As an advertiser, marketer, or leader at an e-commerce business, your primary aim is to get the greatest return out of every advertising dollar you spend on ad platforms like Meta, Google, etc.
But it can be tough to make more informed decisions about your marketing strategies, optimize advertising efforts or even help your performance marketing team get better results on your ad campaigns if you don’t understand how these ad platforms work and show your ads to the right audience who can potentially become long-term customers of your brand.
In this article we will discuss how machine learning algorithms of these ad platforms work, so you can have a better understanding and make more informed choices about your marketing strategies by leveraging user data to improve ad campaigns, increase customer acquisition, reduce acquisition cost and drive higher average order values, helping you increase LTV:CAC at your business.
Let’s dive in!
The walled gardens (closed ecosystems allowing access to user data only through their platform) like Meta (Facebook, Instagram), Google Ads (YouTube), TikTok, Snap, Pinterest, and X are powerful platforms for billions of users to find, interact with, and buy from millions of brands.
They are efficient marketplaces that connect users with content and products they are likely to engage with and eventually purchase.
These ad platforms use powerful machine learning and artificial intelligence algorithms to efficiently match specific users and advertisers. And, although the exact workings of these algorithms are proprietary information, let’s reverse engineer how they work.
For illustration purposes, we will use Meta, but the same concepts apply to other platforms as well.
The first place to start is to understand how ad platforms segment user groups.
The users that on these platforms can be broadly segmented into three groups, defined based on user behaviour across millions of advertisers and tens of thousands of types of advertisers (segments).
These users see all ads on platforms but usually don’t interact with them. From a brand exposure standpoint, this group is valuable, but not so for performance marketing if your goal is to convert users who see your ads into purchasers within a specific time window. Which is why, the platforms assign a lower monetary value to such users.
These users engage with advertisements through clicks and comments; however, the advertisers aren’t able to get a transaction from this group of users. The monetary value of this user-group is higher than never-engagers.
Unlike “Never Engagers” & “Non Converters”, this is the highest value user-group that engages with the ads and also takes the desired action (e.g., making a purchase) within a given time window.
Given the value of such users is higher for advertisers, they are willing to pay more to ad platforms for their ads to be shown to this user-group.
Hence, platforms assign a higher monetary value to this group of users, which makes competition fierce among advertisers for ad impressions from these users.
As an advertiser you shouldn't mix up these groups with “In Market” or “Movable Middle. The main difference is that the above groups are defined across millions of advertisers.
However, as an advertiser, you might want to know who is in-market or movable middle for your own brand, helping you craft better ad campaigns to target users that convert into paying customers.
An In-Market user for advertisers and ad platforms like Meta is a high-value user who is targeted based on their browsing history and purchase behaviour.
These users are individuals who are actively considering purchasing products within a specific product category and are therefore more likely to respond positively to ads targeted to them from same category that they have been engaging with.
Hence, if retargeted with ads from related product and brand categories, leveraging cross-category information, Meta and similar ad platforms keep them engaged and increase the likelihood of conversion.
Moveable Middle Users are individuals who have shown interest in a product but have not yet completed a purchase, but can convert with some additional nurturing.
For example, if a user who clicked on an ad, visited the website, and engaged with the product but did not make the purchase immediately. Their behaviour and demographics are analyzed to estimate a high chance of conversion, and they are targeted with ads and other marketing touches to encourage the further to complete the purchase.
Let’s get a deeper understanding of “In-Market User” and “Movable Middle User” with specific uses cases and how an ad platform like Meta engage these users with ads.
Let’s look at Beth. She is in her mid-30s, living on the East Coast, in a relationship, and shops for herself, her family, and significant other. Meta has segmented her into the high-value category because she has taken desired actions from some ads shown to her on Meta.
Ad platforms like Meta have hundreds of millions of high-value users, given their scale, advertisement networks, and deep engagements they see from the users on their platform.
Let’s say, Beth was considering gifting men’s trendy shorts to her significant other for their Birthday. She browsed the Ten Thousand website, browsed and added an item (pair of shorts) to the cart, however, decided to purchase later after doing her own research. With this user-intent shown by Beth, Ten Thousand wants to retarget Beth on Meta, so they share her data with Meta’s ad platform for retargeting.
Based on the information that Meta has received on Beth from Ten Thousand, they deemed her as someone with a high estimated action rate (EAR) for a pair of shorts: i.e. she is likely to take the desired action (purchase) if shown an ad for shorts.
She also may be interested in patio furniture as another category based on her browsing history for other products and events that Meta may have received from other merchants. However, the platform further extrapolates that given she is interested in shorts, she might be interested in other men’s clothing products — T-shirts, jackets, trousers, jeans, polos, etc. and she might even consider a subscription to a try-before-you-buy merchant.
