised clustering](https://docs.aws.amazon.com/redshift/latest/dg/tutorial_k-means_clustering.html) is also available.

### Level 0 - raw API

A very rich set of services, from Google's [Vertex AI](https://cloud.google.com/vertex-ai) to [OpenAI's Enterprise-friendly models](https://openai.com/enterprise), exist as HTTP APIs. This is the lowest common denominator and is accessible from all systems. It is also particularly appropriate for just-in-time calculations such as generative prompts used by chatbots or online fraud predictions using random forest models mounted alongside [feature stores](https://cloud.google.com/vertex-ai/docs/featurestore) for instant response. When planning an AI feature for a VSaaS, it might seem necessary to master an entirely new API. However, low-level systems are often better suited for specific real-time use cases, while asynchronous or batch workflows should first explore the potential of level 2 or level 1 solutions.

## Choosing a level

A few questions can suggest where to start:

- Is there already an out-of-the-box prediction infrastructure for your use case? Companies should always evaluate level 2 purpose-built solutions before committing to building low-level systems.
- What is the time between a prediction being created and its use? For example, if you're incorporating search, people expect results immediately; using an existing level 2 vendor may be the fastest approach. However, if no solution aligns with your industry needs, you may need to skip Level 1 and develop a [level 0 solution](https://cloud.google.com/vertex-ai/docs/predictions/get-predictions). Conversely, if you plan to rank a lot of records in asynchronous batch use cases, a level 1 solution is likely sufficient.
- Is there industry data that can extend a client's own data? For example, fraud detection vendors incorporate geographic fraud data in their scores. Faraday has a [national dataset of all US consumers](https://faraday.ai/features/data) that automatically enriches predictions made on its platform.
- What kind of scrutiny will your AI receive? For example, if you are making predictions about people, consider how you will implement [Responsible AI](https://www.brookings.edu/events/what-is-responsible-ai/). Some level 2 vendors [automatically build in reportable controls and metrics](https://faraday.ai/features/responsible-ai); some level 0 vendors [let you build your own](https://www.datarobot.com/blog/bias-mitigation-with-datarobot/).

## The imperative

If you were starting a VSaaS product from scratch, what would be your wedge? Some form of AI is a good guess. And if you are a VSaaS incumbent, how do you defend your position? You have an advantage in the data and semantic layer you already hold. A thoughtful approach to AI infrastructure options can boost your efforts and minimize your time to value.

Vertical SaaS (VSaaS) is software built exactly for a particular industry. The next generation of VSaaS is incorporating exactly the AI that people need.

A new imperative in vertical SaaS: built-in AI features

Companies are looking for the [fastest way to build AI into their software](/blog/new-imperative-for-vertical-saas-built-in-ai). Using techniques like RFM ([recency-frequency-monetary](<https://en.wikipedia.org/wiki/RFM_(market_research)>)), many are implementing AI features that learn from how people use websites, respond to emails, and pay for products. This process is called "first party data-based modeling". Companies like [Twilio](https://www.twilio.com/blog/first-party-data-marketing) and [Klaviyo](https://www.klaviyo.com/blog/predicting-churn-risk-our-new-model) have used this simple approach to rapidly add features to existing large products.

At the same time, increasing privacy regulation (did you know [Colorado](https://leg.colorado.gov/bills/sb21-190) and [Virginia](https://law.lis.virginia.gov/vacodefull/title59.1/chapter53/) have their own versions of [CCPA](https://oag.ca.gov/privacy/ccpa)?) and the [retirement of "third party cookies"](https://developer.chrome.com/en/docs/privacy-sandbox/third-party-cookie-phase-out/) have made companies [think "first party data only" is the only option left](https://digitalmarketinginstitute.com/blog/why-first-party-data-is-more-important-than-ever-and-how-to-use-it). It turns out this is not quite right - the very same transaction and customer records that make up first party data can be linked to [traditional, legal, consented offline data](https://www.acxiom.co.uk/blog/understand-your-online-and-offline-data-to-build-defined-data-strategy/) to create more powerful technology. Over time, companies that start with RFM [upgrade to using offline data too](https://www.postpilot.com/lookalike-audiences).

# Problem

Why is first party data not enough? **The total information contained in a user's activity may be insufficient to predict their future behavior** - the model may be weak. Over time, repeat customers _may_ build up enough data to be useful. But new users will always be a mystery. And once a first party model is built, it must be constantly monitored for new patterns in relatively noisy data. In other words, building AI features into software but limiting them to first party data makes them weaker, less widely applicable, and less stable.

