The reporting table has forty columns, most of them are on by default, and the two that actually answer your question are not next to each other. So you end up reading the numbers that are easiest to see rather than the ones that matter, and a campaign that is fine looks worrying while a campaign that is quietly wasting money looks busy and healthy. Reading results well is mostly knowing which columns to ignore.
This guide is about what to look at and what to distrust. It is deliberately concept-led: Ads Manager rearranges its columns and renames its metrics often, so rather than pointing at a layout that will change, it names the ideas that stay true and links Meta's own documentation for the current definitions. The one thing to carry throughout: the platform reports on its own performance, so read its numbers as evidence, not as a verdict.
Read the column that matches your objective
The single most common reporting mistake is judging a campaign by a metric it was never optimising for. If you told the system to get you cheap reach, its cost per purchase is not a fair grade. It was not trying to get purchases. If you told it to get purchases, its impressions are almost irrelevant.
So before you read anything, name the objective, then read the result column that belongs to it.
- An awareness or reach objective is judged on how efficiently it reached people and how many (cost per thousand impressions and reach), not on clicks or sales.
- A traffic objective is judged on link clicks and cost per link click, and on whether those clicks actually arrive as landing-page views rather than bounce before the page loads.
- An engagement objective is judged on the engagements it was set to chase, holding in mind that engagement is an activity metric, not an outcome.
- A leads objective is judged on cost per lead and, more honestly, on what those leads are worth once your own system sees them.
- A sales objective is judged on cost per purchase and return on ad spend, checked against your own records.
Meta defines each metric precisely and maps which ones attach to which objective in its Business Help Center. Read the current definition there before you build a report around a column, because a metric's name does not always mean what it sounds like: a "result" is defined by the objective you chose, so the same column heading means different things across two campaigns.
Distinguish the activity metrics from the outcome metrics
Some columns describe what the platform did. Others describe what happened to your business. They sit in the same table, in the same font, and the difference is everything.
Impressions, reach, video views at a platform-defined threshold, and bundled engagement counts are activity metrics: they measure delivery, not attention or outcome. They are not fake, and they are useful as diagnostics, but a campaign optimising towards them will faithfully produce them without necessarily producing anything you care about. Link clicks, landing-page views, leads and purchases sit closer to an outcome, and cost-per-outcome and return on ad spend sit closest of all. The further from an outcome a metric is, the less weight it deserves in a scale-or-kill decision, no matter how large and green it looks. This is the same distinction that reading ad results without fooling yourself works through in more detail, and it is worth reading alongside this one.
Use breakdowns to find the truth inside an average
A campaign-level number is an average, and averages hide the thing you most need to see. Breakdowns split a metric by a dimension (age, gender, placement, region, time, and others), and the split is where the decision usually lives.
The most valuable breakdown for most brands is by placement. A blended cost per result can look mediocre while one placement is carrying the campaign and another is dragging it down, exactly the mismatch described in the creative and placements guide. Splitting by placement tells you whether to fix a creative for a specific surface or stop delivering there. Age and region breakdowns can reveal that your results come from a narrower slice of the audience than you targeted, which is a targeting decision hiding inside a performance number.
One caution the honest reporter keeps in mind: some breakdowns cannot be combined cleanly with some attribution settings, because a conversion attributed across days or devices cannot always be assigned to a single breakdown value. Meta documents which breakdowns are available and how each interacts with attribution; check its breakdown documentation rather than assuming every split is exact.
Attribution: what the numbers can and cannot tell you
This is where confident reading goes wrong. A reported conversion is a claim made under an attribution model, not a measured fact, and the model has known blind spots.
Two matter most. First, the attribution window (the period after a click or view within which the platform will claim a conversion) is a setting, and the same campaign reports materially different numbers under different windows. Comparisons across accounts or across time are meaningless if the windows differ, so note the window on every report. Second, since Apple's App Tracking Transparency changed what platforms can observe on iOS, a portion of conversions cannot be directly measured and is instead modelled: estimated statistically rather than counted. Meta is open about this in its documentation; the practical effect is that some of the numbers in your table are modelled approximations wearing the same styling as counted events. Meta explains its current attribution and modelling approach in its Business Help Center, and reading their account of it is the responsible baseline.
The question no attribution model can answer is the important one: would these people have converted anyway? Reported return on ad spend, especially on retargeting, sits above the true incremental effect for exactly this reason. Treat platform-attributed revenue as an upper bound to be checked against your own system, not as money the campaign caused. Your order system, CRM or analytics is the denominator that keeps you honest.
A decision rule: scale, hold or kill
Reporting is only worth reading if it ends in a decision. A concept-led rule, applied per campaign rather than to a blended average:
Scale when a campaign has cleared its learning phase, is delivering the objective's result at a cost you are happy to pay, and has done so stably for long enough that the number is not noise. Scale by degrees, because a large budget change can send the ad set back into learning (covered in the budgets and bidding guide), so raise it in steps rather than all at once and let each step settle.
Hold when the read is not yet trustworthy: the ad set is still learning, the sample is small, or the window is too short to mean anything. Holding is a real decision, not indecision. Most campaigns are killed during a period when the only honest answer was "we do not know yet".
Kill when a campaign has had a fair trial past learning, on an adequate budget, and is delivering the objective's result at a cost you cannot sustain, and when a placement or audience breakdown does not reveal a fixable pocket inside the average. Kill the campaign, not your conclusion about paid social, and write down why, so the next test starts ahead of this one.
The discipline underneath all three is patience against your own reflexes. The system needs a stable period to be read at all, and reacting to noise is how brands both scale losers and kill winners.
Where Acumin fits
Acumin is deliberately honest about this surface. Today you read your ad results in Meta's own reporting. Automatic read-back of those results into Acumin is pending Meta's app review, so nothing in the product promises a closed loop that does not exist yet. Where Acumin does help is on either side of the number: keeping measured results separate from modelled claims, comparing only campaigns that have real recorded results and only flagging a reallocation when the gap is genuinely large, and never attributing revenue to a campaign through a model. A number produced that way would be an invention dressed as a measurement, which is the one thing the whole product refuses to do.
Do this today
Open your reporting table and delete every column that is an activity metric your campaign was not optimising for. Add the one result column that matches your objective, the cost per that result, and the attribution window. Then split your best-spending campaign by placement. Most people find, in five minutes, one placement quietly spending money it should not, and that is a real decision you can make today rather than a dashboard you can admire.