B2C companies operate in highly competitive markets, where customer acquisition and retention are vital for sustainable growth. Nowadays, it can be extremely difficult for brands to acquire customers while maintaining healthy customer acquisition costs.

Successful B2C companies are investing heavily in modern data and machine learning solutions, like predictive lead scoring, to drive efficiencies and improve outcomes across their customer acquisition, engagement, and retention programs.

If you are just starting to learn about predictive lead scoring or evaluating solutions for your B2C business, this article is for you. Here’s what you’ll learn:

- What is predictive lead scoring?
- How B2C companies use predictive lead scoring
- Core components of predictive lead scoring for B2C
- Fundamentals of a good B2C lead scoring dataset
- How to implement predictive lead scoring using Faraday

## What is predictive lead scoring?

Lead scoring is a data-driven method of assessing and prioritizing leads based on their potential to convert into paying customers, and it’s a critical component of several B2C marketing and sales organizations. Simply put, leads with higher scores are considered more likely to convert, allowing sales and marketing teams to focus their attention on the right leads.

**Traditional B2C lead scoring** models often rely on a point-based system where humans determine the lead scoring criteria—typically a combination of basic demographic information and online behaviors such as pageviews, email engagement, or form submissions—and assign points to leads when they meet it.

While these point-based lead scoring models can help marketing and sales teams prioritize their efforts, they come with significant drawbacks. Lead scoring criteria are often arbitrarily decided, resulting in lead scores that may not accurately reflect a lead's true potential. Additionally, because they rely heavily on first-party engagement data, lead scores often end up being a lagging indicator of a lead’s conversion likelihood.

**Predictive lead scoring** leverages the power of machine learning to analyze historical lead data and build statistical models that can predict the likelihood of a lead becoming a customer. By analyzing much larger datasets, machine learning algorithms identify patterns and important features that contribute to a lead's likelihood of conversion.

![Image of Faraday model report features table](/images/blog/product/feature-importance.png)

_Note: the image above illustrates important features surfaced in [Faraday model reports](/blog/faraday-model-reports)._

This takes the guesswork out of choosing lead scoring criteria for traditional point-based models. With the right lead scoring dataset, these models can predict a lead’s likelihood to convert before they ever visit your website or open an email.

## How B2C companies use predictive lead scoring

Implementing the right predictive lead scoring system (one tailored for B2C lead scoring) unlocks powerful lead optimization and personalization options for marketing and sales teams.

- **Lead ranking and suppression:** A simple yet powerful use case for predictive lead scoring. Rank your leads, set a suppression threshold (e.g. bottom 70%), and only surface the likeliest-to-convert leads in your outreach tools. This enables sales and call center reps to prioritize their outreach and improve win rates across the board.

- **Lead routing and distribution:** A slightly more creative use case is to use predictive lead scores to tier your leads and distribute leads from the best tier to your best reps; a great way to maximize your win potential. One of the leading solar installers in the nation [improved call center conversions by 33% with this approach](/blog/solar-call-center-lead-conversion-boost).

- **Product recommendations:** If your predictive lead scoring system supports product-specific modeling, you can predict which products or services a lead is most likely to purchase. These product-specific lead scores enable marketing and sales teams to make informed product recommendations across relevant touchpoints.

- **Customized discounts/promotions:** Most predictive lead scoring models are built to predict the likelihood of a lead converting to a customer. Why not predict the likelihood of leads converting to high-LTV customers and offer custom discounts or incentives to the leads who are most likely to spend a lot? This is one of the most creative use cases we’ve seen.

## Core components of predictive lead scoring for B2C

Predictive lead scoring requires robust data (first-party and third-party), customizable and explainable models, and flexible integrations to bring lead data into the system and push predictive lead scores out to your marketing and sales tools.

### Historical lead data

Historical lead data provides the necessary foundation for building accurate predictive models. This first-party data includes information about past leads’ behaviors, attributes, and their outcomes (e.g., whether they converted into customers or not).

### Lead enrichment data

Lead enrichment data is crucial for predictive lead scoring because it enhances models' accuracy and depth of understanding about leads. For B2C lead scoring, this third-party consumer data typically includes demographic information, financial information, lifestyle information, and property information.

_Note: Given the changing data privacy landscape, it’s crucial that your predictive lead scoring system uses responsibly sourced consumer data. [Avoid data from third-party cookies and social scraping](/blog/death-of-third-party-cookies-strategies)._

### Machine learning capabilities

There are several ways to leverage machine learning for predictive lead scoring, all with varying degrees of complexity.

