With the rise of e-commerce, the lingerie market is seeing more competition than ever. One of Faraday’s clients, an established luxury lingerie brand, knew they needed modern marketing solutions to stay relevant. Over the past decades, the brand has amassed a loyal following that spans the globe, thanks to their focus on providing beautifully hand-crafted garments. But to keep growing, their attention to customers can’t stop at the end of the production line.

## Customer profiling

Focusing on improving their holiday campaigns, the brand leveraged [predictive audiences](https://faraday.ai/features/predictions) to reach high-LTV leads and developed insightful personas to further personalization initiatives. Leveraging rich consumer data and predictive models, Faraday has helped the brand get in front of more high-LTV customers and better engage their existing ones with personalized experiences.

## Personalized messaging

After surfacing deep customer insights with the [Faraday Identity Graph](https://faraday.ai/features/data), Faraday generated data-driven personas. Unlike traditional customer personas, Faraday’s are generated with an algorithm using enriched customer data. They are also unique in that they can be stored in any marketing system to inform future engagement and make executing persona-based campaigns a breeze.

Faraday’s predictive personas allow brands to improve personalization initiatives and can help brands stay relevant to their customers. With the Faraday platform, the lingerie brand [pushed their personas seamlessly into Iterable](/integrations/predictions-with-iterable), their email marketing platform. Storing them there makes it easy to segment audiences for future engagement campaigns to include curated creative and messaging.

## Seeding audiences on Facebook

To capitalize on increased traffic around the holiday season, the brand used their newly built personas to seed high-LTV audiences on Facebook, making sure to include only those leads who were the likeliest to convert and stick around for the long-run.

Spending their Facebook budget on the personas that showed the most promise or high-lifetime value resulted in 19% higher click-through-rates, 5% higher average order value, and 43% higher conversion rates.

Discover how this luxury lingerie brand ran a highly personalized, hyper-targeted holiday campaign driven by consumer predictions.

How a luxury lingerie brand drove personalization and high-LTV audience seeding with predictions

Founded in 2016, organic skincare brand [Ogee](https://ogee.com/) relies heavily on social media for brand awareness and acquisition. Between 2018 and 2019, the brand was using Facebook Lookalikes and, as a result, saw a surge of new customers. However, Facebook didn’t give them the crucial insights needed for improved audience targeting and personalization.

As the customer base grew, Ogee wanted to focus on high-LTV customers to maximize their ROAS. So, they teamed up with Faraday to learn more about their customers, predict customer lifetime value, and target high-LTV audiences on Facebook.

## Enhancing first-party data to understand customers outside of purchases

Ogee’s customer base quickly evolved as the brand grew. To understand more about their new customers, Faraday dug deeper into the [data](https://faraday.ai/features/data), enhancing Ogee’s first-party data with rich third-party attributes to give the brand a more holistic picture of who their customers are, outside of their past transactions. Key findings included that many customers were college graduates, upper middle class, and had a history in buying beauty products.

## Using LTV predictions to identify best customers

To understand which types of customers were still worth going after, Faraday [predicted](https://faraday.ai/features/predictions) the customer lifetime value (LTV) for three distinct groups of customers, segmented by their date of first-purchase. Faraday predicted how much each group would spend over the following 6 months and over the following 12 months, using a probabilistic model that took into account each group’s transactional history. Calculating LTV can be a challenge for many brands, but it’s a crucial metric that influences revenue forecasting, ad spend, and fundraising efforts. Traditionally, the margin of error for these calculations is pretty high, but Faraday came in with an impressive 7% MOE when predicting Ogee’s customer LTV.

## Propensity models targeting high-LTV customers

To continue to attract the kind of high-LTV customers growing their brand, Ogee worked with Faraday to leverage propensity models, scoring the entirety of the U.S. to see who would most likely become a high-LTV customer. The top 1% of the scored group comprised the [custom Facebook audience](https://faraday.ai/predict/pipelines/likely_buyers_in_facebook) pushed from the Faraday platform. Ultimately, the predictive audience outperformed Ogee’s traditional Facebook Lookalike audiences. Spending their Facebook budget on predicted high-LTV leads lowered Ogee’s cost-per-result and significantly lifted the brands ROAS on Facebook.

Faraday's custom lookalike audiences and customer lifetime value predictions allowed Ogee to see massive results in their Facebook campaigns.

