We get it, when you’re working quickly, it can be tempting to attach a file to an email and call it a day. But when the data can identify an individual human being, email is not the right channel.

At Faraday, data security is something we take extremely seriously ([just check out this blog](/blog/common-data-security-questions)) but it’s also a shared responsibility, and small process choices can make a big difference. That’s why we want to make sure all our customers, clients, and partners are on the same page about how we protect [Personally Identifiable Information](https://www.dol.gov/general/ppii) (PII). Accordingly, this is a quick guide to what counts as PII, why we don’t want it sent over email, and the secure ways to get that data into Faraday so you can get to the fun parts.

## What is PII

In the most basic sense, PII is any piece of data that can identify a specific person, either on its own or in combination with other fields.

Within Faraday, the easiest way to think about this is to look at the identity fields we use to recognize individuals and households during identity resolution. In dataset identity sets, the core fields you should treat as PII include name, location information, and contact details. These are the same building blocks we expect clients to map when they want strong match rates.

In practical terms, that means a file containing any of the following should not be emailed to us:

* First name and last name  
* Street address, city, state, or postcode when tied to a person or household  
* Email address  
* Phone number  
* Email hashes

This is also why “we removed the name” is not a safe workaround. A full address, phone number, or email can still identify an individual. Even less complete combinations can be enough to narrow someone down (ie there might only be one "Maureen" in zip code 05400).

If you want a more technical view of how these identifiers work together, our [datasets documentation outlines](/docs/abstractions/datasets) the identity resolution priority order and the standard identity fields we recommend mapping.

## What is usually ok to share

It is worth noting that some data can be sent through email. Aggregated data is often the distinguishing factor. For example, counts or summaries at a ZIP code or regional level are typically fine, as long as that data cannot be used to identify a specific individual. 

But when in doubt, treat anything that looks like a row-level customer record as PII. And if you’re unsure DON’T just send it. Reach out to support with questions when you have them before sending potentially sensitive information.

## How you should send data to Faraday

So we’ve established the rule: **do not email us files that contain PII**. 

But we still need those identifiers to resolve identities, append customer context, and build predictive models. The difference is that we need them coming through the secure paths designed for sensitive data.

To send PII to Faraday, use one of these options: 

* upload through the Faraday app  
* upload CSVs via your normal dataset workflows  
* use an established integration or connection  
* or upload programmatically through the methods you already use with Faraday. 

These options are built for secure transfer and handling of sensitive customer data. If you’re unsure which route is best for your setup, reach out to support and we’ll help you pick the right path.

## Why we care about PII

This is not about slowing anyone down. It’s about protecting your customers and keeping your data flows aligned with the security standards both of our teams expect.

Email is convenient, but it is not designed for handling sensitive, row-level customer identifiers. The secure upload paths in Faraday are. Using them protects your customers, reduces risk for your team, and helps keep our shared data workflows clean and consistent.

And this isn’t where our commitment to data security ends. Faraday is also SOC 2 audited, compliant with all federal and local regulations, and we treat secure data handling as an operational discipline. We store data with trusted cloud vendors, encrypt it in transit and at rest, and control access with measures like multifactor authentication.

All of this only works as intended when PII comes through the secure paths built into the platform, rather than unencrypted email.

## If you think you have an exception

If your current process can’t conform to these guidelines, reach out to support. We’ll help you find a secure path forward.

## In closing

At the end of the day, we’re here because we all care about the same thing. Protecting our customers’ private information.

If you take one thing from this post, make it this: PII should come to Faraday **through the platform**, not through *email*. That simple choice helps keep your team, your customers, and your data operations safer, while still allowing you to get the full value of customer context and predictive modeling.

A quick guide to what counts as PII, why you should never email it to Faraday, and the secure ways to upload sensitive customer identifiers into the platform.

Personally Identifiable Information (PII), and the right way to upload your customer data to Faraday

Every CMO feels it: growth only happens when you know exactly where your opportunity is. But consumer behavior keeps shifting. Acquisition costs keep rising. Your best regions last year aren’t guaranteed to be your best regions next year. And the pressure to deploy budget intelligently (across markets, personas, and channels) has never been higher. 

Most teams do not have a reliable way to see where their next customers actually are.

Today, market clarity isn’t optional. It’s the foundation of any GTM plan and you need the right tools to solve this problem.

## Market Opportunity Analysis (MOA) is the tool you need to get that clarity

Faraday’s [Market Opportunity Analysis](https://faraday.ai/docs/abstractions/MOA) is a modern, data-driven way to understand your market at a level your CRM or BI team can’t show you. Now, you can finally see:

* Where you’re overpenetrated and hitting diminishing returns  
* Where real growth potential remains across the U.S.  
* Which personas offer disproportionate upside  
* Which states, metros, and postcodes are quietly becoming your highest-value regions

It connects your customer data to the [Faraday Identity Graph](/features/faraday-identity-graph), our nationwide consumer dataset, which includes thousands of behavioral, demographic, financial, and predictive signals that never appear in your CRM. And because FIG represents the full U.S. population, it shows not just who you’ve already captured, but who you’re missing, and how large the opportunity truly is.

[![Three market opportunity analysis screenshots organized as header image](/images/blog/MOA-menu/moa-the-menu-blog-header-image.png)](/images/blog/MOA-menu/moa-the-menu-blog-header-image.png)

## Delivered in days, not months

Even better, Faraday can deliver completed MOA reports in days (assuming your data is structured). Marketing leaders working in 2025 know they do not have time for slow delivery or long analytical cycles. They need accurate intelligence, and they need it quickly. It’s as simple as this:

1. Upload your historical customer data. That is the only lift on your side.  
2. Faraday enriches and analyzes the dataset to generate a MOA analysis unique to your customers  
3. You receive a clear, executive-ready view of your total addressable market, penetration, and where your strongest opportunities live. 

This speed to delivery makes MOA an ideal planning tool, giving leaders the clarity they need right when decisions are being made.