Most CRMs and marketing automation platforms now offer built-in predictive scoring capabilities. These solutions typically leverage blackbox (unexplainable) machine learning and rely heavily, if not entirely, on the first-party you have in the system. They’re easy to implement, but lack third-party consumer data which is crucial for B2C lead scoring.

If you have data scientists and engineers with enough bandwidth to spare, you can try taking a homegrown approach. However, in-house development costs can be substantial so unless you already have machine learning pipelines and are licensing consumer data, we recommend [using our API to score leads instead](/developers/quickstarts/score-leads).

Predictive analytics and customer prediction platforms like Faraday operate somewhere in the middle, providing users with better control of their models while streamlining laborious machine learning processes. With Faraday, users connect their lead and customer data and tell the system what to predict (e.g. lead conversion likelihood). The system handles the rest (data enrichment, model building, score deployment, etc.). We’ll show you [how to implement predictive lead scoring with Faraday](/blog/predictive-lead-scoring-b2c#how-to-implement-predictive-lead-scoring-with-faraday) later on.

### System integrations

Integrations are necessary for predictive lead scoring to impact revenue. Predictive lead scoring systems need to ingest new lead and customer data to improve models and deploy predictive lead scores into your marketing and sales systems. Ensure that your system can connect flexibly with your data sources and marketing and sales platforms.

## Fundamentals of a good B2C lead scoring dataset

For B2C companies, best-in-class predictive lead scoring requires a combination of historical data and third-party consumer data. Historical lead and customer data enables machine learning algorithms to distinguish successful outcomes (e.g. leads converting into paying customers) from unsuccessful outcomes. Consumer data lets the algorithm identify important “real life” characteristics about the leads that convert into customers versus those that don’t. For the best results, you will need the following data:

- **Identity data** (e.g. name, physical address, and phone number) is necessary to match your lead and customer records with the correct consumer profiles acquired from third-party data sources. This matching process is commonly referred to as identity resolution.
- **Transactional data** (e.g. lead acquisition date, purchase date, and order value) is necessary to define important cohorts like customers, leads, and stale leads. The machine learning algorithm will examine these cohorts to identify patterns that indicate the likelihood that an individual will move from the “leads” cohort to the “customers” cohort.
- **Engagement data** (e.g. pageviews, email subscriptions, and free trial signups) enables the algorithm to understand how certain behaviors contribute to a lead’s likelihood of conversion.
- **Consumer data** (e.g. demography, lifestyle, property, and financial information) enables the algorithm to understand how different human characteristics contribute to a lead’s likelihood of conversion. Here’s a small snapshot of Faraday’s [built-in consumer data](/features/data):

![Image of Faraday traits console](/images/blog/leads-traits-console.png)

Now that you’re familiar with predictive lead scoring and what’s required to generate accurate B2C lead scores, we’ll show you how to set up predictive lead scoring pipelines with Faraday.

## How to implement predictive lead scoring with Faraday

With Faraday, you can implement best-in-class predictive lead scoring within a day, without support from your dev team. To do this, you’ll simply upload your lead and customer data and use [predictive lead scoring recipes](/predict/category/lead_scores) to easily configure predictive models and deploy lead scores to your stack.

![Image of Faraday recipes wizard](/images/blog/leads-recipes.png)

You can upload your historical lead and customer data with a CSV file or create a connection to your database or data warehouse. Faraday automatically enriches your data with built-in consumer data profiles, adding hundreds of attributes to your lead and customer records.

![Image of Faraday traits console](/images/blog/leads-persona-set.png)

Once your data is in the system, just choose a lead scoring recipe, add your leads and customers, then hit save.

![Image of Faraday lead scoring in HubSpot recipe](/images/blog/leads-lead-scoring-in-hubspot.png)

Faraday will automatically build predictive models and provide model performance insights to help you determine what kind of results you can expect.

![Image of Faraday outcome score](/images/blog/leads-model-report.png)

Once you’re satisfied with your predictive lead scoring models, add deployment destinations (e.g. HubSpot) to your pipeline and turn it on.

![Image of Faraday lead scoring pipeline and deployment](/images/blog/leads-pipeline.png)

Just like that, you have predictive lead scores flowing into your stack.

