WeChat Lookalike targeting: 20% higher advertising ROI based on who’s already purchasing from you.

06 mars


Focusing your WeChat targeting on location, interests, and demographics isn’t enough to run high-ROI campaigns in an increasingly competitive Chinese environment. Tencent is finally improving their advertising options: they rolled out last year Lookalike audience targeting, a feature Facebook has been offering since 2013. According to digital performance agency Bamboo, the majority of your Facebook budget should go to Lookalike audiences, which consistently show a higher click-through rate. Is it the same with WeChat? How does WeChat Lookalike compare to Facebook’s?

Splio’s technology enables marketers to identify and bind customer data on all channels on WeChat in order to help them reduce and optimize acquisition budgets.


What is Lookalike targeting?

Lookalike targeting uses a known “high quality” population – typically existing customers – as a basis for advertisement targeting. Instead of fixed arbitrary criteria such as gender, location, age, education etc. An algorithm analyses this seed population and builds a target audience of similar people by finding common criteria between their profiles. Lookalike has the following advantages over criteria or “tag” based targeting:

  • Lower manpower: criteria/tag-based targeting requires a lot of trial-and-error playing with optimization of the criteria used, the resulting addressable population, and the actual performance. Lookalike can offer better results for a fraction of the effort
  • Much finer targeting granularity: lookalike is a machine algorithm that considers hundreds of data points to find highly similar profiles, way more than any marketer would possibly consider

lookalike_audience-1Image source: Bamboo.

BUT: like any algorithm, what goes in affects what goes out. The higher the value of the profiles you feed the algorithm, the better the lookalike audience will perform. The highest ROI will be using you best customers as a basis. You can use a CRM platform like Splio to extract the profiles of your best customers matched with a WeChat profile and use them as a seed population. More on that below.

WeChat’s case study of Lookalike targeting for Education First shows a customer acquisition cost lowered by 22%. This is further confirmed by Lu Jin, BD Manager  from Digital Agency Westwin“Lookalike consistently shows higher ROI than regular targeting, we typically see a 20% performance increase and recommend using it if you have sufficient basis for a seed population.”


How is WeChat Lookalike targeting different from Facebook?

While the two interfaces look very similar, the biggest differences with Facebook targeting are

  • A higher minimum budget: the minimum Lookalike population size is 500,000 users, which will cost you a minimum of 50,000 RMB to target, while there is no minimum spend on Facebook and you can start to see results with as little as 500 USD
  • A larger seed population: while you can create a Facebook Lookalike population out of your 100 best customer profiles for a country, the minimum seed population for Tencent is 2,000 profiles. Both recommend a similar seed size range though (5,000 to 20,000).
  • Expansion size: while the top 1% matching population of a country shows best performance for Facebook, there is little documentation yet on the best performing expansion size range. Tencent’s official documentation states that an ideal expansion size for a 20,000 seed database is 1 Million users, and the accuracy significantly drops above 10 Millions.
  • Offline targeting: unlike Facebook, Tencent offers to create seed populations based on MAC addresses, as captured via in-store WiFi for example, which would typically allow for O2O campaign targeting. Moreover, if you use OpenIDs as a seed population, you can use any offline source as your seed population: WeChat WiFi users, followers from store QR codes, offline use Mini Programs etc.
  • Retargeting from previous campaigns: while Facebook only offers retargeting on Lead adverts, you can use as a seed with Tencent anyone who interacted with a previous campaign (clicks or conversions)



The Lookalike expansion interfaces look very similar. WeChat use absolute population size increments, and Facebook a proportion of a country’s Facebook population.

The most effective Lookalike targeting: using your customers as seed audience.

While you could use WeChat followers or people clicking on your ads as a seed basis for Lookalike targeting, using existing customers show the highest efficiency. In their Education First case study, Tencent used existing clients, and for a reason. We recommend to target customers bound on WeChat: since WeChat will match only users across the Tencent universe, you need to use phone numbers of customers that you have identified on WeChat (as followers, Mini Program users etc.) to maximize match of the seeds with actual WeChat user profiles. Splio’s Omnichannel Marketing Platform helps marketers identify and distinguish customers among followers to order to personalize engagement with customers and optimize acquisition strategy:

splio_best_customers_wechat    splio_wechat_filter

Top: creating a segment of your top customers matched on WeChat based on total purchases. Bottom: export the segment result database as a CSV file, ready to use for a lookalike targeting

To be noted: you can also use directly segmented WeChat IDs for targeting (OpenIDs, UnionIDs), although this is not offered on the advertising backend at the moment: you need to connect directly with the Tencent team or one of their accredited agency for such a targeting (they minimal budget required is usually higher than when using directly the advertising backend).

But while your Chinese customer base can be very large, what should you prioritize? Based on our experience running thousands of customer retention campaigns, here are the top sources you should consider:

  • Top purchasers (frequency): a list of people who have purchased from you most frequently
  • Top purchasers (average basket size): a list of people who have the highest spend
  • By location: a list of customers segmented by geographical clusters. Tencent’s Lookalike algorithm gives priority to similarity over location
  • Dynamic Mini Program audience: a list of all customers who have used your Mini Program over the last 30 days

With a 20% higher ROI on advertisement, lookalike targeting is to be considered if you spend above 200 to 500 thousand RMB on WeChat ads to justify for the overhead of segmenting and matching your customer base. It works best for brands having:

  • a customer database sizeable enough in China for a minimal seed size
  • with a significant proportion matched with WeChat profiles.
  • or who can capture high quality offline traffic to stores to be used as the seed population

WeChat Lookalike is quite similar to Facebook’s feature, but with a higher barrier to entry (budget and size of the seed population), and a higher focus on offline data as a source. And just like Facebook, the highest ROI is when using profiles of your actual customers as a basis for targeting. Splio helps brand marketers to segment their most valuable customers matched with a WeChat profile and have a maximal ROI on your WeChat advertisement.


To get regular insights on how to develop and strengthen your Loyalty strategy in China, scan the QR code to follow us on WeChat!