## What Faraday’s MOA can reveal

Decision makers need a clear picture of what they will actually receive. Faraday’s MOA delivers three core categories of insight that mirror the way CMOs structure their planning: Market Size, Market Penetration, and Market Opportunity.

### 1. Total Addressable Market (TAM)

Every MOA starts with a clear view of your total addressable market across the United States. This shows how big your opportunity truly is and how much room you have to grow, broken down by state, metro, and postcode.  

[![MOA TAM analyis showing total addressable market](/images/blog/MOA-menu/moa-tam.png)](/images/blog/MOA-menu/moa-tam.png)

### 2. Penetration analysis

We map your existing customers against the wider population to show where you are saturated and where you still have headroom. This often surfaces high value regions where you have little or no presence.

[![MOA penetration analyis showing where current customers exist](/images/blog/MOA-menu/moa-penetration.png)](/images/blog/MOA-menu/moa-penetration.png)

### 3. Geographic opportunity (State, Metro, Postcode)

With TAM and penetration in place, the MOA highlights your strongest pockets of geographic opportunity at the state, metro, and postcode levels. This gives you a multi-layered view of where your next customers actually are.  

[![MOA geo analyis showing where current customers exist and where lookalikes could be found](/images/blog/MOA-menu/moa-geotargeting.png)](/images/blog/MOA-menu/moa-geotargeting.png)

### 4. Persona opportunity

The MOA identifies which personas represent the largest opportunity for growth and how those personas vary across regions. This helps inform segmentation, messaging, and channel strategy.  

[![MOA geo analyis showing where current customers exist and where lookalikes could be found](/images/blog/MOA-menu/moa-persona.png)](/images/blog/MOA-menu/moa-persona.png)

### 5. Actionable next steps

Faraday provides clear next steps based on the data, including where to allocate budget, which regions to prioritize, and which personas to focus on. These recommendations make the MOA immediately useful in planning.

### 6. Fully custom add-ons

For organizations with more complex questions, the MOA can include fully custom add-ons. These deepen the standard analysis and answer specific strategic questions through tailored cuts of the data. Examples include:

* Shopping behavior differences between retail and eCommerce customers  
* Demographic patterns such as age or gender distributions  
* Product-level performance broken out by persona  
* Store-level penetration and catchment insights

## Three ways to receive your MOA

Faraday offers three MOA delivery formats, each designed to meet teams at different levels of analytical maturity and strategic need.

### Option 1: Just the data

This is the most flexible format and the one preferred by teams with strong in-house analytics. You receive a standardized CSV output containing everything needed for deeper analysis: TAM estimates, persona distributions, and geographic breakdowns at the state, metro, and postcode levels. There are no visuals and no recommendations attached.   
Just clean, structured data ready for your team to explore however they choose.

### Option 2: Essential insights

If you want a fast, executive-ready readout, this is the most streamlined option. Faraday distills the MOA into a set of core visualizations paired with up to four recommended next steps. It gives you a clear, concise narrative of where your strongest opportunities exist without requiring any internal analysis. Many teams use this version directly in planning decks and leadership reviews.

### Option 3: Fully custom

For organizations with more complex questions or multiple business lines, the fully custom MOA provides a deeper layer of insight. Faraday builds tailored cohorts and segmentations to answer your specific strategic questions, whether that involves product-level differences, demographic splits, store-level penetration, or other custom cuts. You also receive up to two rounds of refinement as new questions emerge.

## Closing

Understanding your market shouldn’t require guesswork. Faraday’s Market Opportunity Analysis gives you the data, clarity, and confidence to know exactly where your next customers are, and where your strongest growth potential lives.

If you want clarity before 2026 kicks off, we’re here. Talk to us about running your MOA.

This post explains how Faraday’s Market Opportunity Analysis uses your customer data and the Faraday Identity Graph to reveal your true TAM, current penetration, and where to focus future growth.

Understand your market with Faraday’s Market Opportunity Analysis

For insurance and debt-resolution teams, the high-volume, parallel-dialing era is breaking under modern spam protections and compliance pressure. The old assumption that maximizing dials will eventually produce results is no longer reliable.

Parallel dialing, which made it easy to brute-force a list by firing multiple calls at once, is now a liability. Carrier spam systems are stricter, VoIP behavior is aggressively penalized, and consumers block unknown numbers faster than ever. The harder you dial, the faster your numbers get flagged as “Scam Likely.”

Because of this, many organizations are shifting toward [progressive dialing](https://www.convoso.com/blog/what-is-a-progressive-dialer/), a system where Agents work leads one at a time. This reduces abandoned calls and keeps compliance risk in check. That does not mean high-volume or parallel strategies never make sense. It does mean that, for most day-to-day calling, more intentional, progressive behavior is becoming the norm. But this shift introduces a new requirement: if each call is deliberate, every prospect needs a real chance of converting.

The thing is that progressive dialing doesn’t fix weak filtering; in fact, it makes those weaknesses impossible to ignore. The raw volume safety net is gone. Once volume goes away, the quality of your scoring becomes the deciding factor in whether your call center generates value or falls flat.

## Fewer dials, higher stakes

[![Diagram showing the different flows of leads through parallel or progressive dialers](/images/blog/progressive-dialing-diagram.png)](/images/blog/progressive-dialing-diagram.png)

Parallel dialing allowed teams to get results even with imprecise filtering, as the sheer volume of attempts eventually surfaced the best leads anyway. Progressive dialing changes that dynamic. When agents call leads one at a time, each conversation carries more value. Time spent on a poor-fit record is time not spent on a viable customer and this shift exposes a structural weakness in most systems: 

> The filtering logic in front of the dialer was built for a world where volume covers for imprecision. 

In most organizations, existing filters rely on outdated assumptions, demographic shortcuts, and scoring models that don’t reflect actual conversion outcomes.

This filtration layer often fails because it uses:

* Guesswork filters, like “higher income is better” or “certain ZIPs respond more.”  
* Manual rules added years ago that rarely get revisited.  
* Lead scores that rely only on demographic data alone without factoring in conversion outcomes.  
* A disconnect between real-time conversion data and day-to-day prioritization.