![Image of Faraday lead scores in HubSpot](/images/blog/leads-hubspot-lead-scoring.png)

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

Understand how predictive lead scoring works and what's required to build effective B2C lead scoring models using machine learning.

Predictive lead scoring: Best practices for B2C lead scoring using machine learning

If you’re a marketer in the finance industry, you know you have some heavy regulations to work with on [Google](https://support.google.com/adspolicy/answer/9917652?hl=en), [Facebook](https://www.facebook.com/business/help/2408667629202904), and other platforms, where consumer protections are put in place to help prevent discrimination. These regulations can make it difficult to reach your ideal customers when targeting new audiences, which likely means that you’re spending more than you need to on inefficient acquisition.

Complying with financial advertising regulations doesn’t mean you need to throw in the towel and accept worse performance. You can respect your customers’ privacy and be smart with your marketing strategy at the same time–you just need to think out of the box. That’s where built-in [third-party consumer data](https://faraday.ai/features/data) and [responsible AI](https://faraday.ai/features/responsible-ai) come in.

The data you have on your customers might be good, but purchases, adoptions, and clickstreams aren’t robust enough to give you a true leg up on your competitors, nor is it enough to ensure you’re not unknowingly discriminating against protected classes. Third-party data, when unified with your first-party customer data, can give you a 360º view of your customers, leads, and virtually any other group of people your business cares about.

With this breadth and depth of data in your pocket, you can then use AI to optimize your ad spend by surfacing your most likely-to-buy customers while remaining compliant by:

- Excluding protected classes from your predictions to ensure no unintended bias
- Targeting at the right aggregation level for your industry (counties, rather than zip codes, for example)
- Removing both human bias and error from the equation by using AI
- Monitoring and minimizing bias in predictive models through bias reporting

Faraday enables you to do all of these things–without a data science degree or programming knowledge–and in under 15 minutes. Here’s how.

## Connect your customer data to combine it with third-party attributes

To start, you plug in last quarter’s adoption data in by [mapping the data in Datasets](https://faraday.ai/docs/cohorts/customers_from_hosted_csv#option-a-each-row-represents-a-transaction-event).

![Image of identity set definition](/images/blog/finance-identity-set-definition.png)

Next, you use this new data to define a cohort–a group of people you’re interested in–of these new adopters.

![Image of cohort definition](/images/blog/finance-cohort-definition-booked-date.png)

## Build a predictive outcome to identify likely buyers

Last, you follow the recipes for likely buyer predictions in [Google Ads](https://app.faraday.ai/recipes/likely_buyers_in_google_ads) and [Facebook](https://app.faraday.ai/recipes/likely_buyers_in_facebook_custom_audiences) to create your likely buyers outcome, selecting your _customers_ as your attainment cohort because that’s your end-goal–you want people to become customers and enter your _customers_ cohort.

![Image of new outcome screen](/images/blog/finance-new-outcome.png)

## Exclude protected classes with trait blocking

When describing your outcome, you choose _trait blocking_ to remove protected class traits. This ensures you’re not only complying with financial fair lending regulations, but are also using AI responsibly.

![Image of trait blocking checkbox](/images/blog/finance-trait-blocking.png)

## Choose the right aggregation level for your likely buyer predictions

Once your predictions are ready, set your aggregation level based on how high you’d like to zoom out of a location–such as by _county_ to remain compliant with Google’s restrictions on zip code targeting.

![Image of pipeline aggregation level](/images/blog/finance-pipeline-aggregation-level.png)

## Push high-scoring locations to your ad platform

With your highest-scoring locations in-hand, you plug them into your Google Ads campaign:

![Image of Google Ads campaign location upload](/images/blog/finance-google-ads-campaign-locations.png)

And your Facebook campaign:

![Image of Facebook campaign location upload](/images/blog/finance-facebook-campaign-locations.png)

That’s it! Your best-bet locations are plugged into Google and Facebook, and fewer than 15 minutes were spent making your predictions.

## Real results from a customer in the financial services space

A Faraday client in finance recently followed these exact steps to kick off a new acquisition campaign for an underperforming insurance plan, and they’re expecting a **3.2x lift in conversion rate** in the top 10% of individuals alone.

![Image of Faraday lift table](/images/blog/finance-lift-table.png)

Next up for them is adding [direct mail](/integrations/predictions-with-poplar) into the mix, but the possibilities are endless.

[Sign up](https://faraday.ai/signup) free today.

