Broad vs. narrow targeting: Audience size, the most underrated user acquisition lever for success.

“Should I create a 2% Lookalike in the US with an audience size of 4.7M or is a 3% Lookalike with a 7 Million audience size better?”

“Should I exclude some devices, especially on Android, or even run on a selected list of devices, or should I not apply any device targeting at all?”

“Should I target a specific age range, which I believe is the one that performs best for the game, or should I run without any age targeting at all?”

“How many competitors should I bundle up for this interest targeting campaign?”

These are just a few of the questions & micro-decisions user acquisition managers face on a daily basis. Of course, there are several factors that come into play. But, why are questions like these particularly crucial to the success of our UA campaigns? The basis of all those questions is that they increase or decrease the size of the audience we are targeting. Why is this increasingly important today? We are living in the age of algorithmic marketing. As platforms such as Facebook and Google become more and more sophisticated, using their data, statistical models & machine learning, we marketers are ceding more and more control, trusting them to match the high value users to our games and apps.

With advertising platforms wanting us to give them more and more control of optimizations, a direction which makes sense given the increased sophistication of their products, the nature of a UA specialist’s job is evolving, as many of us know. In addition to focusing on the increasingly important levers that marketers have such as creative, growth managers will also need to turn their attention to giving the platforms the right kind of data they need but also the right volume of data they need to maximize targeting & auto-optimization effectiveness. This is where the audience size & broadness spectrum topic comes into play. How broad should my audience be in order for the algorithms to have a large enough pool of users & therefore enough data to then go and target the most valuable users for us, instead of us, marketers, doing the micro-targeting & optimizations manually.

When a UA specialist optimizes campaigns manually, they rely on aggregate data and averages to analyse performance & then take actions. On the other hand, because of the amount of first-party data they have, platforms like Facebook & Google, know how each individual user is valued before they are even targeted.

“On Facebook this month, Italy is seeing a ROAS D7 of 30% versus our target of 40%”.

A logical action when a UA manager is optimizing manually might be to decrease investments in Italy or even pause activity there completely. Knowing how the Facebook algorithm works, wouldn’t it be more efficient & effective to give Facebook the power to optimize & target only the high value users in Italy which contributed to this ROAS? One way to do that for this specific example is to bundle up Italy in a campaign with a larger pool of GEOS, so that the Facebook algorithm can have a big enough audience to “cherry pick” & target the high value users we want at the scale that we want.

It is not to say the broader the targeting, the better because the algorithms will have more data. On the one hand we want them to have, through aiming for a large audience, enough data to be as effective as possible in their targeting, but on the other hand, we want:

1. To set some kind of boundaries in audience definition to slightly guide the machine.

2. To apply some targeting layers to be able to tailor our message & creative.

Marketers need to constantly try & keep the right balance between narrow and broad targeting throughout their UA activity.

1. Setting boundaries to guide the machine

The platforms’ algorithms cannot effectively operate 100% autonomously. Of course, some platforms work more autonomously than others. For instance, Google, famously being called a “black box”, gives marketers less levers to “guide the machine” than Facebook, especially regarding audience size. But still, both platforms do, in different ways. UA managers need to think of audience sizes as a way to set boundaries & slightly guide the algorithms in their optimizations.

We could take a few of the questions we asked ourselves in the beginning of this blog post in order to illustrate this idea:

“Should I create a 2% Lookalike in the US with an audience size of 4.7M or is a 3% Lookalike with a 7 Million audience size better?”

Would we want to limit our audience size to the 4.7M users which look the closest to our seed or are we okay to broaden the Lookalike & go slightly further away from the seed but in turn giving Facebook a bigger audience to then have more room to optimize?

It is worth mentioning, that for Facebook specifically, the machine will need a different audience size based on what it is optimizing towards: Installs, events or ROAS. However, the overarching concept stays the same: What’s a good audience size, here defined by Lookalike size, that won’t be too narrow for FB to have enough data to be effective in their auto optimization or won’t be too broad for Facebook to be efficient?

“Should I exclude some devices, especially on Android, or even run on a selected list of devices, or should I not apply any device targeting at all?”

Is it better to limit the list of targeted devices based on the averaged data we have on device performance or does it make more sense to not apply any device targeting in order to give algorithms more data to then do the optimizations on their side with the user level data they have. The answer might lie somewhere in between.

A certain device model might show from our data analysis, lower performance in the form of a lower average ARPPU for instance, but Facebook might have a pool of users which have this exact device that actually would possibly generate a relatively high ARPPU, reaching our performance targets. This pool of users would automatically be left out if we were to blacklist this device. Therefore, does it make more sense to leave this decision to Facebook to make, by still including the device in our targeting?

Of course, if the game or product is not technically compatible with a list of devices, then those should be excluded from the targeting.

2. Narrowing the audience for creative tailoring

Thinking about where we should be on the spectrum of broad vs. narrow targeting, with creative becoming an increasingly important lever for UA, it is vital to also think about assets in the context of audience & audience size and vice versa:

How generic vs. tailored should the creative be?

How generic vs. tailored should the messaging be?

All of the above will depend on the type of game of course, but this is still an important topic to think about, no matter where the game sits on the casual spectrum. Does a generic creative or message resonate well with everyone, even if you consider your overall audience to be very broad? How does this affect the audience size we are targeting in a specific campaign or ad set? With even the biggest hypercasual players such as Voodoo slightly moving towards the other side of the hypercasual to casual spectrum, questions like these will become increasingly important.

How do we set those boundaries in audience definition?

The answer of course is not straight-forward. What must be done is constant testing. Testing different audience sizes in creatives ways, using all the levers that the different advertising platforms have available, layering & not layering certain targeting capabilities which impact audience size, all while also keeping in mind that the creative strategy needs to also be taken into consideration & adapted when thinking about audience size. Different creative strategies need to be also tested for the different audiences in terms of how broad vs. narrow the assets appeal & messaging should be.

A very important question this whole topic also brings is how must we then segment our overall audience & structure our campaigns in order to tackle this subject in the best way.

One way SHAMSCO is addressing this topic with some of our mobile gaming clients is by using one of our frameworks to help them build a robust & structured audience testing plan and campaign structure, which also then helps them shape conversations around their creative strategy.