Lookalike audiences are simple and powerful. Take an audience group that has performed well for you and then use platforms like Facebook or Google to find users that exhibit similar behavior patterns. It’s a great way to prospect new audiences and scale a successful product company beyond its local environment. It is not, however, a one size fits all solution and more often than not, tactics deployed effectively in one scenario may be ineffective in another operating under different assumptions.
Lookalike Audience Seeds For eCommerce
Though mostly Facebook is referenced here, this list of tactics is relevant for any ad platform with a “lookalike” functionality . For example, Google’s similar audiences and Taboola, Outbrains variation on lookalikes to name a few.
While many guides that discuss the nuts and bolts of how to set up a particular lookalike audience for a particular campaign type, few address the tactics behind each maneuver. Below is a birds-eye view of lookalike audiences marketers use and the conditions in which they have been successful.
1. Lifetime Value
- Requires a lot of purchase data over a long period of time
- Great for breaking an established product/service into a new market or country
For a company that has a backlog of purchases and customers, lifetime value is possibly the best metric to use when creating a lookalike audience. This tactic involves taking the creme de la creme of your customers, the top percentile in terms of lifetime value and use them as a lookalike audience seed, creating an audience that is likely to generate very positive results.
Facebook now offers this kind of custom audience natively. And for building something similar on Google Ads you’ll want to import a csv of your customer data filtered by the top 20% LTV customers.
2. Past Purchasers
- Requires continuous purchase data
- Great for expanding an existing customer base
A very common “best practice” for businesses. When deployed site-wide, this tactic is amazing for businesses that have a solid sales volume and feature a few products that cater to a specific user journey. For example, a band selling merchandise. In situations where ‘conflicting products’ are sold under the same roof, it’s important to separate the audiences by product category at least so that each audience shares a specific purchase intent. For example, a pharmacy store that sells both diapers and bodybuilding supplements might find that these product categories are best segmented into separate lookalike audiences, one for each category and presented with ads for products that present a natural continuation of the user journey.
3. Highest Cart Value
- Requires a lot of purchase data across to achieve cart size variance
- Great for finding more of your most profitable customers
Another option for companies with an extensive backlog is focusing on customers who checked out with particularly large carts and using them as a seed for lookalikes. This creates a lookalike that’s likely to repeat this binge shopping pattern. This tactic is especially effective in seasonal / holiday context.
4. Add to Cart Events
- Requires some data to be effective, but less than actual purchases
- A best practice tactic, employed by many eCommerce businesses
In remarketing, a visitor that’s added a product to cart is indicating both clear purchase intent and a sense of immediacy, making add to Cart audiences some of the highest quality audiences to work with. In Lookalikes, these audiences are an acceptable alternative to audiences based on previously completed purchases, for businesses operating on a smaller scale, or trying to develop a new product.
5. Product Page Views
- Doesn’t require vast amounts of historical data
- When done right can be effective on recently launched products
Using visitors that viewed a page as a lookalike seed sounds like a great idea, but in reality, takes skill to pull off well. The challenge is rooted in the fact that people may visit the page by accident, click on the wrong link, or otherwise arrive at the page without any purchase intent. Your media budget will be entirely wasted on these visitors and it’s important to find ways to filter them out of your campaigns.
6. Website Engagement
- A great predictor of purchase intent and doesn’t require a lot of historical data
- Can creates substantial overhead if not using an external solution
Though it is notoriously difficult to measure, engagement is a great indicator for purchase intent. Basic engagement metrics like time on site and scroll depth are fairly simple to set up with GTM, with Facebook going so far as to offer metrics like top percentage of time on site within their ads platform, allowing marketers to build campaigns based on site-specific engagement patterns. By creating a threshold which filters out unengaged audiences, a metric like Page views can become a viable audience.
In some cases, engagement is the only metric available, for example, in case of a content-oriented page like a blog or magazine. For these websites, understanding engagement is key to effectively taking advantage of lookalike audiences and remarketing in general.
Pro Tip: To build a lookalike audience based on website engagement metrics, you need to identify what level of engagement is ideal for flagging intent. Use Google Analytics and create a filtered segment for converters. Then inspect that average time-on-site, number of pages viewed, and scroll-depth events to better understand engagement benchmarks of converters… and use them as the basis for your new lookalike audience.
7. Data-Science Fueled Engagement Audiences
Although you could set up event tracking manually through Google Tag Manager, you can also automate and improve your engagement based lookalike audiences by leveraging plug-and-play data-science.
Google Smart Lists
Google offers conversion optimization to websites that have 500+ monthly eCommerce transactions and 10k+ page views. A steep curve. Websites that don’t have that kind of volume, receive optimization based on statistical data from similar websites.
Facebook Top Percentile Lists
Facebook allows you to target a percentile of visitors that stayed the most on your website. For example “Visitors by Time spent – Top 25% – Time on Site”, and other combinations. Being a facebook solution it’s only available within the Facebook ad platform.
Fixel Visitor Scoring
Fixel uses machine-learning that looks across 60+ data points to create a unique model for each website and deliver powerful engagement audiences scored from high to low intent – syncing directly into most ad platforms.
Bonus Tip: Facebook Lookalike Audiences: 1% vs 10%
One question worth addressing is the question of similarity percentage. The optimal lookalike similarity percentage is often debated. The chief tradeoff being between the audience scale and compatibility. In simpler terms, you choose between a smaller audience that’s expected to deliver great performance or a larger audience that’s expected to deliver “ok” performance.
Fortunately, we don’t have to rely on gut feeling on this. AdEspresso ran an experiment on the subject and we’re happy to share their conclusions.
- 1% lookalikes outperformed the 5% and 10% in a split test
- When targeting a large audience, campaigns will perform well even after a 14-day runway
- It’s optimal to target 1%-2% and not more than that
Based on how much prior data you have, your level of traffic, number of transactions, you can choose different lookalike audiences and run campaigns for them.
The more data you have, the more effectively you will be able to find your ideal customers.If you have low traffic and low level of transactions, start with “Website Engagement” lookalike audience and test 1% and 5% audiences on Facebook.
PS. Be careful with creating lookalike audiences. If you identify the wrong segment of the audience on your site, you may end up targeting people who are not exactly interested in buying your products. You must be sure to segment people with clear buying intent. Or, ideally, those who purchased of course.