Under parallel dialing, this was “good enough” but with progressive dialing, that safety net disappears. 

> If your system can’t surface your most viable prospects early, agents waste their best hours and expensive leads get burned without meaningful lift.

## What better filtration looks like

So you might find yourself asking: what filters do I actually need for a financial-services call center?

The answer isn’t a single score or one magic rule. It is a handful of signals that give your tools and teams the context they need to decide which prospects are worth prioritizing, which deserve immediate outreach, and which should take a different path entirely.

Better filtration in insurance and debt-resolution usually comes down to a few core areas:

* **Quality at the moment of purchase:** Teams need a way to understand which leads are likely to engage or advance, and which should be down-bid or skipped altogether. This decision happens before anyone makes a call, and it depends entirely on having meaningful context at the point of purchase.  
* **Early engagement signals:** A significant amount of agent time is lost on people who are not in a position to meaningfully engage. Early behavioral or modeled signals, whether from form data or enrichment, give teams the context they need to focus agents on prospects who are more prepared for a productive conversation.  
* **Likelihood to convert:** Once a lead is in your system, the order you work it in matters. Teams rely on predictive context to decide which records to prioritize during peak windows so the highest-value prospects get attention when it is most likely to count.  
* **Matching prospects to the right path:** Not every lead should go straight to an agent. Context helps teams decide which prospects warrant immediate outreach, which should enter warm-up flows in email or SMS, and which should move into slower, lower-cost nurture until their behavior changes.  
   (The routing itself happens in the CRM, dialer, or marketing tools. Faraday provides the signals those systems rely on.)

When these decisions line up, the entire funnel tightens. Teams use their resources more effectively, agents spend more time in productive conversations, and performance improves without adding volume. The difference comes from having the right context in place to drive smarter decisions.

## Where Faraday fits in

For insurance and debt-resolution teams, Faraday regularly sits at the very top of the funnel. Many teams connect us through [LeadConduit](https://activeprospect.com/leadconduit/) where we score inbound leads in real time and help prioritize which records deserve immediate attention. Many also use our credit score proxy at this stage, since it gives an early signal that informs routing decisions and focuses agent time without requiring a hard pull. That alone keeps teams engaging the people who are most likely to respond or stay in the conversation.

Once a lead is accepted, that same score carries downstream into your CRM or dialer. Teams use this context to decide who should be worked first, who should enter warm-up through email or SMS, and who should move into slower, lower-cost nurture until their behavior changes. Faraday provides the intelligence that supports these decisions while the actual routing happens in the tools teams already use.

The impact isn’t just a cleaner scoring step, it’s the continuity it gives the entire funnel. From buying the lead to deciding who should be called next, the same prioritization signal flows through. Teams avoid overspending on records unlikely to engage, agents spend more of their day in meaningful conversations, and the call center starts to feel productive again instead of overloaded.

## The bottom line

Parallel dialing lets you get away with weak filters. Progressive dialing won’t.

 As the industry standardizes around progressive behavior to stay compliant, filtration becomes the real competitive advantage for most day-to-day calling. There are still situations where higher-volume strategies are appropriate. The point is to match the dialing approach to the job at hand and support both with strong intelligence in front of the dialer. 

If your 2026 plan involves dialing fewer people more thoughtfully, the key question isn't which dialer you use. It’s whether the intelligence in front of it is smart enough to make every call slot count.

As parallel dialing collapses under compliance pressure, teams win by pairing progressive dialing with smarter, outcome-driven filtration that prioritizes the right prospects at every step.

When every call is intentional, context matters more than ever: Progressive dialing makes smarter filters critical for your call center

**What happens when you combine contextual consumer data, lead scoring, intelligent dialing, and a brand finally ready to rethink its customer journey?**

In this panel (moderated by Faraday COO [Robin Spencer](https://www.linkedin.com/in/robinbspencer/)), [American Standard](https://www.americanstandard-us.com/)’s [Eric Kozak](https://www.linkedin.com/in/eric-kozak-11b32550/), [outboundIQ](https://www.outboundiq.com/) CEO [Jason Shatzkamer](https://www.linkedin.com/in/jshatzkamer/), and Faraday CEO [Andy Rossmeissl](https://www.linkedin.com/in/rossmeissl/) break down the operational changes behind American Standard’s call-center transformation.

They discuss what changed inside the organization, how prioritizing the right leads created 3× higher conversion opportunity, and how contact rates rose from ~5.5% to ~20% by pairing smarter data with smarter outreach.

🎯 **In this video, you’ll learn:**
- Why home-services contact rates are collapsing, and what actually fixes them.  
- The consumer data signals that reveal which leads will convert (and which won’t).
- How dialing strategy, not dial volume, drives real contact.
- The system American Standard used to turn a silent call center into a high-performing one.
- The biggest opportunities home-services marketers are missing heading into 2026.

Ready to learn more? Check it out! 👇

<iframe width="560" height="315" src="https://www.youtube.com/embed/cDnNu8Mx4EQ?si=THZJ2YVk0l4SjwPJ" 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>

A practical look at how American Standard used predictive scoring and smarter dialing to lift contact rates, prioritize high-value leads, and modernize its call-center funnel.

American Standard’s new contact playbook: Optimizing the lead funnel with smarter data and dialing

Picture this: You’ve just landed a new CMO job at your dream home services brand. National distribution, huge budgets, and an even bigger mandate to drive real growth, fast. 

Congratulations, **and good luck.**

Because here’s the truth: you don’t have six months to understand your new market. You probably don’t even have six weeks. Everyone from the CEO to the sales team wants to know how you’re going to grow, and they wanted to know yesterday. In fact, it’s probably why they decided to hire a new CMO in the first place.

That means you need to work efficiently from day one. Before you touch a campaign or sign a media contract, you need a clear, quantitative view of where your opportunity actually lies; not vaguely, not anecdotally. Quantitatively. 

That’s where [Market Opportunity Analysis (MOA)](/docs/abstractions/MOA) comes in.

