Facebook has seen a lot of changes recently, most of which are fueled by the iOS shifts in Apple’s App Tracking Transparency (ATT) update brought on by iOS 14. Facebook estimates conversions have been underreported by approximately fifteen percent, with more impact in Q3 than in previous quarters. This is perhaps the largest hurdle to overcome, as lack of accurate data for each advertiser makes it difficult to scale and target considering over sixty percent of iOS users are opting out. Apart from the lack of conversions, Facebook recently announced more changes that affect audience targeting and campaign attribution. We’ve outlined the main sets of changes below and how best to respond:
Here are the times Facebook has changed attribution.
- Historically – to April 27th: 1-day view, 28-day click (actual data)
- April 27th to June 30th: 7-day click (7-day click actual data,1-day actual and modeled data)
- June 30th – September 23rd: 7-day click (7-day actual and modeled data)
- Starting September 23rd: 1-day view, 7-day click (for new campaigns)
With all these changes, it can seem daunting to keep track of and plan for next steps, but we’re here to help.
- First, it is important to log these changes, and monitor any performance fluctuations and gaps that have occurred.
- Second, advertisers should use any available third-party tracking systems such as Adobe, DCM and Google Analytics (GA) to use as supplemental sources of data.
- Most importantly, stay aware of these changes and their potential impact.
FMPS and other supplement sources and reporting should be keeping the same windows for consistency and end of year reporting, and advertisers need to account for all these different windows with some caveats for comparison reference points. Special attention should be given to review data before April 27th, comparing conversions in different attribution windows, to provide a reference point by looking at the percentage of conversions attributable to the period between seven and twenty-either days.
Facebook made updates to conversion campaign audience targeting, as well, in order to mitigate the impact of data loss. While this will help more advertisers benefit from the platform’s powerful AI tools, it is not necessarily a benefit to all advertising use cases. While the opt out rate is high, Facebook in response has made updates to the targeting feature in platform.
Starting this week, Facebook will update all conversion campaigns to have both detailed targeting expansion and lookalike expansion on in all ad sets.
Unlike in previous iterations, there is no option to toggle these features off, and Facebook touts that this expansion will reduce costs, increase audience sizes and increase campaign performance. This is like the conversations surrounding Campaign Budget Optimization (CBO), with Facebook pushing towards more automation for advertisers, in an era where there is little transparency from Apple and more modeling will be necessary to fill in the gaps.
With Detailed Targeting Expansion, the AI for audience targeting will target individuals with interests similar to the those you have already selected, but only if the system predictably finds it is likely to have a positive impact on performance. Lookalike Expansion works the same way, except it expands your Lookalike audiences beyond the standard 1-10% range. Overall, the impact is that Lookalikes and Detailed Interest segments will increase and expand beyond audience parameters we set.
However, what is new is that Facebook will now automatically opt advertisers and brands into adopting this. This automation, and trust in the platform optimization, will add a layer of system accountability, and lessen the need for super granular targeting as the system will discover users outside of the segment. It also reduces control of your targeting. Please note that age, gender, and location will still be set and cannot be altered, and the automated Detailed Targeting and Lookalike Expansion only applies to Conversion campaigns optimizing to conversions, value, or app events (these features have never been available for Awareness or Reach campaigns).
We will continue to monitor impact and can exclude interests/segments to try to reduce some broad segments. Before launching Facebook did do some initial testing with positive results on costs. We are staying positive, but vigilant with changes:
- Conversion campaigns using Detailed Targeting Expansion, saw a 37% decrease in median cost per incremental conversion compared to using Detailed Targeting only.
- Conversion campaigns using Lookalike Expansion with a Website or Mobile Custom audience seed, saw a 17.3% decrease in median CPA compared to using Lookalike Targeting only.
- Conversion campaigns using Lookalike Expansion with a Customer List Custom audience seed, saw a 10.1% decrease in median CPA compared to using Lookalike Targeting only.
- Make sure you update your windows to the latest attribution, as these have been updating these rapidly.
- Do not make many frequent edits to existing campaigns, to allow the system to optimize and avoid continuously going into the learning phase. Analyze reporting at campaign level, if possible, since estimated conversions are aggregated with a delay, which can impact ad performance.
- Make sure you prioritize the lowest funnel events in event configuration
- Consider implementing Conversion API, as this will be mandatory in 2022.
- Analyze Reporting after a period of at least 72 hours and on a broader scale (such as on the ad set or campaign level).
- Consolidate ad sets to give campaigns more data to work with and facilitate the learning phase.
- Test and learn to continue to improve campaigns, and implement finding from split tests, across the board.
- Costs should go down for conversion campaigns due to Lookalike and Detailed Targeting Expansion. However, if you’re using quality metrics like applications, or third-party segments like GA for supplemental data, be sure to monitor quality as well as cost with the expansion changes.
- Monitor the impact automated expansion has on your campaigns going forward. Consider implementing a few pertinent exclusions right away (such as income brackets that may not be relevant to your audience), but don’t exclude too much.
If you have additional questions or any concerns about these changes, please contact us. We are happy to help!
POV By Gellena Lukats, Director, Paid Social.