Optimizing Facebook Ads: Quality Data vs. High EMQ Scores

Most Facebook ad buyers try to have the highest possible EMQ scores in their Meta Events Manager. What they don’t realize is that a little bit of great data is better than a lot of bad data.

There are two goals of sending data back to Meta via the Conversions API:

1️⃣ IDENTITY 💥 -> 🙋‍♂️
Help Meta tie website events back to a matched Meta user. This solves IDENTITY, and helps Meta know who to find more of.

2️⃣ ATTRIBUTION 💥 -> 📊
Help Meta tie website events back to a matched Meta ad click. This solves ATTRIBUTION, and helps Meta know which purchases to take credit for.

EMQ scores essentially give you more points for sending more data. And you generally want to send as much data as possible, as that will increase the probability of Meta matching to a user and matching to an ad click.

But let’s look at where the “more is better” mindset can break down.

✅ Sending a single parameter that actually matches to a user is all you need. If you only sent external id or only sent phone number, that’s likely enough to match to a user, as both of those are deterministic.

(Email is good, too, but it might not be the email Meta has on file for a person, given we all have so many email addresses)

✅ Sending a valid Facebook click id (FBC) is all you need to match back to an ad click. Literally that is it. And given every Facebook click id is unique to a single user’s click, it solves for both Attribution and Identity in one.

(Remember, FBC is only available for site traffic that started from Meta ad clicks, so its not available 100% of the time unless 100% of your traffic comes from Meta ads)

If you had a website data connection that only sent email, external id, and FBC for every purchase, you would probably have embarrassingly low EMQ scores around 5.0, but you’d match 100% of all possible users and ad clicks.

Meanwhile, I know of many data connectors that artificially inflate EMQ to be in the 8.5 - 9.5 range with AI-modeled data, fuzzy guesses, and enriched data that will never match to a user or a click.

🚩 platforms that artificially attach FBC to purchases: all this bad click data will create significant over-attribution

🚩 platforms that guess someone’s identity with ip-address-based fingerprinting: Meta will get mis-trained and find more of the wrong person

🚩 platforms that don’t correctly format phone/email values: Meta will give you EMQ score credit for sending a hashed value, but those values won’t ever match to a user

Remember, Meta doesn’t provide a score on the quality of data you are sending. They never tell what data actually matched a user. They don’t tell you which data was actually used for attribution (especially not 1-day view). They just give you a score for how much data you sent.

If you want to have a high score for the sake of having a high score, knock yourself out. If you want to have better ad performance and better ad attribution, make sure you have a quality data connection that gets all of these details right.