How Ogee used LTV predictions and predictive acquisition audiences during growth

The pandemic has destabilized many consumers’ financial situations, forcing them to reprioritize their finances in order to stay out of debt. Because of this, businesses offering creative solutions that allow individuals to pay back debt, invest, or easily start a savings account have become increasingly popular.

One of these businesses is [Happy Money](https://happymoney.com/), a fintech company on a mission to turn borrowers into savers, specializing in helping people pay off their credit card debt and save up for the future. Featured by publications like [Forbes](https://www.forbes.com/sites/donnafuscaldo/2019/09/05/happy-money-raises-70-million-at-a-valuation-of-nearly-500-million/?sh=6b37d1bd5777), The Wall Street Journal, and Fast Company, Happy Money has over 100K members and has helped pay off over \$2 billion in credit card debt by partnering with financial institutions willing to lend money.

Happy Money started working with Faraday to improve their marketing campaign performance by connecting to the people across the country who are in need of their financial services.

## Predicting optimal locations for expansion

In order to reach new leads across their marketing channels, Faraday built [predictive models](https://faraday.ai/features/predictions) using rich consumer data to find potential Happy Money members, identifying candidates interested in paying down credit card debt. This gave Happy Money access to high-performing audiences that they were able to push directly to Facebook during Q1 and Q2 2020.

This is particularly exciting because, despite the economic downturn in the beginning of the pandemic that accompanied high rates of job loss and hardships for many businesses, Happy Money was able to reach the consumers who would benefit personally from the brand’s services.

## Taking a multichannel approach

Facebook isn’t the only platform where Happy Money advertises. Like any smart brand, they know that reaching new leads on one social media platform just doesn’t cut it — they need to utilize their audiences across multiple acquisition channels.

Faraday’s custom audiences are unique in a number of ways. First off, they allow brands to get in front of consumers that they may otherwise not reach via traditional segmentation or Facebook’s own lookalikes. Second, the audiences are portable, meaning they can be optimized for and deployed to any marketing channel — like Facebook, Snapchat, direct mail, and ad display networks like Verizon Gemini — to support a strong omnichannel strategy.

These custom audiences aren’t just one-off lead lists, which is how a number of lead generation tools work. The Faraday platform delivers new, updated audiences every month to Happy Money’s acquisition channels to ensure that Happy Money is always reaching the most qualified leads.

## Using customer personas to seed new audiences

Within every brand’s customer base, there are multiple types of customers, which can be analyzed and grouped by persona. To hone in on ideal members, Happy Money is working with Faraday to generate audiences based on existing member personas.

[Faraday’s personas](https://faraday.ai/predict/category/persona_assignments) are developed with machine learning models to mitigate human bias and eliminate the chance of overlooking a significant customer type. Seeding audiences with these personas has many advantages, arguably the most important of which is the ability to personalize ad content and campaign messaging to maintain relevance with each audience.

As Happy Money continues to strengthen their omnichannel approach, these multichannel audiences will hopefully prove to be useful in reaching more people who look like the members that have grown Happy Money’s business the most.

How the booming FinTech company reaches the right consumers with Faraday's predictive targeting solutions.

How Happy Money optimizes omnichannel marketing with data-driven audiences

During the pandemic, interest rates for home mortgages reached all-time low. As people were spending significantly more time at home, buying a house — especially for those who relocated out of denser cities — has become a more practical option than renting. In August 2020, [home sales shot up](https://www.businessinsider.com/how-2020-broke-the-housing-market-inventory-could-run-out-2020-9) to reach a 14-year high.

Millennials in particular have made their mark on the real estate business, as they now make up the largest share of home buyers in the country, according to the [National Association of Realtors](https://www.nar.realtor/research-and-statistics/research-reports/home-buyer-and-seller-generational-trends). However, many businesses that service the home, like HVAC contractors, report that they [do not focus on millennials](https://www.achrnews.com/articles/144348-ignore-millennials-if-you-dont-want-to-reach-homeowners) for their marketing campaigns. This is their loss, especially since millennial homeowners are at their fingertips on social media and digital platforms, where targeting younger demographics is a breeze.

So, if you’re in the business of home renovation, DIY tools, residential energy efficiency, or HVAC, consider the following when marketing to millennials — a generation eager to take you up on your services.

