Metrics and analysis
Seven metrics that mislead and what to look at instead
ROAS, CTR, CPL, conversion rate and reach are all true and still lead to wrong decisions. Why each one misleads, and which metric to decide with.
Also available in: Português · Español
The metrics below are not false. They measure exactly what they say they measure. The problem is that what they measure is not what you are deciding — and using them as a criterion leads to systematically wrong choices.
1. ROAS
What it measures: attributed revenue divided by spend.
Why it misleads: the highest ROAS in the account is almost always remarketing, which talks to people already about to buy. And high ROAS on low volume is usually worth less than medium ROAS on high volume — 4 sales at 12x make less profit than 30 sales at 5x.
What to look at: absolute profit for the period, and ROAS split by function (prospecting, remarketing, brand). Comparing prospecting ROAS to remarketing ROAS is comparing different things.
2. CTR
What it measures: attention.
Why it misleads: curiosity drives clicks. A creative saying "did you know this?" has great CTR and terrible conversion. A creative stating the price in the first line has lower CTR and filters out people who were not going to buy.
What to look at: cost per conversion. CTR enters as a diagnostic when cost worsens — to tell whether the drop came from attention or from conversion.
3. CPL
What it measures: what it costs for someone to leave their contact details.
Why it misleads: it says nothing about who left them. A $12 CPL with a 3% close rate is worse than a $40 CPL with 20%.
What to look at: cost per sale. And if sales volume is too low to read, cost per qualified lead — but never CPL alone.
4. Landing page conversion rate
What it measures: how many of those who arrived filled in the form.
Why it misleads: it rises when traffic gets warmer, not necessarily when the page gets better. Cutting prospecting raises the rate and lowers the volume — and the report records it as an improvement.
What to look at: conversion rate by source. The site average mixes people arriving from brand search with people arriving from an Instagram video, and those two never convert alike.
5. Reach and impressions
What they measure: how many people, how many times.
Why they mislead: they are delivery metrics, not result metrics. They appear in reports because they are big and impressive. No budget decision should come out of them.
What to look at: nothing, in a performance report. Impressions and frequency enter as saturation diagnostics, never as results.
6. Platform conversions
What they measure: what the platform claims within its own window.
Why they mislead: each platform uses its own model and window, and all of them claim what they can. Adding up conversions from three platforms produces more conversions than there are sales.
What to look at: CRM sales with a source. And the ratio between what the platform claims and what the CRM recorded, which is your account's discount factor. See attribution windows.
7. Any average
What it measures: the center, hiding the distribution.
Why it misleads: an average cost per sale of $600 can be three campaigns at $400 and one at $1,800. The average is correct and the decision that comes out of it is wrong.
What to look at: the distribution. Median beside the average, and always the by-campaign detail before any conclusion about the account.
The metric that does not mislead
revenue recorded in the CRM for the period ÷ total spend for the period
It depends on no window, no attribution model, and no question of which platform claimed what. It includes what was never attributed, because it uses total spend in the denominator — which makes it conservative, the right direction in which to be wrong.
It is the metric that survives a meeting with finance. All the others are diagnostic tools for understanding why that one went up or down.
How to use each without being misled
| metric | good for | not good for |
|---|---|---|
| ROAS | comparing campaigns of the same function | deciding total budget |
| CTR | diagnosing fatigue | picking a winning creative |
| CPL | tracking volume | judging quality |
| Conversion rate | optimizing a page, by source | measuring overall improvement |
| Reach | nothing in performance | a results report |
| Conversions | platform optimization | reporting sales |
| Averages | a quick summary | deciding anything |
The general rule: platform metrics are for operating the platform; business metrics are for deciding money. Confusing the two is the origin of most bad decisions in paid media.
The one-question test
Before putting a metric in the report, ask: what decision do I make differently if this number doubles?
If the answer is "none", it is not a result metric. It is context, and its place is the appendix.
For the report structure that applies this rule, see the report clients actually read.
Frequently asked questions
What is the most reliable metric in paid media?
Revenue recorded in the CRM divided by total spend for the period. It is the only one that depends on no attribution window and no platform model.
Is a high ROAS always good?
No. High ROAS in remarketing usually reflects audience quality, not campaign merit, and high ROAS on low volume can be worth less than medium ROAS on high volume.
Why does CTR mislead?
Because it measures attention, not intent. Creatives that spark curiosity have high CTR and low conversion; creatives that filter have low CTR and high quality.
Is an average metric useful?
Rarely on its own. An average hides the distribution: an acceptable average cost can be a mix of one excellent campaign and one terrible one.