Metrics and analysis
Quality events vs volume events: which one to optimize for
Optimizing for signups fills the funnel with bad leads. Optimizing for sales can stall the campaign for lack of data. How to choose, and when to switch events.
Also available in: Português · Español
Every performance campaign makes a silent choice: which moment in the funnel counts as "success" for the algorithm. That choice defines what kind of person the platform will go looking for — and it causes half of the funnels that generate plenty of leads and no sales.
The trade-off in one sentence
A volume event (key page view, simple signup, WhatsApp click) happens a lot. The algorithm learns fast and delivery stabilizes. But it will go looking for people who complete that event — and completing a short form is something plenty of people with no intent happily do.
A quality event (sale, signed contract, paid subscription) happens rarely. The algorithm learns slowly or not at all. But when it does learn, it looks for people who buy.
The choice is not "which is better". It is which one you have enough data to sustain.
The 50-a-week rule
Meta's benchmark — and Google works similarly — is around 50 conversions per week per ad set to exit learning with stability.
Do the math before deciding:
ad set weekly budget ÷ estimated cost per event = conversions/week
| ad set spends | a sale costs | sales/week | verdict |
|---|---|---|---|
| $7,000 | $120 | ~58 | optimize for sales |
| $7,000 | $600 | ~12 | intermediate event |
| $1,400 | $600 | ~2 | intermediate event, no question |
Below 50, the ad set lives in limbo: irregular delivery, swinging costs, and no reliable read on which creative works.
The middle path: the intermediate event
Most accounts do not have the sales volume to optimize for sales, and do not want the junk that comes from optimizing for signups. The way out is an intermediate event: a step that happens more often than the sale and predicts sales better than the signup.
Examples, by business type:
| business | too much volume | intermediate | real quality |
|---|---|---|---|
| Local service | form submitted | appointment confirmed | service delivered |
| B2B | ebook downloaded | meeting booked | contract |
| High-ticket ecommerce | added to cart | checkout started | purchase |
| Insurance | signup | quote completed | policy |
| Education | free class signup | class attended | enrollment |
The middle column is almost always the right answer for anyone spending between $1,000 and $10,000 a month.
How to tell whether your intermediate event is any good
An intermediate event is only worth using if it correlates with sales. That is verifiable, not a matter of intuition:
- Take 90 days of CRM data.
- Separate those who completed the event from those who did not.
- Compare the sale rate of the two groups.
If people who book convert at 25% and people who merely fill in a form convert at 3%, booking is a good event. If both groups convert the same, the event is decorative — it carries no information about intent, and optimizing for it is optimizing for volume with extra steps.
Run this check by cohort, not on the cumulative total: an event may have been predictive last year and stopped being so after a site change.
When to move up the event (and when not to)
Move from volume to intermediate when cost per lead is good and sales do not show up. It is the classic symptom: the ads manager looks great, sales complains about quality.
Move from intermediate to sale when you comfortably clear 50 sale events per week in the ad set and offline attribution is working.
Do not move when the only reason is a desire to "optimize for what matters". Switching the event puts the ad set back into learning, and an ad set permanently in learning for lack of volume is worse than a stable ad set on an imperfect event.
Never switch the event mid-creative-test. You lose the whole comparison: you will not know whether the change in results came from the piece or from the event.
Send sales back, even without optimizing for them
There is one step almost nobody takes that improves everything: send the sale back to the platform via the Conversions API, even while optimization stays on the intermediate event.
The platform receives the quality signal, enriches the model, and starts to understand which profiles among your intermediate leads turn into sales. You get part of the benefit of sale optimization without paying the price of low volume.
That requires the CRM wired to the platform — the same work described in connecting Meta Ads to your CRM.
The summary
Pick the event closest to the money that still gives you volume. Check the correlation with sales instead of assuming it. Send sales back through the API even when you do not optimize for them. And only switch events when there is a reason, because every switch costs learning.
If the symptom is good CPL and no sales, the full diagnosis is in low CPL and no sales.
Frequently asked questions
How many conversions does an ad set need per week?
Meta's benchmark is around 50 per week per ad set. Below that, optimization is unstable and the algorithm cannot learn with confidence.
Can I optimize directly for sales?
If sales volume per ad set gets close to 50 a week, yes, and it is the best route. Below that it is safer to optimize for an intermediate event that correlates with sales.
What is an intermediate event?
A funnel step that happens more often than the sale and predicts sales better than a signup: qualified lead, booked appointment, conversation started, quote requested. The middle ground between volume and quality.
Does changing the event reset learning?
Yes. Switching the optimization event puts the ad set back into learning. So switch when there is a clear reason, not as a casual test.