Based on this, Meta expects her estimated action rate (EAR) for these product categories will also be higher. And, if Meta gets additional engagement data from Beth (such as - if she is clicks, comments, saves, or watches 25% or more of a video ad), then that will send additional signals to Meta’s ad platform about her preference for these extended product lineups.
Using Ten Thousand’s data on Beth, Meta shows her a retargeting ad with social proof and a personalized discount offer. If Beth doesn’t engage with this ad, Meta will continue to use cross-category information and Beth’s (EAR) to show her ads from similar merchants. This will continue until Beth engages with another ad (satisfying her dopaminergic urge) or she has seen enough ads from the category.
As a result of this, Beth will be seeing these ads from the men’s clothing category - from shorts, tees, shirts, jeans, trousers, khakis, blazers, and including an ad from a try-before-you-buy subscription brand.
Meta deems she is an “In-Market” user and is likely to engage with these product categories based on her interest in Ten Thousand’s shorts — as these categories are adjacent to shorts.
And at some point, as Beth feels convinced with Ten Thousand’s pair of shorts or another product in adjacent product category, she is more likely to convert and make a purchase.
Now, let’s consider another user — Jack Smith, also in his mid-30s, married with kids, tech-savvy, does his own research before making purchase decisions. Jack was browsing the internet using his latest iPhone model connected to his home Wi-Fi in Los Angeles around 8 pm.
While he was browsing, he clicks on an ad for Ten Thousand with compelling social proof and subsequently visited Ten Thousand’s website and engaged with the website (e.g. - filled out the True Fit size quiz, selected a 5-inch inseam liner with size L).
He also added a pair of tactical training shorts with over 5,000 reviews to his cart (valued at $68). But, before he makes a purchase decision, he would prefer to do some more research.
Hence, Jack didn’t complete the purchase in the same user-session, as he needed more time to make up his mind.
This behavioural pattern from a user makes them a “Movable Middle” user, someone who is not in the market to purchase a product, but if an ad is shown to them that piques their interest, they might engage with that ad, click on the website, etc. but won’t convert in the same user-session or even in the next few days.
But considering they have shown interest in an ad by engaging with it, the brand in consideration might just be able to convert them into a paying customer with retargeting and additional nurturing.
Both “In-Market” and “Movable Middle” segments of the market have the potential of becoming your brand’s customer. The pool of “In-Market” buyers is much smaller than the “Moveable Middle” for any product category, hence, advertisers must identify and tap into the “Moveable Middle” segment for them to efficiently scale paid social marketing.
Although it’s definitely easier to target “In-Market” users, that does not mean it’s easier to convert them into a paying customer. The reason is simple: if a user is in-market and actively looking to purchase a product in a category similar to yours, be assured that your competitor brands are also competing in the ad auction for their ads to be shown to the same user. Hence, the brand that outspends to win the ad-auction has higher chances of eventually converting the customer.
“Movable Middle” is a unique segment of users that haven’t shown buying intent yet for your products but might have similar characteristics and user behaviours like your best customers. Meaning, if this group is further nurtured, they might just convert, purchase your product, and become a paying customer of your brand. And, the best thing is, your competition will have no idea of such users' existence if they can’t identify them, hence they can’t advertise to them.
Now you might be thinking, if your competition can’t identify the “Movable Middle”, how can you?
That’s where the power of machine learning and artificial intelligence comes in.
Remember Jack from the "Movable Middle" user considering Ten Thousand's pair of shorts example above?
Well, Based on the browsing data Ten Thousand has on Jack, along with additional demographic attributes Angler has on him (age, gender, interests, purchase insights). And, Angler’s AI conversion model estimates Jack has a 28% chance of completing a purchase in next 7-days and he falls in the 90th percentile of Angler AI score.
If Ten Thousand sends this additional user-data and signal to Meta, it can help them use Jack’s Angler AI score to trick Meta to assign higher estimated action rate (EAR) for him, and he is likely to see an ad from Ten Thousand (or other advertisers from the same category).
When Jack engages and clicks on the ad, he has 7-days from that point to complete a purchase for the ad to take credit for the purchase. Jack may even receive another marketing touch (email, SMS etc.) from the advertiser in the mean time, however, from first touch attribution basis the click on Meta will still get the credit for the purchase.
By understanding how ad platforms like Meta use machine learning algorithms to segment users and target them with ads to connect them to various brands, you, as an advertiser or a leader in your E-com business can leverage user data to maximize engagement and increase conversions on your ads.
Additionally, armed with this knowledge, you can use Angler’s predictive AI to optimize your ad spend towards both "In-Market" users and "Movable Middle" users. This will allow you to acquire more high-value customers who might even not be in the market for your products yet but will convert easily with some nurturing, and then purchase frequently with higher average order values from there on helping you increase customer LTV while lowering acquisition costs.