## Find millennial homeowners on the right channels

Millennials are a digitally native generation. A [global survey](https://www.gwi.com/reports/millennials) conducted by the Global Web Index in Q3 2020 looked at more than 66,000 internet users and found that millennials ages 24–37 spent an average of 2 hours and 30 minutes each day on social media alone.

If your brand is not capitalizing on social media, digital ads, and OTT options like streaming channels to reach millennial homeowners, you’re missing out on crucial brand awareness and conversion opportunities.

In addition to actively targeting specific audiences, be sure your SEM strategy is up to par, as millennials are big online researchers. Back in the day, newspaper and yellow pages ads may have cut it, but that’s all switched over to online search engines now. Ensure that your areas of interest and/or operation are prioritized in any search your brand may be relevant to.

And once you’ve got your lead on the hook, you can ease their journey to conversion by communicating on the platforms that are most convenient for them; this may be SMS text, video chat, or a website chatbox. From getting a quote for a solar installation to matching a desired paint color, millennials value seamless communication that meets them where they already are in the digital realm.\

## Cash in on millennials' renovation obsession

To accommodate growing families, remote work, and even houseplant obsessions, millennials are big into remodeling and DIY renovations — they’ve even been coined the “renovation generation.” Compared to both older and younger groups, millennials are [3X more likely](https://www.homeadvisor.com/r/wp-content/uploads/2020/12/DP6355-StateOfHomeSpending-2020-R4.pdf?ref=faraday-blog) to take on home office projects and additions. It’s imperative that contractors and home service providers take these trends into account when fleshing out marketing campaigns.

This could mean kitchen remodelers including messaging around spending more time cooking in the kitchen since dine-in options are still limited around the country; or perhaps construction companies advertising the ease of adding on a home office to accommodate remote work. Even video-based consultations can be a deciding factor for homeowners concerned about social distancing or eager for ease of access to an expert.

No matter the product or service, using digital marketing campaigns to introduce ideas for how millennials can improve their at-home lifestyles may spark interests homeowners didn’t know they had, bringing them closer to converting.

## Emphasize eco-friendly products and services

The millennial homeowner is no stranger to “going green” — this messaging has been prevalent throughout much of their life, so much so that [73% of millenials](https://www.inc.com/melanie-curtin/73-percent-of-millennials-are-willing-to-spend-more-money-on-this-1-type-of-product.html) are willing to pay more for sustainable products. Green features are especially popular when they’re sure to lower utility bills in the future.

[During the pandemic](https://porch.com/advice/home-improvement-trends-covid), 34% of homeowners have made their homes more environmentally friendly, with half of that percentage being millennials and Gen Z. On average, millennials are going to spend $10,000 this year alone on home improvement projects, and they recognize that their return is greater if those projects include eco-friendly materials and products.

So whether you’re offering the installation of more environmentally friendly heat pumps, solar panels, or even sustainably harvested hardwood flooring or low VOC paints, make a point to emphasize that your products are eco-friendly to further appeal to millennials. This angle may be even more effective if used to support social media campaigns, as those platforms serve as a hub for eco-conscious consumers proliferating sustainable home improvement solutions.

## Tap into trends to market to millennials

As a digitally native generation, millennials easily take to social media trends, especially those that are good for their wallets, the environment, and lifestyles at home. Tapping into the right channels with messaging relevant to millennial interests can improve the performance of your campaigns and attract the consumers who hold the most spending power in the economy.

Digital campaigns, relevant messaging, and eco-friendly products are key to wooing the largest share of homebuyers in the country.

Marketing to millenials: Tips for home services companies

Work-life balance has been put to an extreme test this past year, with many people working from the same place they spend their free time. And for those with family obligations around schooling — as well as entertaining kids who aren’t doing their usual extracurricular activities — the notion of “free time” may hardly exist. In the sweet moments of respite, self-care is queen.

Over the last few years, self-care has evolved to be more than a fleeting trend in the health and wellness industry — it’s an entire market that spans fitness, nutrition, personal care, and beauty products. From books to vitamin supplements, it encompasses a variety of products, setting up a wide range of brands for entry into the market.

As more options pop up for customers, engaging leads and building brand loyalty may prove to be difficult. Luckily, there are steps marketers can take to capitalize on the growing self-care industry and reap the benefits of its popularity among consumers.