## Activity vs affinity

First party activity data is traditionally thought of in terms of "recency" (time stamps), "frequency" (aggregation over time), and "monetary" (revenue impact). Product choice can be added to this (this person bought a pillow!) Does this contain sufficient data to predict their next purchase or how much they will spend over their lifetime? Often, it does not. It would be more predictive to know, in addition to RFM, their [lifestyle preferences and financial resources](https://faraday.ai/features/data). Activity data is a relatively incomplete view into a person's overall affinity for a product.

## Cold start problem

The first time a person shows up at your business, you don't know anything about them - so you can't make any first party-based decisions about them. This is called the "cold start problem" with first party data. Over time, you will know more about their purchase behavior, activity on your website, opens/responses to emails, etc., and RFM models will become more powerful (though still not as powerful as the same model with lifestyle and financial data mixed in). However, if a software company's clients need to make predictions at first contact, first party data will not help.

## Noisy data

Activity data is noisy. Many random factors can affect when a user opens a browser, comes to your site, stops, buys a product, or opens an email. First party-based models must be able to adapt quickly to the same changing environment. In some industries, like fraud detection, the ability to update models in real time to address ever-changing threat actors is a key feature. In other industries, the need or capacity to constantly refresh models based on noisy data it not necessarily positive. For example, a model that uses lifestyle or financial aspects to predict next best product is likely to exhibit more stable performance over time, especially if it incorporates (but does not rely solely on) first product purchased and time spent on site.

# Solution

What possible solutions are there? One is to use offline data - a supply chain that has existed for decades. Ever since sewing magazines realized their mailing lists offered valuable insight into people's hobbies, they have been asking their readers for permission to share their data with marketing partners. In terms of regulation, this data is consented - unlike web scraping or even some forms of first party data. Sewing is only one of the many niches with dedicated providers who can derive affinity data from purchases. And every major data aggregator - who has relationships with all the niche providers - has a vested interest in defending its continued use and teams dedicated to adapting to changing laws.

There are a few challenges to using offline data directly. The contracting process is lengthy and requires a very secure infrastructure to receive and store the data - not something all companies have (or even want). Also, there is a lot of data - hundreds of columns. It can take data science teams years just to understand what data is valuable to each market segment and build repeatable workflows to create models. Even if this is done, issues like proxy discrimination and bias propagation arise - and may disqualify the product's use in certain industries.

Are we back to square one - first party data only? Not necessarily. AI infrastructure companies like [Faraday](https://faraday.ai) that have offline data built-in are solving contracting, security, model discovery, and responsible AI concerns with standardized components. The most important benefit is time to value - software companies feeling the imperative to add AI can release it an order of magnitude faster.

## Contracting

Traditionally, buying offline data has been a multi-month contracting process with "what the market will bear" pricing. AI infrastructure companies like Faraday have standardized contracts and usage-based pricing, making it affordable to try new AI product initiatives without high upfront data costs.

## Security

Data security requirements for holding offline data are high. A data breach may release not only customers' order history but also more personal data. Having a SOC 2 that specifically audits for the protection of consumer data is a common requirement. AI infrastructure companies store offline data on behalf of other companies, letting them use it in batch and real time use cases without ever coming into possession of it.

## Legal and regulatory

Holding offline data always increases legal requirements. For example, if you hold more than first party transaction and activity data, responding to CCPA (and other) data access and data deletion requests becomes much more expensive. It also reclassifies a company as a data broker, forcing it to register in many states. AI infrastructure companies respond to such requests on behalf of users and are registered data brokers so that other companies don't have to be.

## Time to value

Data scientists need time to figure out what data matters, but data usually arrives in an unusable state. It must be cleaned and reformatted. Different techniques are then used to measure each data point's value to the business. Experimental models are built in Jupyter notebooks; finally, the whole process is handed over to a development team to make ready for production. AI infrastructure companies can accelerate this process, for example, by automatically matching first party transaction data to offline data and generating information gain metrics and models that can be backtested, field tested, and iterated on.

## Responsible AI

In addition to general public concern about malicious use of AI, specific industry regulators and trade groups have been adding requirements. Longstanding requirements - such as fair housing and fair lending - apply to AI just as much as traditional methods. Other requirements - such as model transparency and elimination of proxy discrimination - take on new forms when using automatically generated models. AI infrastructure companies can provide these as out-of-the-box reporting and mitigation tools instead of individual company data science or compliance teams implementing them on a one-off basis.

# Don't leave money on the table

First party data is not a magic bullet; offline data has a steep learning curve. AI infrastructure providers like Faraday, Twilio Segment, and Snowflake can help add AI to existing products faster and with less complexity, while also bypassing the limiting first party data-only stage.

1st party data-based AI modeling struggles with limitations like noisy data and the cold start problem. By integrating 3rd party data through providers like Faraday, companies can enhance predictions and accelerate AI deployment.

Tech