## Why Market Opportunity Analysis matters in your first week

The unfortunate reality is that a lot of CMOs walk into new roles and inherit fragmented data: CRM exports, agency reports, dashboards that don’t line up. It’s a recipe for confusion and risk: marketing and sales chasing different audiences, agencies optimizing for the wrong KPIs, leadership making calls on partial truths.

MOA cuts through that noise. It takes your existing customer base, maps it against the [total addressable market](https://corporatefinanceinstitute.com/resources/management/total-addressable-market-tam/), and reveals where you’re overperforming, underpenetrated, or completely invisible.

For a home services brand, that means understanding where your most profitable jobs come from, down to the ZIP code, address, or even individual. Then finding more just like them. You can uncover patterns in income, housing age, household size, or hundreds of other datapoints that predict demand. 

The result? You can start targeting growth with precision, speed, and confidence.

## Real insights fast

Once you’ve seen how MOA reveals your market reality, the next question is why it matters: why it’s the first thing every new CMO should reach for. In a role where time and clarity are everything, MOA delivers both. It doesn’t just show you where the opportunities are; it gives you the confidence and direction to act on them immediately.

1. ### Speed to decision-making

MOA compresses the time between orientation and action. By combining your first-party data with Faraday’s pre-built national consumer dataset, you get a complete view of your market in days, not months. 

For you, this means walking into early leadership meetings armed with hard numbers (penetration rates, opportunity values, and clear maps of where to grow next) earning trust and alignment from day one.

2. ### Richer context and clarity

Most CMOs start out surrounded by data, but without context. MOA changes that by layering your internal records over [Faraday’s Identity Graph (FIG)](/features/faraday-identity-graph), giving you a complete view of how your customers fit within the broader market.

From there, it reveals the lookalike consumers you haven’t reached yet and the patterns driving demand: where you’ve over-indexed, who you’re missing, and which factors actually matter. That clarity turns scattered metrics into meaningful insight.

3. ### Tactical direction and alignment

Every MOA ends with concrete recommendations you can act on immediately: which regions to prioritize, which audiences to target, and how to allocate resources for maximum ROI. It replaces intuition with quantifiable strategy and gets marketing, sales, and leadership aligned fast around the same growth plan.

The result is more than analysis: it’s acceleration. You’re not just understanding your new market; you’re steering it.

## How to use MOA to guide your next moves

Understanding your market is one thing but acting on it is where the real value shows up. MOA isn’t a static report; it’s a decision engine that turns market clarity into measurable outcomes. Here’s how to put it to work from your first week onward:

1. ### Prioritize geographies

Focus your paid media, outreach, and sales enablement in areas with high lookalike potential: regions where your best customers have close behavioral or demographic twins in the broader market but low current penetration.

Faraday’s Market Opportunity Analysis identifies these lookalike clusters at the entity level, aligning them to real households and properties through the Faraday Identity Graph. This precision makes it easy to prioritize spend where incremental investment drives exponential growth.

2. ### Align with sales

Bring MOA findings into your next territory or franchise planning session. When marketing and sales share the same data-backed view of opportunity, you can align resources quickly and avoid wasted effort.

3. ### Refine messaging

Use MOA insights to tailor creative and offers to the consumer profiles that dominate your highest-value areas, and to extend those learnings to lookalike audiences with similar behaviors and needs.

This approach replaces gut feeling with precision. It sets a precedent for analytical leadership from day one and gives your team a clear, evidence-based framework for growth.

## The takeaway

Your first few weeks as a CMO set the tone for everything that follows. Market Opportunity Analysis is your shortcut to situational awareness. It’s a fast, objective way to see the whole board before you make your first move.

When you walk into that first strategy meeting, you won’t just be the new CMO with fresh ideas. You’ll be the one who already knows where the growth is hiding, how much there is, and the exact steps to capture it. That turns day-one clarity into your first real win.

## FAQ

### What is Market Opportunity Analysis (MOA)? 

MOA is Faraday’s framework for mapping your customer data against national consumer data to identify untapped market potential.

### How fast can a new CMO get insights from MOA? 

Most analyses run in a few days, not months, helping new leaders make confident, data-backed decisions quickly.

### Why does MOA matter more than standard reporting or dashboards? 

Dashboards tell you what’s already happened. MOA shows where to go next. By layering Faraday’s consumer data over your internal CRM data, CMOs gain a forward-looking view of opportunity that’s immediately actionable: where to invest, which segments to prioritize, and how much potential is still on the table.

### How does MOA connect to Faraday’s other predictive tools? 

MOA is the foundation. Once you know where opportunity lies, Faraday layers predictive scores—like Likelihood-to-Convert or Credit Score Proxy—and persona insights onto those markets. From there, the Identity Graph maps that intelligence down to real households and individuals, turning high-level opportunity into precise, actionable targeting.

### How does MOA improve marketing and sales alignment? 

Because MOA outputs are inherently shareable (like maps, penetration metrics, and opportunity rankings), it gives both teams a shared source of truth. Sales knows where to focus outreach, marketing knows where to invest, and leadership sees the projected return.

For new CMOs under pressure to grow, Market Opportunity Analysis cuts through data chaos to deliver a quantitative map of untapped markets. It provides a clear, actionable growth plan in days, not months, aligning marketing and sales from day one.

The CMO's shortcut to growth: how Market Opportunity Analysis delivers day-one wins

While there are 1,500+ demographic features available per American adult (from companies like Faraday), feeding all this raw data to an AI agent would overwhelm its "attention budget" - the limited cognitive capacity LLMs have to process context. Anybody who has used a code copilot like [Cursor](https://cursor.com) or [Claude Code](https://www.claude.com/product/claude-code) knows the feeling when its attention has wandered... after too much information, the LLM is distracted, not helpful, pissed off, stuck up a creek without a paddle. This is because of attention scarcity:

> LLMs have an “attention budget” that they draw on when parsing large volumes of context. Every new token introduced depletes this budget...

