News, tips and ideas on accelerating growth for apps and brands.
Facebook lookalike audiences have emerged as an extremely effective tool in the UA marketing toolbelt. Providing both scale and high ROAS when used correctly, Facebook lookalike audiences are a form of affinity-based marketing that can achieve game-changing results for your app marketing campaigns if you take the time to understand how it works and avoid the potential downside.
To understand how to optimize Facebook lookalike audiences for your app campaigns it helps to understand how lookalike modeling in general works. There are three major parts to lookalike modeling to consider:
Qualities analyzed for affinity
Step 1: Generate a Seed Audience
For non-Facebook lookalike modeling, source data is generally limited to your existing user base. The first step is segmenting your base to find the ones that are most desirable (most engaged, biggest spenders etc.) and using that subset of users as a seed audience.
Step 2: Analyze for affinities
Using data about your ideal users, a DMP will analyze the specific qualities about users to determine what correlations in affinity for your app exist. That could include things like what other apps someone downloads, demographic information and so on.
Step 3: Map affinities and deliver ads
Once correlations in affinity are identified, a DMP can act on those insights to deliver your ads to a relevant audience.
Step 4: Measure and Optimize
Once your ads are in-market, it’s important to closely monitor ROAS and make adjustments to the affinities being used to determine a lookalike audience. For example, If your model was initially based on a specific set of demographics, but your ROAS is taking a hit, you might try targeting based on common apps installed.
You are modeling based on your best customers, so in theory you should acquire more users with similar qualities
Better than a “pay and spray” approach where you take your best guess and make adjustments based on ROAS
If you are launching a new app or in a new market, this approach doesn’t work
Limited ability to scale
Depending on the technology you use, a lack of transparency might make it difficult to adjust and optimize your campaigns
For the most part, Facebook uses the same formula to produce it’s own lookalike audience. However, the thing that makes Facebook’s Lookalike model so powerful is the universe of data and insight into user behavior that it incorporates. Variables like app usage, app page likes, allow Facebook to produce a more nuanced profile of your seed audience that is about as good as it gets at the 1% level, though that dilutes as you add scale to your campaign.
Can’t beat breadth and depth of behavioral data
More nuanced view of seed audience
Incredibly powerful at 1% level
Facebook doesn’t provide any insight into what affinity variables it is optimizing for making it impossible to leverage insights across platforms
Impossible to optimize exclusively for parameters that deliver ROAS
Facebook Lookalike Audiences are essentially a black box, but you can still take steps to ensure your ad dollars are being allocated as efficiently as possible.
Focus on the 1%
The biggest risk involved with Facebook’s Lookalike model is that it becomes increasingly less effective as you scale beyond 1%. If you find yourself mid-campaign and realize you need more scale at the same ROAS, you are going to be out of luck. Do the math before you launch to make sure you can hit your goals before you exhaust the 1% audience and you will avoid a big headache.
Layer on additional affinity-based tactics
It’s risky to put all of your eggs in any one basket for a major UA campaign. If you like the lookalike model, we suggest complementing the campaign with other affinity-based tactics for similar results. Industry leaders like MZ, for example, are using Appnique to target high affinity users across Facebook, Google and Apple Search Ads and improving ROAS along the way.
Get your seed audience right
Any lookalike model is only as good as it’s seed audience. Make sure you feel confident that the data you put into the model is representative of who your best users truly are. If you are launching a new app or in a new market and don’t have seed data you feel confident about, consider tools like Appnique that analyzes publicly available data so you can get a complete picture of your target audience without incurring the risks that come with using private user data.
Facebook lookalike audiences are just one part of an affinity-based approach to user acquisition. If you are looking for ways to scale across Facebook, Apple Search Ads, and Google AdWords check out our on-demand webinar or contact us for a free consultation.