In finance, remaining compliant to fair lending regulations doesn’t mean you need to accept worse acquisition performance. Learn how to use AI and third-party data to optimize ad spend by surfacing  your most likely-to-buy customers in under 15 minutes, while ensuring you're compliant to any relevant regulations.

How to optimize ad targeting and comply with financial advertising regulations

We are excited to announce that [AccuLynx](https://acculynx.com/), a leading provider of cloud-based business management software for roofing contractors, has launched a new Lead Intelligence feature that uses AI to predict homeowner buying behavior.

[Lead Intelligence](https://acculynx.com/roofing-sales-lead-intelligence/), powered by [Faraday's prediction API](https://faraday.ai/developers), taps into responsibly sourced consumer data and machine learning to enrich and score Acculynx customers’ leads every week, and provides users with up-to-date lead rankings that can help prioritize their sales efforts and close more business efficiently.

“Lead Intelligence is a game-changer for roofing companies,” said Mark Rattin, VP of Product at AccuLynx. “AccuLynx customers no longer have to guess which leads to call first, or only rely on their experienced sales people to quickly close deals. Now they can improve their sales win rates by focusing their efforts on the best leads.”

## Here’s how it works

Lead Intelligence uses Faraday to analyze hundreds of data points on all of their customers’ leads, including demographic, financial, property data and more, which then provides a score that indicates the likelihood of a lead converting into a customer.

By leveraging Faraday's API, Acculynx was able to develop the Lead Intelligence feature in a matter of months, providing users with a game-changing tool included in all subscriptions and saving significant engineering and data science resources that would otherwise be required.

We at Faraday are proud to support Acculynx’s mission of providing roofing contractors with easy-to-use technology tools that help them work smarter and grow their business.

How AccuLynx uses Faraday’s prediction API to power a game-changing Lead Intelligence feature for roofing companies across the US.

Acculynx launches AI-powered lead intelligence with Faraday prediction API

If you have at least 2000 payments, Stripe data probably contains your buyer signal - the unique patterns that predict who will buy from you.

Last month, a friendly brand exported about 10,000 payment records from Stripe as a CSV. Then they uploaded them to Faraday and, within 24 hours, got their results:

![Buyer signal showing importance of different features](/images/blog/buyer-signal.png)

Let's focus on one feature - "Science/new technology." If a person is interested in this, they are much more likely to buy this brand's products:

![Likelihood to buy based on science interest](/images/blog/science-directionality.png)

Faraday is batteries-included - we know things like whether people like science thanks to our [built-in consumer data](https://faraday.ai/features/data).

Now, this brand can predict the likelihood of more than **240 million Americans** to buy their product. Faraday can take any list of people and sort it by likelihood to buy... or Faraday can generate the list in the first place, using a ["Deployment"](https://faraday.ai/docs/abstractions/deployments).

So... how did they get this? First, by doing a [Stripe Payment Report](https://support.stripe.com/questions/exporting-payment-reports). In particular, doing a [**Successful Payments**](https://dashboard.stripe.com/payments?status%5B%5D=successful) report.

![You want Succeeded payments](/images/blog/stripe-payment-export-succeeded.png)

Faraday only needs the following fields:

![PII, date, and amount fields](/images/blog/stripe-custom-columns.png)

(Later, you can set up an automated integration. But for now, since this is a relatively stable model based on inherent consumer characteristics, it should be stable for at least a few weeks.)

In Faraday, you map the fields:

![Address mapping](/images/blog/stripe-dataset-identity-mapping.png)

Then, [follow the standard process for importing transactions](https://faraday.ai/docs/cohorts/customers_from_hosted_csv#option-a-each-row-represents-a-transaction-event).

And finally, you make a [likely buyers model](/docs/abstractions/outcomes).

In total, if this takes longer than 15 minutes, [email the CTO](mailto:seamus@faraday.ai) because that's a critical bug!

Happy Signal Hunting!

-Seamus

PS. When this brand loaded our model into Facebook, we immediately started beating Facebook's internal model, even though we were still in learning phase...

![Beating Facebook](/images/blog/stripe-adset.png)

[Sign up](https://faraday.ai/signup) free today.

If you have at least 2000 payments, Stripe data probably contains your buyer signal - the unique patterns that predict who will buy from you.

How to determine buyer signal from Stripe payments

Purchasing jewelry is traditionally positioned around the moment. Whether it’s for romance or a special occasion, jewelry campaigns tend to focus on everything but the actual wearer. [gorjana](https://gorjana.com/) is the fine jewelry company that says otherwise by providing jewelry you’ll live in.