## Account for customers’ shifting finances

There’s nothing worse than being targeted by a brand for a product that you can’t afford. Since many consumers’ finances have taken a beating during the pandemic, it’s imperative to recognize that the people you included in target audiences a year ago may not be able to afford your products now.

To mitigate this issue, many self-care brands have begun partnering with fintech lenders to offer post-purchase payment plans. Not only does this make products with a higher price point available to a larger swath of consumers, but it also indicates to your customers that your brand recognizes upfront payments can be a challenge for some individuals. This type of signal can be especially important if your product is considered a luxury or nonessential item.

Another way to mitigate consumers’ change in financial standing is to segment audiences by financial bracket or income, pushing ads and promotions to each segment featuring products that are actually feasible purchases. This could improve your ad campaign performance, increasing conversion rates and repeat purchases.

## Get to know new types of customers

As brand loyalty shifts and the online presence of self-care products and services grows, many brands have seen an influx of new customers that don’t necessarily fit the mold of their existing audiences. Engaging these new customers may be different from what you’re used to.

One way to understand shifts in your audiences is to leverage your historical customer data to generate [personas](https://faraday.ai/features/predictions). With insight into your various types of customers, old and new, you can craft messaging that takes into account their respective needs and lifestyles. This may manifest as certain product recommendations in email campaigns, or perhaps direct mailers that feature promotions specific to each persona’s interests.

Additionally, several brands are using data science to predict customer lifetime value, which helps them be more strategic about discounting, CAC thresholds, etc. You may be surprised which customers are likely to take their business elsewhere, and which are going to be loyal for the long run. Not only does this analysis help brands predict future revenue more accurately, but it also allows marketers to focus on engaging the customers who will contribute most to that revenue.

## Determine which customer engagement strategies work for your brand

This article is by no means an exhaustive list of what marketers advertising self-care products and services can do to improve their customer engagement. And not everything touched on here fits every brand’s marketing strategy. By testing new channels, messaging, and audience segmentation, you can figure out what works best for your brand and your customers.

Whatever you end up doing, be sure to include data in the process; from building audiences to generating personas, data-driven marketing is what will set you apart and grow your self-care brand most efficiently.

As the self-care industry becomes increasingly competitive, brands must leverage data intelligently to keep customers engaged.

Customer engagement strategies for self-care brands

Audience overlap is a common issue marketers face when advertising on platforms that build native audiences or lookalikes. Often marketing teams will create two separate ad sets for a campaign, but end up reaching more or less the same audience despite trying to target different audiences. So rather than expanding their reach, brands are inadvertently oversaturating their audiences — which can frustrate consumers and result in wasted spend.

## What causes audience overlap with Facebook Ads?

Honestly, we aren’t sure. Many advertising platforms that generate audiences using brands’ uploaded first-party customer data in combination with the platform’s third-party data — Facebook, for example — don’t provide much insight into how they build those audiences. Faraday Data Scientists have seen time and time again Facebook pulling the same people into audiences, regardless of the target result (e.g. add-to-carts, clicks). We chalk it up to the training data.

## How audience overlap impacts brand reach

Audience overlap costs marketers valuable opportunities to reach potential customers. To serve the best ad to as many of the best leads as possible, brands must know who their customers are — but tracking just consumers’ digital footprints to guide ad creative isn't enough. One way to mitigate this issue is to upload custom audiences built with real-world, offline data.

We've seen a significant reduction in overlap between Faraday's custom audiences (from our third-party consumer database) and Facebook Lookalike audiences.

This approach has two key benefits: it addresses the problem of oversaturation and expands advertisers' potential reach beyond the ad platform's built-in capabilities.

## Tools to monitor and reduce audience overlap on Facebook and other channels

When ad sets overlap, they can compete against each other for visibility and unnecessarily increase marketing costs. You can (and should) monitor [overlapping audiences in Facebook](https://www.facebook.com/business/help/1679591828938781?id=176276233019487) monitoring overlap in the Facebook Audience Overlap Tool and uploading custom audiences that limit any duplication will be key here when thinking about ways to optimize spend and lower customer acquisition cost.