Anthropic's Applied AI team recently wrote a [short essay about the evolution from "prompt engineering" to "context engineering"](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents). The upshot is that you must continuously manage the information available to the LLM, keeping it focused by discarding unnecessary tokens but keeping "high-signal" tokens. Advanced AI agents like Cursor and Claude Code are rapidly getting better at this. It's much harder to tire them out now, thanks to strategies like:

- Tools - using vector search and `grep` to find only relevant code, instead of reading the whole codebase.
- Sub-agents - "sub-agents [handling] focused tasks with clean context windows [returning] condensed, distilled summar[ies] of [their] work."
- Compaction (compression) - "the practice of taking a conversation nearing the context window limit, summarizing its contents, and reinitiating a new context window with the summary" - often with the output of the tools or sub-agents.

This is relevant to my industry, [demographic data](https://faraday.ai/features/faraday-identity-graph) and [customer behavior predictions](https://faraday.ai/features/customer-behavior-predictions), because the amount of data available about every American adult - at least 1,500 features are available from Faraday alone - would easily overwhelm most LLM attention budgets. If you want to [build AI agents that interact with American consumers](https://hightouch.com/demo-ai-agents) and take advantage of third party data, you need to keep the signal while discarding the rest.

Let's detail 3 ways to do context engineering with demographic data. The upshot is to use quantitative methods as tools - really, as deterministic sub-agents - to compress the surfeit of raw demographic features into high-signal tokens and maximize the attention budget left over for AI agents to do their job:

## Decision tree models

Decision tree models compress thousands of demographic features into a single propensity score by learning which combinations of features predict a specific outcome. They work by recursively splitting the population based on feature values—much like a flowchart—to find patterns that distinguish people who achieve an outcome from those who don't. **Instead of forcing an AI agent to reason about every demographic detail, the model distills this complexity into one number that captures everything relevant to the prediction.**

The input is historical examples: people who converted versus those who didn't, along with demographic and behavioral features describing them. The output is a percentile score (0-100) or probability for each person, indicating how likely they are to achieve the outcome. This compression is dramatic—from potentially 1,500 input features per person down to a single score that the agent can easily incorporate into its decision-making.

[![Decision trees efficiently compress thousands of demographic datapoints into high-signal tokens for LLMs](https://developers.google.com/static/machine-learning/decision-forests/images/DecisionTree.png)](https://developers.google.com/static/machine-learning/decision-forests/images/DecisionTree.png)
_Decision trees efficiently compress thousands of demographic datapoints into high-signal tokens for LLMs ([source: Google ML](https://developers.google.com/machine-learning/decision-forests/decision-trees))_

Decision trees give you numeric scores and lift that can go straight into your context:

[![Help the LLM understand how convinced the user is](/images/blog/finance-lift-table.png)](/images/blog/finance-lift-table.png)
_Help the LLM understand how convinced the user is ([source: Faraday blog](/Users/seamus/fdy/src/magic/src/content/blog_posts/how-to-optimize-ads-for-compliance-in-finance.md))_

Consider a credit union building an AI agent to help members with financial planning. Without propensity models, the agent would need to process 1,500 demographic fields per member every time it considers whether to recommend opening a savings account. With a decision tree model, the agent simply receives a score indicating each member's likelihood to open an account, preserving its attention budget for actual conversation and advice.

## Recommender systems

Recommender systems compress interaction history—purchases, clicks, views—into ranked predictions of which product or action comes next. These systems analyze the matrix of who-interacted-with-what across your entire customer base to find latent patterns: "people who bought A and B tend to buy C next." **Instead of an AI agent processing raw transaction logs and product catalogs to figure out what to recommend, it receives a pre-computed ranked list that captures all that complexity.**

The input is a transaction or interaction matrix (rows are people, columns are products or actions) plus any product metadata. The output is a ranked list of recommendations for each person—typically the top 5-10 options most likely to resonate. This compression is especially powerful because it captures collaborative filtering patterns (what similar people liked) without forcing the agent to analyze the entire interaction history every time.

[![Recommenders use matrix factorization to efficiently compress product affinity and transaction history into high-signal tokens for LLMs](https://developers.google.com/machine-learning/recommendation/images/Matrixfactor.svg)](https://developers.google.com/machine-learning/recommendation/images/Matrixfactor.svg)
_Recommenders use matrix factorization to efficiently compress product affinity and transaction history into high-signal tokens for LLMs ([source: Google ML](https://developers.google.com/machine-learning/recommendation/collaborative/matrix))_

Imagine an e-commerce site building an AI shopping assistant. Instead of the agent analyzing every past purchase across thousands of products to determine what to suggest next, it receives "most likely next 5 products: [Product A, Product B, Product C, Product D, Product E]" and can focus its attention on how to present these recommendations conversationally based on the current chat context.

## Cluster models

Cluster models group similar people together and identify the defining characteristics of each group, compressing individual-level variation into persona-level patterns. Rather than treating each consumer as a unique snowflake across thousands of attributes, clustering finds natural segments where people share meaningful similarities. **An AI agent can then work with a handful of personas instead of reasoning about every individual's complete demographic profile—a massive reduction in cognitive load.**

[![Don't tell your LLM all the dots... tell it the cluster](https://developers.google.com/static/machine-learning/clustering/images/clustering_example.png)](https://developers.google.com/static/machine-learning/clustering/images/clustering_example.png)
_Don't tell your LLM all the dots... tell it the cluster ([source: Google ML](https://developers.google.com/machine-learning/clustering/overview))_

The input is your customer population with demographic and behavioral features. The output is typically 3-7 personas, each with defining characteristics and a label for every individual indicating which persona they belong to. For example, a model might identify distinct groups like "Budget-Conscious Families" and "Premium Singles," automatically determining which datapoints (income, household size, urban vs. rural) meaningfully distinguish these personas.

Consider a subscription service building an AI agent for customer support. Instead of the agent processing each customer's unique demographic and usage profile from scratch, it receives "this customer belongs to the Budget-Conscious Families persona" and can immediately apply relevant context about typical concerns, preferences, and communication styles for that segment. The agent's attention is freed up to focus on the specific support issue rather than figuring out who this customer is.

