The advanced targeting technique proven to drive 10x improvements in Facebook advertising performance
"+1" to let us know.
Facebook has become one of the most effective advertising channels thanks to the amount of data it has on its users. This data allows for a wide array of targeting methods for advertisers.On one hand, the variety of options can be a blessing since it allows Facebook marketers to get more granular with targeting than they can anywhere else. For marketers who know exactly who they want to target, this is great.
On the other hand, the large number of options can also be a curse, since it’s very easy to waste ad spend testing various targeting combinations in an attempt to arrive at the best-performing audience segments.
At Custora we are big believers in the power of continuous testing and optimization. But we’ll let you in on a secret that we have found through our work with some of the largest and most sophisticated retailers in the world. Combining predictive customer lifetime value with Facebook Lookalike Audience targeting consistently delivers 5 to 10x improvements in Facebook advertising performance.
Why Use Predicted Customer Lifetime Value?
Lifetime value is the revenue a customer generates over time. By harnessing the power of (P)CLV to inform more intelligent and efficient media buying, you can move away from proxy metrics like click-through rates and cost per action to true ROI. The impact can be very powerful. Here is a simple example.
Let’s say you’re serving an ad for your cat t-shirts to two Facebook users, Sonya and David. David’s cost per acquisition (CPA) is $9, while Sonya costs $6 to acquire. Sonya may have been cheaper to acquire, but she only buys one $9 shirt, and then never shops with you again. David, on the other hand, continues to purchase from your business. Over the course of his life with your brand he will buy $36 worth of cat t-shirts.
So while David may be 33% more expensive to acquire than Sonya, he purchases 400% more and generates 6X greater ROI than Sonya. In the long run, customers like David are more valuable to your business, even though they cost more to acquire. Now the question is, how do you use these insights to improve the performance of your Facebook advertising?
Adding The Power of Lookalike Audience Targeting
Facebook’s algorithm for building Lookalike audiences sifts through its data on an input audience (called a seed audience) to find patterns between the seed audience and the overall Facebook user base.
This algorithm takes into account many facets of information that Facebook has on its users, information which transcends simple user input data and includes inferred attributes. Facebook doesn’t go into detail about what exactly is included in its Lookalike modeling, but it’s fairly safe to assume it includes demographic data, interests, and observed online behaviors of a given seed audience.
Facebook then extrapolates patterns to build a larger Lookalike Audience of users that most resemble the patterns inferred from the seed audience. Since the Lookaike algorithm is mostly black box, it requires you to give up some control over who ends up in your target audience, as you have less ability to explicitly choose the targeting parameters for each audience.
Building Your High CLV Lookalike Audience
In order to create a Lookalike Audience of your highest value customers, first you need to calculate the CLV for each customer. Customer Lifetime Value is a prediction of all the value a business will derive from their entire relationship with a customer.
Many brands use historical sales performance to measure CLV. But CLV is a forward-looking metric. It is the estimated future cash flows that you can expect from each customer. Historical CLV, since it is backward looking, can produce misleading results when the company, the market, or both have changed. In addition, historical CLV is limited when trying to measure the CLV of customers using new channels or tactics. Luckily, there are several methods that can be used to predict CLV, including extrapolation, supervised learning algorithms, and probabilistic modeling.
If your company is lucky enough to have a data science team, that team has likely built CLV models that you will need to access for Lookalike Audience targeting. If not, you will want to purchase a solution like Custora which accurately predicts CLV for each customer based on advanced statistical methods like Bayesian Inference and Pareto/NBD.
(If you would like to learn more about predicted models for CLV please check out the very comprehensive overview in Custora U.)
To create a Lookalike Audience of high CLV customers in Facebook, you need to segment your customer file by predicted CLV and pull out the top 5 or 10% of that list. These high value customers make up your seed audience. Seed audiences in Facebook must contain at least 100 members, but you should aim for much more than that. The reason for this is that Facebook will be able to build a more similar and higher quality lookalike audience if you feed it more data points.
However, too many data points can actually lead to pattern degradation. Facebook identifies the optimal audience size to be 10,000 – 50,000, but we’ve seen the greatest success with seed audiences in the 2,000 – 5,000 range.
Selecting the Audience Size – Quality over Quantity
After creating your seed audience and uploading the list into Facebook (get our step by step guide to lookalike targeting on Facebook here if you’d like to know how specifically to do this), you will need to select the size of the Lookalke Audience that you want to create.
Here you have an important choice to make—how large should you make the Lookalike Audience? The audience size can range from 1% to 10% of the population of Facebook users in the target countries you select.
The team at Adespresso just published the results of an interesting experiment that they ran to answer this question.
They created Facebook Lookalike Audiences based on past high value purchasers.
All of their Lookalike Audiences were targeted in the USA, meaning that even the audience targeting 1% of the country’s population included 2.1 million people.
Their first Lookalike Audience targeted 1% of US population on Facebook that most resembled their best customers.
The second Lookalike Audience was larger, targeting 5% of the US population.
The third Lookalike Audience they tested targeted 10% of people in the US most closely matching their best customers.
During the 14 days of their campaign test here’s what they found:
- The 1% Lookalike Audience had a cost-per-lead of $3.748
- The 5% Lookalike Audience had a cost-per-lead of $4.162
- The 10% Lookalike Audience had a cost-per-lead of $6.364
The 10% Lookalike Audience had a 70% higher cost-per-conversion than the 1% Lookalike Audience. The results of their experiment indicate that quality is more important than quantity when doing lookalike Audience targeting. Of course we have not run this test ourselves, but we would assume the audience that is most like your high CLV customers would also have the highest lifetime value in addition to a lower cost per lead.
The Performance Impact of High Value Customer Lookalike Targeting
Lookalike modeling is proven to significantly improve online advertising results. A study by Exelate found that:
Lookalike modeling “results in double or even triple the results of standard targeting, according to the 30 percent of advertisers and more than half of agencies who reported using the tactic.” – Exelate Study
Custora customers typically see Return on Ad Spend improvements that range from 5x to 10x over more traditional targeting approaches. If you’d like to see how one leading online retailer is using high CLV Lookalike Targeting on Facebook take a look at our Calendars.com case study.
The bottom line? If you are looking for a new way to get a better return for your advertising spend do high value customer Lookalike Audience campaigns on Facebook. You won’t regret it.
The Adobe 2018 Digital Trends report was recently released and it gave its readers some interesting insights into this year’s top marketing priorities around data, design, and technology. The report, which consisted of close to 13,000 participants from various industries including retail, e-commerce and the marketing agency sector, revealed the trends that distinguished the […]