Overlap can happen on any channel where you build lookalike audiences, whether it’s Facebook or Hulu or Snapchat. Some platforms have taken measures to decrease the risk of oversaturating audiences by installing [frequency caps](https://www.advertisingweek360.com/hulu-a-better-experience-for-viewers-a-better-home-for-advertisers/?ref=faraday-blog) on ads. Hulu, for example, now makes sure that no ad is shown more than twice in an hour. But some customers have still reported seeing the same ads many times. This may be because of audience overlap — when Hulu runs an ad against multiple audiences, customers that are in more than one audience could see it again and again.

So, just because platforms are working to make ad experiences better doesn’t mean that brands don’t also have to put in the work. Because ultimately, this sort of viewer exhaustion can drive customers away, rather than reinforcing brand recognition and leading to conversions.

Learn how custom third-party audiences help reduce audience overlap across Facebook Ads campaigns.

Reducing audience overlap with custom audiences

Consumer spending has changed drastically during the pandemic. We saw a boom in household supplies and food categories as quarantine lockdowns took effect, and the trend continued through 2020. Discretionary spending became a luxury for many as unemployment rates hit reached [nearly 15%](https://sgp.fas.org/crs/misc/R46554.pdf) in April 2020. With such drastic changes, businesses had to shift their tactics and appeal to a new type of audience: the at-home consumer, who is fueling the “homebody economy.”

As the name suggests, these are the consumers who are spending the majority of their time at home. They’re likely working remotely, and certainly making most of their purchases online. Because of this, e-commerce has taken center stage for retailers, forcing brands that hadn’t implemented strong consumer-facing digital strategies before 2020 to adapt quickly.

But this adaptation goes beyond setting up an e-commerce platform — many businesses have adjusted their digital strategies to understand the swaths of new customers coming in the door and reel in at-home consumers with more personalized ads. And now consumers expect better online experiences, with [68% saying](https://www.forbes.com/sites/forbestechcouncil/2020/12/17/how-the-homebody-economy-is-shaping-digital-customer-experience/?sh=5807ede562eb) they have higher expectations of brands' digital abilities since the start of the pandemic. So, how can brands deliver?

## Get to know your new audience

The adage “Know your audience” has never been more relevant, with somewhat unpredictable consumer trends changing the audiences brands may have thought were set in stone. As more consumers were forced to shop online because of the pandemic (e-commerce now makes up 16.1% of all U.S. sales, up from 11.8% in the first quarter of 2020) businesses everywhere encountered unexpected customers.

To provide the enjoyable customer experience people expect, marketers need to dig into their customer data to better understand who they’re serving. Personalization isn’t just about meeting the expectations of the customers you already know and love — you’ve got to capture (and retain) the interest of these new customers, too.

Leveraging responsibly sourced consumer data can help round out brands’ understanding of their customers and aid in personalization efforts that go beyond web-based activity. By enhancing customer profiles, third-party data also bolsters targeted marketing strategies by making segmentation by customer type or persona more precise.

## Advertising to the homebody economy

When executing targeted marketing strategies, an omnichannel approach is more necessary than ever. Right now, brands have the chance to get in front of consumers they may never otherwise engage with because so many marketing channels are in play.

Between social media, OTT, and digital ads, brands should be capitalizing on the availability of the at-home consumer audience. Consumers’ media consumption is at a high, opening up lucrative marketing opportunities — not because there’s necessarily more ad space, but because there are more eyes on screens these days.

Radio and podcast ads aren’t out of the question either — though not as many people are commuting to work, [Nielsen reported](https://www.forbes.com/sites/bradadgate/2020/08/21/nielsen-how-the-pandemic-changed-at-home-media-consumption/?sh=765b07e45a28) that 75% of people say they listen to music during the workday, and 40% of those listeners do daily. In light of this, brands should consider including placing ads where their audiences tune in.

As for offline marketing channels, direct mailers are a great way to deliver personalized promotions to the at-home consumer — at home. It’s been reported that [70% of mailers that were personalized](https://www.sequeldm.com/direct-mail-trends-v) in some way as marketers vied for consumers’ attention. They were even integrated into digital strategies: QR code usage in direct mailers rose 69% in 2020 from 2019, making attribution easier for brands relying on mailers to convert customers online.

## At-home habits die hard

Even as Covid-19 vaccinations become more widely available and businesses are able to return to previous modes of operation and capacity, the shift to a homebody economy is most likely here to stay. [According to McKinsey](https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/survey-us-consumer-sentiment-during-the-coronavirus-crisis), 3 out of 4 Americans have tried new shopping habits (changed brands, stores, or way of engaging) and intend to stick with them after the pandemic. So, businesses must continue to adapt because meeting the customer where they are is more urgent than ever before.