## Faraday implementation notes

For readers interested in how these compression techniques are implemented in practice, here's how Faraday handles each approach:

**Decision tree models**: Faraday [Outcomes](https://faraday.ai/docs/abstractions/outcomes) handles this compression using multiple algorithms including [random forests](https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-random-forest) and gradient boosted trees. The system trains separate models for different customer lifecycle stages (how long someone has been in the sales funnel), handles both vendor-enriched demographic data and customer-only data, and applies fairness adjustments to ensure protected demographic groups aren't disadvantaged by the predictions.

**Recommender systems**: Faraday [Recommenders](https://faraday.ai/docs/abstractions/recommenders) uses a two-stage approach: [matrix factorization](https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-matrix-factorization) finds collaborative filtering patterns, then propensity models (similar to Outcomes) refine those recommendations using demographic data. The system handles cold-start scenarios (new customers with no purchase history) by falling back to demographic-based predictions, and adapts to different data availability levels—working whether you have rich behavioral data or just basic customer information.

**Cluster models**: Faraday [Persona Sets](https://faraday.ai/docs/abstractions/persona_sets) implements this using [k-means++](https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-kmeans) clustering. The system automatically selects the optimal number of groups (balancing too few vs. too many personas), removes outliers before clustering to prevent extreme cases from distorting the personas, and identifies which specific datapoints distinguish each persona—importantly, not forcing every datapoint to matter when some simply don't differentiate between groups. It also builds lifetime value predictions for each persona, helping prioritize which segments deserve more attention.

## Conclusion

When you use AI copilots like Cursor or Claude Code, you can feel them efficiently quantifying and summarizing your code base. They use tools like vector search and `grep` to pull out the high signal tokens and leave themselves lots of attention budget to write code. This is similar to what's necessary with AI agents that want to use third party data to interact with consumers. You have to selectively compress information about the consumers; decision trees, recommenders, and clusters are a simple quantitative way to do this.

Just as code copilots like Claude Code use tools like vector search to pick out relevant code (saving attention for writing code), consumer-facing AI agents should use predictive models to pick relevant demographic signals (saving attention for conversation and decision-making).

Context engineering and demographic data

## Who this guide is for

This post is written for marketing and operations leaders in the home-services industry—the people responsible for filling calendars, scheduling appointments, and keeping call centers busy at roofing, solar, and plumbing companies. Whether you’re running campaigns in [ServiceTitan](https://www.servicetitan.com/), [Housecall Pro](https://www.housecallpro.com/), [Salesforce](https://www.salesforce.com/) or [HubSpot](http://hubspot.com), your challenge isn’t collecting more leads. It’s figuring out which ones will actually book jobs.

If you’re a CMO, VP of Marketing, or Marketing Ops leads in home services, this piece will show how a predictive data layer can plug into your existing [CRM](https://www.salesforce.com/crm/what-is-crm/) to help your team focus on high-intent prospects—and stop wasting time on leads who were never going to convert.

## The problem: Your CRM knows what happened before but not what happens next

Your CRM is great at logging calls, tracking appointments, and storing customer info. But it’s not helping you anticipate what will happen next, who is likely to convert, or how much they might buy. Your team is still guessing who to call first. Campaigns are firing blind. Reps are wasting hours chasing low-quality leads.

The problem isn’t your CRM—it’s what it’s missing: predictive context. You’ve got historical data, but not behavioral or financial signals that show who’s ready to book. That’s where adding data AND predictions comes in.

## The solution: The predictive data layer: what it is and why it matters

A predictive data layer adds real-world behavioral and financial signals drawn from verified consumer data to each contact in your CRM. In addition to first party data you collect from web form fills and past interactions, you get third party data points and scores based on how likely each lead is to book right now or in the next 30, 60, or 90 days.

Faraday’s predictive data layer integrates directly into your CRM—no replacement required—and automatically adds key datapoints that reveal intent, readiness, and fit. The result: your reps spend more time on leads that close, not leads that clog up the queue.

## What smarter CRMs deliver

Companies that add predictive enrichment to their see immediate results:

* [Fewer wasted calls as reps focus on high-intent prospects](/stories/call-center). If you could sort your call list, why wouldn't you?  
* Higher close rates across campaigns. By avoiding bad leads, or leaving them for last, you get better results sooner.  
* Cleaner funnels and more predictable revenue forecasting. You can calculate an "estimated value" (EV) for every lead, even if it just appeared in your pipeline.

## Faraday has the datapoints that change everything

Most CRMs weren’t designed to predict—they were designed to record. Faraday fills that gap by adding a predictive data layer that plugs directly into your existing tools. You don’t need a data science team, a new platform, or an overhaul. Just a connection to the [**Faraday Identity Graph**](/features/faraday-identity-graph), which powers over 1,500 verified consumer datapoints—everything from financial readiness to life-stage signals.

While our platform is chocked full of data that drives value, some are particularly useful for home services brands like yours. Here are the high-impact datapoints Faraday adds to your CRM:

**1️⃣ Credit score proxy:** Predicts a household’s financial readiness without a hard credit pull. Perfect for screening prospects before offering financing options.

**2️⃣ Likelihood-to-Convert:** Scores leads based on real behavioral patterns, showing who’s statistically most likely to book a job this week.

**3️⃣ Dwelling type:** Distinguishes renters from homeowners—critical for knowing whether to push service contracts or new installs.

**4️⃣ Household income & equity:** Adds key context for premium upsells, like full HVAC replacements or solar upgrades.

### So in short, with Faraday’s data, you can:

✅ Predict who can pay  
✅ Predict who will book  
✅ Predict what to offer next

## How to add predictive data to your existing CRM

Adding the datapoints you need to your CRM with Faraday is easy and fast. In a few careful steps, you can turn your existing customer database into a predictive engine that shows you who’s ready to book, who needs nurturing, and who’s not worth chasing—without touching your current workflows.

