TL;DR
Most Meta ads metrics are noise. The 12 that matter: CPA and ROAS (outcomes), CTR and thumbstop rate (creative), CPM (auction), frequency and first-time impression ratio (fatigue), CVR and AOV (downstream), spend and impressions (sanity checks), and MER or blended ROAS (the truth serum). Everything else is either a derivative of these or a vanity metric designed to make dashboards look busy. If a metric can't change a decision, stop tracking it.
The 200-column problem
Open Ads Manager's column customizer and you'll find around 200 metrics. This is not generosity. It's a trap. The more metrics on screen, the more likely you are to find one that's going up, feel good about it, and ignore the three that are going down.
We run every account on roughly 12 metrics. Not because we're minimalists, but because every additional metric has a cost: attention. Each column you add is attention stolen from a column that matters. Here's our cut, grouped by what each metric actually tells you.
Outcome metrics: did we make money?
1. CPA (cost per acquisition). The price you paid per conversion. The single most important number for lead gen and most ecommerce. Everything else is commentary on why this number is what it is.
2. ROAS (return on ad spend). Revenue divided by spend, as reported by Meta. Useful, but remember it's platform-reported revenue, which means it's subject to every tracking caveat in existence. Always cross-check against your actual books. ROAS that looks great in Ads Manager while the business bleeds is the oldest trick in the agency playbook.
These two are the scoreboard. Every other metric exists to explain movements in these two. If someone presents you 40 metrics without CPA or ROAS front and center, they're hiding something.
Creative metrics: is the ad any good?
3. CTR (click-through rate, link). The percentage of impressions that became link clicks. CTR measures whether the ad earned attention and interest. Falling CTR with stable CPM means the creative is tiring, full stop.
4. Thumbstop rate (3-second video views divided by impressions). For video, this isolates the hook: did the first three seconds stop the scroll? We judge hooks on this before judging anything else. A video with a great thumbstop rate and weak conversion is fixable. A video nobody stops for is firewood.
5. Hold rate (thruplays divided by 3-second views, roughly). Of the people who stopped, how many stayed? This separates hook problems from body problems. Great thumbstop plus terrible hold means the opening works and everything after it doesn't. Most creative "analysis" never gets this granular, which is why most creative feedback is useless.
Auction and delivery metrics: what did the market cost?
6. CPM (cost per thousand impressions). What you paid for eyeballs. CPM moves for reasons unrelated to your creative: seasonality, competition, Meta's inventory shifts. When CPA rises but CTR and CVR hold, CPM is usually the culprit. Check it over 30 to 90 days, not 7.
7. Frequency. Average impressions per person. The fatigue early-warning system. Climbing frequency plus falling CTR is the classic creative exhaustion signature. In prospecting, sustained frequency above 3 with CTR decay means your audience has seen enough.
8. First-time impression ratio. The share of impressions reaching people who haven't seen your ads before. When this drops, you're paying to re-show ads to the same users. Meta buries this in the Inspect tool, which tells you how little they want you thinking about it.
Downstream metrics: did the business hold up its end?
9. CVR (conversion rate). Clicks that became conversions. When CTR holds but CVR drops, the problem is downstream: landing page, offer, checkout, or competitive pressure. This metric is how you stop blaming creative for landing page problems and vice versa.
10. AOV (average order value). Revenue per order. Rising CPA hurts less when AOV rises with it. AOV also reveals whether your ads are attracting the right customers or just bargain hunters. If AOV is sliding while conversion volume holds, check your offer and audience, not your ads.
Sanity metrics: is the data even real?
11. Spend and impressions. Boring. Essential. Before analyzing any ratio, check the raw volumes. A 50% CTR improvement on 200 impressions is a rounding error, not an insight. Every percentage without a volume check is a potential lie.
12. MER / blended ROAS (marketing efficiency ratio). Total revenue divided by total ad spend, across all channels, from your books, not Meta's dashboard. This is the truth serum. Platform ROAS can look heroic while MER declines, which means you're just moving credit between channels. If MER is healthy, the business is healthy. Everything else is instrumentation.
The vanity metrics hall of shame
Now the other side. Metrics that look important, change nothing, and exist mainly to fill agency reports:
- Relevance score / quality ranking. Meta's own grades, computed opaquely, unactionable in practice. You can't optimize a black box's opinion of you.
- Reach. "Look how many people saw it!" Saw it and did what, exactly? Reach without outcomes is just expensive wallpaper.
- Engagement (likes, shares, comments) on conversion campaigns. Nice when it happens, meaningless as an optimization target. The most engaging ad we ever saw sold nothing.
- Video views (undifferentiated). A "view" can mean 3 seconds of autoplay with the sound off. Thumbstop and hold rate tell you what happened. Raw views tell you Meta's video player functioned.
- Landing page views as a success metric. It's a diagnostic step between click and conversion, not an outcome. Celebrating landing page views is celebrating people who almost did something.
The test is simple: if a metric moves and you wouldn't change a single decision because of it, it's vanity. Delete the column. Your dashboard will be shorter and your thinking will be clearer.
How to actually read them together
Metrics are useless in isolation. The skill is reading them as a system:
- CPA up + CTR down + frequency up = creative fatigue.
- CPA up + CTR stable + CVR stable + CPM up = expensive auction.
- CPA up + CTR stable + CVR down = landing page or offer problem.
- ROAS great + MER flat = you're stealing credit from other channels.
- Everything volatile + low volume = you don't have data yet, stop analyzing noise.
Learn these five patterns and you can diagnose 90% of account situations in about 60 seconds. The remaining 10% is where audits earn their keep.
FAQ
How many metrics should I have in my Ads Manager view? Enough to see outcomes, creative health, auction costs, and data sanity at a glance. Our default view is the 12 above. Anything more is usually procrastination disguised as analysis.
Is ROAS or CPA more important? Depends on the business model. Ecommerce with varied cart values: ROAS. Lead gen or single-product with fixed value: CPA. Anyone who tells you one is universally correct is selling a template.
Why does my Meta ROAS not match my actual revenue? Attribution windows, view-through credit, tracking gaps, and the fact that Meta takes credit generously. Use platform ROAS for directional optimization and MER from your books for business decisions. Never confuse the two.
What is a good CTR for Meta ads? It varies wildly by industry, format, and placement, which is why universal benchmarks are mostly useless. What matters is your CTR trend: stable or rising is fine, steadily falling is fatigue. Compare against your own history, not someone's blog post.
Should I track different metrics for prospecting vs retargeting? Same metrics, different expectations. Retargeting will show higher frequency, higher CTR, higher CVR. That's normal. The diagnostic patterns still apply, just calibrated to the warmer audience.
What is thumbstop rate and where do I find it? Three-second video views divided by impressions. It measures whether your hook stops the scroll. You'll need a custom column or calculated metric in Ads Manager, which is frankly ridiculous given how important it is, but here we are.
Written by the MetaMaxd team. We run Meta ads for brands that are tired of agency theater: no vanity metrics, no "hacks," just accounts engineered to make money. If your dashboard has 200 columns and no answers, that's what our audits are for.
