How to Read an Ad Report Without Being Fooled

Most ad reports are not dishonest. They are just badly sorted. Thirty numbers arrive in one PDF with no indication of which ones matter, and the eye lands on the biggest figure — usually reach or impressions — because big numbers feel like progress. Meanwhile the number that should decide next month's budget is sitting in row 14, unlabelled and unexplained.

Every metric in an ad report belongs to one of three groups. Vanity metrics describe the size of the megaphone. Diagnostic metrics tell you where a campaign is leaking. Decision metrics are the only ones that should make you spend more, spend less, or stop. Sort the report into those three piles and it becomes readable in about four minutes.

The vanity pile

These are real numbers. They are measured accurately. They just do not answer any question you have.

• Impressions and reach. How many times your ad was served, and to how many people. If you raise the budget, these go up. That is all they tell you. They cannot go down while spend goes up, which is exactly why they make such comfortable reporting.
• Engagement and engagement rate. Likes, shares, comments and reactions on a paid post. Useful as social proof on an ad that will run for months. Not useful as a performance verdict, because engagement on a paid post is often driven by people who like arguing, not buying.
• Video views. Meta has historically counted a view at three seconds. Three seconds is roughly the length of a thumb slowing down. A report proudly listing 80,000 video views is telling you 80,000 thumbs slowed down slightly.
• Followers gained. A traffic campaign will pick up followers incidentally. Nice. It has no bearing on whether the campaign made money.
• Click-through rate on its own. CTR — the share of people who saw the ad and clicked it — is genuinely diagnostic, but only when read next to what happened after the click. Reported alone it is vanity, because a shock-value creative can double CTR and halve revenue.

What I would not buy: a monthly retainer whose deliverable is a deck of platform screenshots totalling reach, impressions and engagement, with a line at the bottom saying awareness grew. Any agency can generate that report without touching the account. If the report does not reconcile against something in your own records, it is a wallpaper.

The diagnostic pile

These numbers do not tell you whether the campaign works. They tell you where it stops working, which is far more useful when something is wrong. Read them as a chain, in the order a stranger actually moves.

Cost per thousand impressions (CPM)

What it costs to be seen a thousand times. This is your auction price. If CPM jumps 40% in a fortnight with no change from you, something external moved — festive season bidding, a new competitor, a shrunken audience. It explains a rising cost per lead that has nothing to do with your creative.

Frequency

Average number of times one person saw the ad. On a small local audience — say a 5 km radius around a Laxmi Nagar showroom — frequency climbs fast. When the same person has seen the same creative nine times, performance decays and no amount of bid tinkering fixes it. Frequency is the metric that tells you the problem is creative fatigue, not targeting.

Click-through rate, read properly

Low CTR with normal CPM means the creative or the offer is not interesting to this audience. High CTR with poor outcomes means the ad is promising something the page does not deliver.

The gap between link clicks and landing page views

This is the single most under-read row in any Meta report, and it is where I would look first. Link clicks count the tap. Landing page views count the page actually loading. When 1,000 clicks produce 640 landing page views, you paid for 360 people who gave up while your page loaded. That is not an ads problem, it is a site speed and page build problem — often a hero image straight off a phone at four megabytes. Google's Core Web Vitals guidance treats a Largest Contentful Paint above 2.5 seconds as needing improvement; on a mid-range Android over 4G, an unoptimised page can easily be triple that.

Conversion rate on the landing page

Of the people who genuinely arrived, how many did the thing. If this is weak while CTR is strong, stop editing the ad and edit the page.

The decision pile

Two or three numbers belong here. They are the ones tied to money that actually entered your business.

• Cost per qualified lead. Not cost per lead — cost per lead your sales team would agree is worth a call. A campaign at ₹180 a lead where four in five are students doing research is worse than one at ₹520 a lead where half convert.
• Cost per sale, and the margin on that sale. If your average order contributes ₹1,400 after cost of goods, a ₹900 cost per sale is a business. A ₹1,600 cost per sale is a hobby, however healthy the dashboard looks.
• Return on ad spend, reconciled. Only for e-commerce, and only after you have compared platform-reported revenue against your own order records for the same dates.

That reconciliation is the whole game, because platform-reported conversions and your books will not match, and the mismatch is not fraud. It is attribution. Meta's default setting has commonly been a seven-day click and one-day view window, meaning someone who saw your ad on Monday and bought on Saturday gets counted — and the same sale may also be claimed by Google Ads if they searched your name before buying. Both platforms also model some conversions, filling gaps left by iOS privacy changes and cookie loss with estimates. Add those up and you can get 140% of your actual revenue claimed by your ad accounts.

I will say plainly: nobody has a clean solution to this. Anyone selling you perfect cross-channel attribution for a business doing a few lakh a month in revenue is selling confidence, not measurement. The practical answer is cruder and works: use your own records as the denominator. Total ad spend last month against total new customers last month, from your own CRM or order register. Platform figures then become directional — useful for comparing two ads against each other inside one platform, not for judging the channel's total worth.

A sorting table you can keep

Three questions to ask whoever sends the report

You do not need to learn the platforms to hold a report to account. Ask these.

1. Which number here would make you stop a campaign? If the answer is vague, the report has no decision metric in it and is decoration.
2. What attribution window is this, and does it include view-through? The person running the account should know without checking.
3. How many of these leads did we actually speak to, and how many bought? This is the question that ends vanity reporting permanently. It also requires that lead source be captured at the point of enquiry — a hidden source field on the form, or a simple automation that stamps every enquiry with where it came from. Without that, nobody can answer honestly.

One caution about killing campaigns too fast. Both platforms need a volume of conversion events before delivery stabilises; Meta's learning phase is documented as needing roughly 50 optimisation events per ad set per week. A brand-new ad set judged on three days of data is being judged on noise. Give it a fortnight or a defined spend before the decision pile gets a vote.

What to do next

Open last month's report. Take a pen and write V, D or X next to each line — vanity, diagnostic, decision. If there are no X's, that is your finding, and it is worth more than the report. Then pull your own count of new customers for the same month and divide total spend by it. That single figure, however rough, is the honest one.

If you want a second pair of eyes on an account and a reporting format that leads with decisions instead of reach, we do that as part of paid social and creative work, and you are welcome to send us the last three months of numbers and ask what we would change. More on measurement and ads on the blog.

Questions we get asked