1. **Connect your records.** Faraday integrates directly with platforms like ServiceTitan, HubSpot, or your internal CRM through a secure API. Our [developer-friendly API](/docs/reference) and [detailed documentation](/docs) are a good starting point- and we'll give you a hand getting it set up. You can also give it a try using simple CSV uploads and downloads, to get comfortable with the data.  
2. **Enrich with Faraday’s Identity Graph.** We fill in the missing details—household, property, demographic, and behavioral data that CRMs alone can’t provide.  
3. **Deploy predictive datapoints.** Append fields like Credit Score Proxy or Likelihood-to-Convert so your team can instantly see which leads matter most.  
4. **Automate routing and prioritization.** Use these datapoints to filter and prioritize your call queues or lead-routing automations—whether that’s directly inside your CRM or through integrations like [ActiveProspect’](https://activeprospect.com/)s [LeadConduit](https://activeprospect.com/leadconduit/). This ensures your highest-value prospects are called first and passed to the right reps in real time, reducing manual triage and wasted dials.

> Implementation should take a couple weeks—after that, your CRM doesn’t just record what happened—it starts predicting what’s about to happen next.

## FAQ

1. **Do I need to replace my current CRM (like ServiceTitan or HubSpot) to use Faraday?** No. This is the most common misconception. Faraday is a predictive data layer, not a new CRM. It plugs into your existing system and enriches the customer records you already have. Your team keeps working in the tool they know, but with new, predictive data at their fingertips.

2. **What’s the difference between Faraday’s predictive scoring and my CRM's built-in lead scoring?** Traditional lead scoring is manual and behavior-based (e.g., "opened email" \= \+5 points). It only knows what a lead does with you. Faraday’s predictive scoring is AI-driven and market-based. It compares your leads to a massive, verified consumer dataset to find patterns. It answers a different question: not "Is this lead engaged with my marketing?" but "Does this lead look like someone who will actually book a job?"

3. **Do I need a data science team or data engineers to implement this?** It makes it easier, but not necessarily. Faraday has integrated with hundreds of marketing and operations teams and we can provide more or less help depending on your situation because we connect via pre-built integrations and APIs, you don't need data scientists to build or maintain models. Your ops team can map the new data fields (like Likelihood-to-Convert) into your CRM, and your marketing team can start using them in automated workflows immediately.

4. **How does a call center rep actually use a "Likelihood-to-Convert" score?** Faraday appends this score directly to each lead record in your CRM or routing system.

* **Qualification:** The system can filter out bad leads before they ever make it to your call center, if you want.  
* **Prioritization:** Reps can sort call queues by highest Likelihood-to-Convert, so they start with leads most likely to book—no guesswork or manual filtering.  
* **Routing:** Through integrations like **LeadConduit**, high-scoring leads can be automatically routed to top reps or live-transfer partners, while lower scores move to nurture sequences. The result: reps spend time where it counts—on the leads that actually book.

5. **Is Credit Score Proxy compliant?** The Credit Score Proxy is a modeled attribute, not actual credit bureau data. It is privacy-safe, does not require a "hard" or "soft" credit pull, and does not use or access any data from credit bureaus, making it fully compliant for marketing and prospecting use cases. More information can be found [here](/blog/prioritize-with-credit-score-proxy). 

## Putting it all together

So what should you walk away with? Your CRM isn’t broken—it’s just incomplete. By layering in Faraday’s powerful and varied datapoints, you give your team the visibility they’ve always needed: who’s likely to book, who’s financially ready, and who’s worth a follow-up later.

The result? Smarter routing, cleaner funnels, and a marketing team that can finally prove its impact in booked jobs—not dashboards.

If you’re ready to make your CRM predictive, reach out or take a look at our [API docs](/docs/reference).

Faraday’s predictive data layer turns your homeservices CRM into a tool that shows which leads are most likely to book, helping your team focus on high-intent prospects and close more jobs efficiently.

Your home-services CRM needs predictive data to predict who will actually book

If you’re using predictive models, the number one question you’ll eventually ask is: *Are these predictions actually working?*

At Faraday, we love that question. We’re not here to spin our wheels—we want to generate tangible value for each and every client. That ethos is built right into our sales process: before any client signs on the dotted line, we offer to run a custom [holdout](/blog/holdout-testing-evaluate-prediction-accuracy) analysis on their data to prove our models can accurately predict how their customers behave. That’s **transparency** and **trust** from the very beginning.

But after those lines are signed and once those models are live, it can sometimes become harder for clients to get that same level of visibility. Sure we have a variety of other [explainability and transparency](/features/ai-safety-and-responsible-ai) features but here at Faraday, we aren’t really ones for leaving well-enough alone. 

Trust in AI isn’t built on promises, it’s built on proof. So that’s why we’ve released **Efficacy Reporting**, a new feature in the Faraday dashboard that gives you yet another layer of insight into how your models are operating.

## What it does

Just like a holdout test, Efficacy Reporting measures the correlation between Faraday’s predictions and what actually happens. Reports are autogenerated for pipelines that have an [Outcome](/docs/abstractions/outcomes) in the payload (which must have eligibility explicitly configured). 

> 📌 Note: Efficacy Reports require a live, regularly updated dataset connection. One-off uploads won’t generate meaningful reports, since we need to verify outcomes over time.

Each report breaks down performance by decile—the top 10% of predicted scorers, the next 10%, and so on—so you can quickly see how outcomes line up against model expectations. A strong model will show the classic “lift” shape: higher attainment in the top deciles, tapering off toward the bottom.

[![Model efficacy report in Faraday dashboard](/images/blog/efficacy-reports-in-dashboard.png)](/images/blog/efficacy-reports-in-dashboard.png)

## Where to find it

Efficacy reports live in the [Pipeline tab](https://app.faraday.ai/pipelines) of the Dashboard for any qualifying pipeline. As they’re auto-generated, there’s no extra setup required, and they’re tied directly to the models they evaluate, so you’re always looking at performance in context.