With this focus on the wearer, it comes as no surprise that gorjana wanted to know who their ideal customers were, so they teamed up with Faraday to validate their customer profile and reach more potential customers across digital channels and direct mail.

We recently chatted with [Elton Graham](https://www.linkedin.com/in/eltonxgraham/), Chief Marketing Officer at gorjana, to further explore their challenges and how they worked with Faraday to solve them.

## Gaining customer insights by clustering for personas

Previously, gorjana had a pretty good hunch about who their core customers were, but were looking to confirm and deepen their understanding of them with more data. In came Faraday, unlocking a wealth of data that they could leverage to not only confirm their profile, but validate their core customer segments with the people-first focus they were looking for.

To do this they leveraged the [Faraday Identity Graph](https://faraday.ai/features/data), enriching their first-party data with 1500+ responsibly sourced attributes on nearly every adult consumer in the U.S. With this combination of data, gorjana was able to build brand-specific customer personas using Faraday’s built-in machine learning capabilities. These quantitatively developed personas enabled the gorjana team to identify similarities and differences between their core customer segments to further understand how their retail-only, ecommerce-only, & omni-channel customers compared.

## Calculating total addressable market (TAM) with potential buyer predictions

[Faraday’s potential customer predictions](https://faraday.ai/features/predictions) are often used to create custom audiences for customer acquisition campaigns. The gorjana team took it a step further and leveraged those predictions to calculate their TAM and penetration rates in designated market areas across the U.S. This prediction-based TAM analysis is currently helping to inform go-to-market decisions and campaign strategies across the country.

## Reaching more potential customers across paid digital and direct mail

With a stronger understanding of their customers, the gorjana team leaned into their potential customer predictions to create custom audiences of high propensity to buy consumers for their digital ad campaigns. The team saw meaningful improvements in cost per acquisition and return on ad spend with these custom audiences from Faraday.

When asked what’s on the horizon for gorjana, their response was to expand their use of Faraday’s potential customer audiences to direct mail, and to build new predictive outcomes to identify likely gifters.

We love working with gorjana - they’re a people-first brand making fine jewelry for everyday wear. [Check them out](https://gorjana.com/) to find the perfect pieces for your jewelry stack.

At gorjana, Elton Graham’s team used Faraday to gain customer insights, validate core customer segments, and reach more potential customers with data-driven campaigns.

How gorjana fuels growth with responsible data and AI

AI matters when it makes something you already do work better. Faraday is a [Google Cloud partner](https://cloud.google.com/blog/products/ai-machine-learning/building-an-open-generative-ai-partner-ecosystem?ref=faraday-blog) that helps hundreds of US brands grow their revenue using AI. These brands did growth initiatives, lead scoring, LTV maximization, acquisition campaigns, personalization and churn reduction before Faraday—but with Faraday, they improve their margins, finding hidden signals that may cover every adult in the United States.

Built on [Google Cloud AI](https://cloud.google.com/products/ai?ref=faraday-blog), Faraday lets brands use the world's most advanced artificial intelligence (AI) and machine learning (ML) tools by choosing from tried-and-true recipes. Users sign into the self-service app at https://app.faraday.ai, pull key data from their existing data stacks, define their business objectives, and let Faraday - powered by Google Cloud’s [BigQuery ML](https://cloud.google.com/bigquery-ml/docs?ref=faraday-blog) - do feature engineering, model optimization, cross-evaluation, backtesting, bias detection, and scoring. What's more, Faraday is batteries-included, with built-in life event, demographic, financial, and property data on 240 million US adults. This consented, opt-in data is included by default in Faraday's models, without clients ever having to come into contact with privacy-sensitive data.

Seamus Abshere, Faraday CTO says that Faraday could only be built on Google Cloud:

"Google Cloud is 3-5 years ahead of the competition because its integration of AI with data warehouse is already mature. For years, our data scientists and developers have been working in the same system - BigQuery ML. Google Cloud's approach to AI breaks down traditional barriers between data analysis and software development. It makes our data scientists happy to directly develop infinite scale production systems," Abshere said.

Faraday works with customers like Boll & Branch, Advia Credit Union, Aspiration, Snap Finance, Tula Skincare, Ethos Life Insurance and ADT to grow revenue. Boll & Branch increased its DTC revenue using persona predictions, one of the options available in the Faraday app:

"Using Faraday, we saw a 30% lift in conversion with our personalized email campaigns," says Katia Unlu, Chief Digital Officer at Boll & Branch.

