Everyone knows that great data is needed for strong ad performance on Meta and TikTok - but 90% of ad buyers donโt actually know how to assess their data quality. This is why EMQ score obsession causes more harm than good.
The biggest mistake is a more-is-more mindset that falsely assumes bigger numbers are better (while too much bad data leads to over-attribution). Instead you should be striving for a goldilocks mindset (not too much, not too little) with strong signs of data alignment.
When I run pixel, data, and attribution audits for brands, here's what I look for with the Meta or TikTok Events Mangers:
๐๐๐ฒ๐ป๐ ๐๐ผ๐น๐๐บ๐ฒ ๐ฎ๐น๐ถ๐ด๐ป๐บ๐ฒ๐ป๐
Check the volume of actual site traffic or actual purchases from your site analytics. Does this match the volume of events you're receiving? If not, you have duplicate connections or pixel install issues.
โ
Look for 1:1:1 alignment between volume of actual purchases, browser purchase events, and server purchase events
โ Too little data will lead to weak ad performance
โ Too much data will lead to over-counting issues
๐๐น๐ถ๐ฐ๐ธ ๐ถ๐ฑ ๐ฐ๐ฎ๐ฝ๐๐๐ฟ๐ฒ ๐ฟ๐ฎ๐๐ฒ ๐ฎ๐น๐ถ๐ด๐ป๐บ๐ฒ๐ป๐
Click ids are arguably the most important data type to send to your ad account. While you want the capture rate of click ids to be high, the natural limit is the percentage of events that actually came from ad clicks.
(ie. 30% of site visitors from Meta ads defines a natural limit of 30% capture rate of FB click ids)
โ
Look for 1:1 alignment between your volume of site traffic from paid ads and the volume of events with click ids
โ Low capture rates will lead to missing data and weak ad performance
โ Capture rates that exceed the natural limit suggest click id tampering (and will cause massive over-attribution)
(I won't name names, but two popular CAPI connection providers who entered the market in the past year are guilty of this. Beware of platforms who artificially game EMQ scores, but are actually just causing over-counting.)
๐๐ฟ๐ผ๐๐๐ฒ๐ฟ ๐๐ ๐๐ฒ๐ฟ๐๐ฒ๐ฟ ๐ฒ๐๐ฒ๐ป๐ ๐ฎ๐น๐ถ๐ด๐ป๐บ๐ฒ๐ป๐
Ideally you want to have your data connection handle both browser events and server events to ensure the same data is being sent from both sources and with 100% effective deduplication. (Popsixleโs latest release provides fully aligned browser + server events).
โ
Look for event ids being passed 100% of the time with deduplication rates that are over 90%
โ Deduplication rates in the 70-90% range are sub-optimal
โ Deduplication rates below 70% may lead to over-counting issues
โโโ
Know someone who needs help with data alignment? Drop a comment, DM, or use the booking link on my profile to set up time together.