## Who it’s for and how to use it

Efficacy reporting is designed for the teams and leaders who rely on predictive insights to make decisions. 

* Marketing and growth teams can use it to build confidence that their campaigns are aligned with reality.   
* Data teams can use it to validate that models are holding up over time without pulling ad‑hoc reports.   
* Executives can use it to get clear, ongoing evidence that predictive AI is delivering ROI.

In practice, efficacy reports let you validate Faraday models over time without extra manual work, spot potential performance issues early, and build confidence with internal stakeholders that predictions are worth scaling.

## What’s next for transparency at Faraday?

Efficacy reporting is just one step toward making the performance of your predictions as visible and explainable as possible. We’ll keep expanding this framework so you can always see, in concrete terms, how Faraday is driving real outcomes.

Because at the end of the day, predictive AI should be more than a [black box](/blog/the-ai-blackbox-is-a-liability) — it should be something you can measure, trust, and build on.

Efficacy reporting lets you see how well your Faraday models are performing in the real world by comparing predicted outcomes to actual results.

What's new: Efficacy reporting makes modeling even more transparent

We’re excited to launch a new email notification feature in Faraday that keeps you seamlessly informed when your most important resources complete a build or refresh, or enter an error state. No more manual checking; now you’ll know exactly when things wrap up (or need attention).

### What’s new
- **Email notifications for key resources** — Receive an email when important resources, like Outcomes or Datasets, finish building, refresh, or encounter an error.  
- **Selective rollout** — Everyone has access to the new *Settings → Notifications* feature, but only some users will have notifications pre-enabled. All other users can opt in at any time.  
- **Opt-in controls** — Notifications are off by default for new users.   

### Why it matters

Resources in the Faraday dependency graph often pass through intermediary states and may take hours—even a day or more—to complete. Until now, users had no automatic way of knowing when a resource built, refreshed, or failed, which could leave downstream workflows hanging. With these notifications, you can stay aligned in real time and act faster when things finish (or fail).

### Get started

If notifications aren’t already enabled for your account, visit *Settings → Notifications* in the Faraday app to turn them on and start receiving alerts.

Faraday now supports customizable email notifications, letting users opt in to alerts when key resources finish a build, refresh, or encounter errors.

What's new: Email notifications for key resource states

If you’ve spent five minutes in the identity resolution world, you’ve probably heard someone throw out a number like *“we match 99% of records, guaranteed\!”*

Sounds great, right? Until you make campaign projections for your CMO that … don’t entirely pan out.

The thing is, flashy numbers might close deals, but do they drive value? Only if they’re accurate. 

So sure, Faraday can hit 99% matching in certain conditions (we’ll get to that later). But does that mean it’s the number you should plan your campaign around? Well, that depends: on what identifiers you’re looking for, and what you put in to find them.

[![Jack Black magic number gif - school of rock](/images/blog/magic-number.gif)](/images/blog/magic-number.gif)

## Match rates aren’t one-size-fits-all, and they’re not a magic number

We’ve written before about why [match rates matter](/blog/identity-matching-makes-or-breaks-predictions) (because they’re the foundation of enrichment and predictions) and why [transparency and accuracy matter even more](/blog/is-your-id-match-rate-real) (because vanity numbers break downstream models). But here’s the thing: when it comes to match rates, there’s no universal “magic number.” Anyone selling a consistent 90%+ is probably doing one of three things:

* Counting duplicates (two bad emails for the price of one\!)  
* Leaning on easy identifiers like household address and presenting that rate as always true  
* Or—a personal favorite—straight-up inflating numbers.

In reality, match rates depend almost entirely on **what identifiers you hand us**. Think of it like cooking: bring fresh ingredients, and you’ll get a great meal. Show up with a half-empty pantry and a can of expired beans, and we can only do so much.

Moreover, it’s not just about whether something matches—it’s about *at what level* they match. Faraday can resolve identities at the person or household level, and in some cases at a broader level like ZIP code or metro area. Which one matters depends entirely on your use case. Sending direct mail? Household-level matches are usually perfect. Calling or emailing individuals? Then person-level accuracy is what you need.

## So what can you actually expect?

With all of that context in mind, here’s what you can actually expect with Faraday:

* **Email-only datasets** → **\~60%+** individual match rate. Not bad, considering how many emails are throwaways, duplicates, or churned.  
* **Add phone numbers** → things stabilize. Old numbers still get recycled, but you’re closer to **\~65%+**.  
* **Bring addresses into the mix** → now we’re cooking. Household match rates can climb into the **80–99%**+ range depending on data quality, with **70–80%** at the individual level.

**And here’s an important point**: yes, on a clean dataset with addresses, we can technically match **99% of residences**. But do we advertise a 99% match rate? 

No. Because that would be misleading. Of course we can recognize residence level matches on that dataset—we literally have the addresses—but what we’re actually proud of in that example is that we can match **\~80%** of *individuals* on the same dataset. Is that less flashy? Sure. But is it accurate, actionable, and potentially value driving? Absolutely. 

And that means you won’t be backtracking later when projects built on inflated numbers don’t pan out. 

## Why this matters for CMOs

If you’re a CMO you already know where this is headed. Inflated numbers might look good in a pitch deck, but they don’t hold up when you’re presenting results to the board. Every padded “match” that isn’t real will:

* Waste spend in acquisition  
* Poison your enrichment  
* Undermine predictive models you rely on

That’s why Faraday takes the harder road: we dedupe, we verify, and we only count matches that are actually usable in production. For you, that means you can confidently plan campaigns and allocate budget knowing the foundation is solid—no backtracking when the numbers don’t pan out.

## So what’s the punchline?

At the end of the day, match rate isn’t about chasing the biggest number—it’s about trusting the one that will actually power your campaigns. Faraday’s approach gives you the clarity and reliability you need to move forward with confidence.

Stop planning campaigns on vanity numbers. Let's see what *actual* match rates look like with your data.

Faraday delivers real, verifiable match rates—not inflated vanity numbers—so CMOs can confidently plan campaigns on accurate foundations that actually drive value.