People are spending more time than ever at home so brands are adapting to reach them there. Here's data-driven guidance for the homebody economy.

How to reach at-home consumers in a homebody economy

This spring, amidst widespread Covid-related lockdowns, Microsoft CEO Satya Nadella remarked that “we’ve seen two years’ worth of digital transformation in two months.” While COVID clearly shocked the economic system, it also expedited a direct-to-consumer trend across retail and CPG that has been gaining momentum for years.

[Ritual](https://ritual.com/) is one of the brands leading this revolution. Since launching in 2015, it has sold over a million bottles of vitamins through its subscription service. It’s brought transparency, quality, and reliability to an industry ripe with pseudoscience and half-truths. When you buy from Ritual, you can trust the source, ingredients, and science behind their vitamins.

Over the last year, Ritual has worked with Faraday on a few core initiatives to better understand their customers and grow their brand.

## Improving engagement with customer insights and personas

Ritual has used customer insights reports, the [Faraday Identity Graph](https://faraday.ai/features/data), and [custom personas](https://faraday.ai/features/predictions) generated through a [k-means clustering model](https://faraday.ai/blog/k-means-clustering-customer-personas) to break down their customers into logical segments, supporting targeted and personalized engagement efforts. This has resulted in improved customer engagement metrics across their email campaigns.

## Optimizing Google Ads spend with geo-bid modifiers

Ritual runs a sophisticated customer acquisition strategy across major digital, offline, and out-of-home channels; a significant piece of this is search engine marketing. Faraday works with Ritual to optimize their ad spend on [Google Adwords](https://faraday.ai/predict/pipelines/likely_buyers_in_google_ads) by employing geo-bid modifiers produced by the propensity models the brand built using our platform. The geo-bid modifiers are applied at the zip code level, automatically increasing or decreasing Ritual’s bid based on the propensity of people in the area of a Google search to subscribe to Ritual. Using pre- and post-performance analyses, Ritual saw a marked improvement in CPA efficiency across the metros where bid modifiers were activated. After the successful tests, they’ve chosen to roll out the bid-modifiers across all non-branded keyword campaigns, resulting in propensity-based geo-bid modifiers being activated across 10,000 zip codes.

## Driving new product adoption with customer predictions

Ritual has a long history of serving women – its founder Katerina Schneider started the company after struggling to find a source of prenatal vitamins she could trust. Earlier this year, they rolled out a line of vitamins for men. Ritual used Faraday to support this effort, identifying who within their existing customer base had a family member or partner that could benefit from this new product line.

## Check them out!

Ritual represents a company that we love. They’ve brought a great set of products to market, and they use data in a thoughtful and strategic way to improve not only their bottom line, but the customer experience as well. We strongly recommend you [give them a try](https://ritual.com/)!

It pays to get into the nuts and bolts of consumer prediction. From custom personas to geo-bid modifiers, Ritual is taking advantage of AI to drive growth.

How Ritual leverages customer insights and geo-bid modifiers for growth

Headlines abound on the [opportunities](https://www.techrepublic.com/article/companies-are-doubling-down-on-artificial-intelligence-and-machine-learning-due-to-pandemic/) and [dangers](https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing) of machine learning. Now armed with [plenty](https://www.propublica.org/article/breaking-the-black-box-when-algorithms-decide-what-you-pay) of [examples](https://www.nytimes.com/2019/08/20/upshot/housing-discrimination-algorithms-hud.html), the Data Science community is at least aware that our data lakes and algorithms can make life more difficult for real people if not properly used. But the root problem can be summed up in pretty simple terms:

> An algorithm contains the biases of its builder.

Examples of bias include:

- **Reporting bias** _(at the source)_ — Maybe the folks who filled out your online survey were mostly young, male, and computer-literate.
- **Implicit bias** _(the call is coming from inside the building)_ — Perhaps your image of an ideal job candidate fits into a mental archetype of which you're only semi-aware, and that image excludes those who don't match the archetype.

At Faraday, we have a handful of approaches we use to minimize these effects.