Advia Credit Union doubled auto loan conversions using Faraday propensity predictions, sending direct mail and email promotions to the best-fit members.

“Faraday’s platform gives organizations a framework for applying AI to improve critical programs like marketing or consumer campaigns,” said [Manvinder Singh, Managing Director, Partnerships](https://www.linkedin.com/in/manvinder/?ref=faraday-blog) at Google Cloud. “We’re delighted to support Faraday’s platform and growth on Google Cloud, and to make their platform more accessible to customers via the Google Cloud Marketplace.”

The Faraday platform was built on Google Cloud, and utilizes Google Cloud AI and ML services to deliver more capability and value to customers.

Faraday is now available in the Google Cloud Marketplace.

Faraday is a proud Google Cloud partner and uses Google Cloud AI to help US brands grow revenue with cutting-edge AI and ML tools.

Faraday gives clients access to Google Cloud AI using tried-and-true recipes

Predictions are complex, and consumer predictions for marketing aren’t an exception. In many cases, you might find yourself given the green or red light to go after a lead, prevent a churn, or approach a customer in another way that’s specific to a business goal your org has. Having access to that knowledge is great, but it’s a black box–simply knowing whether or not a lead will bite isn’t enough if you want to truly understand your customers and the actions they take.

In Faraday, every predictive model you create for your business is scored based on how well you can expect the model to perform, and includes a full technical report describing how the model was built. While the full technical report includes a swathe of info for data science-minded users, for the sake of this guide we’re going to take a look at how **features of importance** and **directionality** impact a sample **lead conversion** model, which you can find in every outcome and recommender.

## Features of importance

![Image of features of importance in a Faraday outcome](/images/blog/product/feature-importance.png)

In each outcome and recommender, you’ll find a **data** section describing the top features of importance for that model broken down into tenures, or how long an individual existed within the eligibility cohort before entering the attainment cohort. The features that had the most impact in creating the model are listed in the chart, sorted by the level of impact. In the above image, you can see that features such as high ticket home decor quintile, activity, total online dollars quintile, and household income had a major impact on the model. This means that when thinking about who you’re targeting with these predictions, you can tailor creative and messaging appropriately–but that still feels a little broad. Let’s take a look at directionality to see exactly how these features of importance break down.

## Directionality

![Image of directionality in model report](/images/blog/product/feature-directionality.png)

When you take a deeper look at a feature of importance in feature directionality, you’ll find it broken down into ranges if the feature is numeric, or selections if not. In the above example, green indicates positive, and red negative. For **household income** (HHI), it’s clear that consumers with an HHI of over 75k rank higher on average (green), whereas all HHI ranges lower than 75k rank lower (red). For **shopping styles**, _luxury offline_ and _prefers online_ shoppers rank higher, while _retail therapy_ shoppers rank lower.

That’s cool and all, but what’s the takeaway?

In short, these breakdowns tell you that in this lead conversion model, the people that are most likely to convert are largely making more than 75k/yr, and prefer to shop online unless it's for luxury items. Using this info, your marketing to them can confidently personalize toward the positive features so that your content is relevant.

Learn how to analyze your predictive models to deeper understand your customer base.

Faraday model reports: Understanding feature importance and directionality

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.

## 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

[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

With 79% of consumers saying they’re only likely to engage with an offer if it has been personalized, personalization should be at the forefront of your team's marketing strategy year-in and year-out.

Faraday Strategic Account Managers Alexandra Noyes and Nick Haggerty recently gave a deep dive into how their clients are leveraging Faraday's bespoke, AI-generated personas to help them see eye-to-eye with their own customers. Every marketer knows that the right message, in front of the right customers, at the right time, is key, and Faraday enables you to plug these insights directly into your stack in just a few clicks. Hear from the experts as the duo explores personalization use cases that Faraday clients are currently employing, field questions from the audience, and more.

**Check out the recording below, where you'll learn how to**:

- Leverage consumer predictions to create personalized content
- Segment your audience with a focus on customer behavior
- Identify attributes that make leads more likely to convert

Join Faraday Strategic Account Managers Alexandra Noyes and Nick Haggerty as they dig into how clients are leveraging Faraday's bespoke, AI-generated personas.

All posts