How To Optimise Your Facebook Relevance Scores

Most native biddable media platforms include an element of quality assurance in their auction mechanics – the most well-known of these is Google’s Quality Score. Facebook is no different, assigning a Relevance Score (RS) out of 10 to each ad in a campaign. These scores have existed for a few years, and were surfaced in the front end in early 2015, so advertisers can now add a column to view this metric easily in the interface.

The Relevance Score is calculated after an ad has received more than 500 impressions, and is based on the ratio of positive to negative feedback on your ad. Positive feedback consists of actions such as likes, shares and campaign objectives, such as clicks to website, while negative feedback is calculated based on the number of people choosing to hide your ad.

The key difference between Facebook’s Relevance Score and Google’s Quality Score is at what level each score is calculated. While Google’s QS is calculated at impression level, RS is calculated at ad level. If you’re targeting a broad audience on the Google Display Network and on Facebook, your Google QS could vary between different users within that audience as it will be recalculated on each impression, but your Facebook RS is an aggregate score across that entire audience. In practice, this means that if you’re targeting a large audience on Facebook with a few users who respond very well to your ad, but a larger proportion of users who are indifferent or respond negatively, you could end up paying more than necessary to reach those positive users, as the larger negative cohort will bring your RS down.

How do I optimise Relevance Score?

Given the mechanics of Relevance Score, it’s key to segment your audiences where possible. If you’re targeting a 10% lookalike audience, consider splitting out the top 1% or 2% of users into a separate ad set. These users should in theory respond better than the rest of the audience, and by splitting them out, you’ll get a separate relevance score for them, which should hopefully be higher. This would then either reduce your cost to reach that audience, or increase the available reach for your budget, depending on what your aim is. Of course, remember to cross-exclude your audiences, as your top 1% of lookalikes will automatically fall into your 10% audience too. Excluding your 1% audience from your 10% ad set will ensure they filter into the correct ad set.

Once you’ve begun segmenting audiences, think about how to boost engagement with more relevant ads. For example, if an advertiser is selling travel insurance, they could segment their audience by age range and target different creatives at each group. This should boost positive engagement, increasing your Relevance score.

It’s also important to refresh your creative to prevent ad fatigue. If a user sees the same ad cropping up for a sustained period of time, they’re more likely to perceive it as a nuisance and opt to hide it. New messaging and images can help combat this.

Relevance Score in Action: Advertiser reduces CPA by 47%

Last year, we took over a Facebook campaign for a client in the education vertical. Previously the campaign had been running in-house and had heavily relied on one large lookalike audience to target users. It suffered from incredibly low Relevance Scores, with almost every ad having a score of 1 or 2.

We began by splitting out the audience into different levels of lookalike audience, from the top 1% of lookalikes through to the top 10%, giving them their own ad sets. This raised the relevance scores immediately, and the CPA began to come down.

We then mined the best performing audiences using Facebook’s Audience Insights tool to come up with other targeting methods. We combined these with more relevant ads for each audience – for example, we used different USPs for younger users.

Within 9 weeks, the Relevance Scores had increased from an average of 1/10 to 9/10 and CPA had decreased by 47% to just £0.46.

Overall, the key to optimising Facebook Relevance Scores lies in audience segmentation. It's crucial to ensure that you're not overbidding for a highly-responsive audience by including them in a much larger audience which is less inclined to interact with your ad. This, combined with targeted ads which are regularly updated, can lead to healthy improvements in campaign efficiency.

To read this blog written by James Howard on the Periscopix website, please click here.