## Data sourcing

It's possible to get your hands on a pretty broad range of data points these days. Everything from the basics of home value, age, and income to the more esoteric pet ownership and liklihood to buy high-end men's apparel. **Faraday actively excludes data points relating to protected classes**. Race, ethnicity, primary language, and religion are excised from our database entirely and therefore excluded from our predictive models and audience definitions.

Of course, seemingly-harmless demographic variables can often act as proxies for these features, but our stance begins with the most effective levers we have available to pull. And further on in the pipeline we do check _results_ against features like race, but we do it at an aggregated level using standardized census data (more on that below).

## Model priors

We investigate the examples in our training data before they become machine-actionable profiles. In particular, we look at how well they're balanced by age, gender, and geography against the population at large. If there's a significant skew versus the baseline, we know there's a chance the resulting predictive model will show bias, so we rework the inputs.

## Model predictions

When a predictive model is fully-built and ready to be put into action, we typically assign a predictive score to everyone in the country as a matter of efficiency. But this also allows us to check whole populations against statistics that reflect the inclusion or exclusion of protected groups at a census block group level.

If we see - for example - that the lowest scores are concentrated in the block groups with the highest proportions of particular racial or ethnic groups, we'll know that _some_ bias has made it into the model, and we'll revise it.

The bottom line is that these strategies are only a start, and **the best defense against unfair algorithms is multi-pronged and in for the long haul.** If you have questions about how your own applications might be veering away from fairness, I suggest checking out:

- Google's [Marketers' guide to ML fairness](https://www.thinkwithgoogle.com/feature/ml-fairness-for-marketers) or [ML bias crash course](https://developers.google.com/machine-learning/crash-course/fairness/video-lecture)
- ["Abolish big data"](https://d4bl.org/videos/19-abolish-big-data) from [Data for Black Lives](https://d4bl.org/)
- The machine bias case studies at [diversity.ai](http://diversity.ai/#case)

How to reduce the effects of implicit bias in machine learning, from data sourcing to model predictions.

Mitigating implicit bias in machine learning

For years people have griped about the cost of buying and selling homes through real estate agents. The 6% commission paid to real estate agents has remained surprisingly resilient amidst a broad reduction in other financial transaction fees (think stock trade fees). As a homeowner or buyer, depending on your market, the only way to avoid paying agent commissions is through a for-sale-by-owner process. As of 2018, only [11% of American](https://www.nar.realtor/research-and-statistics/quick-real-estate-statistics) chose to sell their home this away and often the result is a lower overall sales price.

This is where our client [Homie](https://www.homie.com/) comes in. Homie’s mission is to make homeownership easy, affordable, and accessible to all. Its technology platform streamlines the home buying and selling processes to reduce fees and commissions, while buyers and sellers work with a real estate agent. To date, they estimate they’ve saved their clients over \$61MM.

Homie came to Faraday when they were in the process of expanding beyond their current markets of Utah and Arizona into Nevada. They were looking to leverage the Faraday platform to support a few key initiatives.

## Predicting optimal locations for expansion

Moving into a new market is always challenging for a company, largely due to the lack of on-the-ground experience and local brand recognition. Simultaneously, data collected from one market is not necessarily representative of the new market. To overcome these obstacles, Homie uses Faraday's location and market intelligence solutions capabilities to guide their expansion across the mountain west. Specifically, Faraday lets Homie predict which zip codes in their target markets contain significant populations of high-propensity customers (homebuyers and sellers), giving the team a better understanding of where they should prioritize their acquisition efforts.

## Reaching likely buyers and sellers, directly and cost-effectively

Working with their Faraday customer success manager, the Homie team built geo-normalized models that accounted for market variances across key attributes, such as home value and household income. Then, they deployed custom lookalike audiences to Facebook, building up additional training data in their new markets. The results from this work were positive, with Faraday-generated audiences doubling conversion rates and reducing acquisition costs by more than 33%.

_Learn more about [likely buyers in Facebook](https://faraday.ai/predict/pipelines/likely_buyers_in_facebook) with Faraday_

## Check them out!

We are excited about the work [Homie](https://www.homie.com/) is doing to give its customers more transparency and control over the home buying and selling process and while getting all the support they need from a whole team of licensed agents, attorneys, and experts.

How Homie uses Faraday to guide its market expansion strategy and reach likely customers, directly and cost-effectively.

All posts