Ad Metrics

For years the answer to "how is the advertising doing?" was "response has been good." The report a platform sent back listed how many times the ad appeared and how many times it was clicked, and then it stopped. What the person who clicked did on the site, how many of them signed up, whether anyone who signed up ever paid — the ad records and the product records lived in separate places with no way to join them.

So quarterly reviews arrived with the middle missing. The marketer would say thirty thousand clicks, the CEO would ask how much money that made, and nobody in the room had the sentence in between. There wasn't even a basis for deciding whether thirty thousand clicks was a lot, so neither praise nor criticism had anything behind it. When the question was whether to keep spending, the loudest person in the room won.

Joining the two records changed the conversation. Once the trail ran from the moment an ad appeared, through who clicked, through whether the person who arrived signed up, through whether they were still paying three months later, performance stopped being one lump and became a series of questions. That is why these numbers are arranged as steps. When results are bad, the point is not to conclude that advertising doesn't work but to find which step people fell out of.

The order of the descent is fixed too. Is it showing at all, do people click what shows, do the people who arrive sign up, do those who sign up return more than they cost — checked from the top down. Fixing the landing page while impressions are near zero wastes the week, and so does swapping creative while impressions and clicks are both healthy. Because the steps exist, the place to touch narrows to one. It also settles the meeting where three people each want to fix a different thing: check from the top step and the disagreement gets decided by a number rather than by opinion.

The metrics split into three groups following the order in which an ad reaches a person: reach, which measures how far it traveled; response, which measures whether the people it touched moved; and profitability, which measures whether that movement left money behind. A weak group makes the numbers below it unreadable. A conversion rate of 2 percent on a campaign with only a thousand impressions says nothing — it means fifty people arrived and one signed up, and a second signup would have printed 4 percent. The ratios lower down only hold still once the counts higher up are large enough.

Reach covers how many times and to how many people an ad appeared. Three numbers interlock.

1. Impressions — how many times the ad rendered on a screen. It counts renders, not people, so the same person seeing it three times counts as three.
2. Reach — how many distinct people saw it. The same person seeing it ten times counts as one.
3. Frequency — how many times the average person saw it. It's impressions divided by reach, so 300,000 impressions across 60,000 people is a frequency of 5.

The three are read together because impressions alone can't describe the situation. If impressions double and reach doubles with them, the campaign is meeting new people. If impressions double while reach holds flat, the campaign is pressing twice as hard on the same people. The second pattern degrades over time: past a frequency of about 3, the click rate on the same creative usually starts sliding, which makes this number the alarm that tells you to swap the creative. It's also why impressions and reach get read side by side whenever budget goes up — to confirm the extra money bought new people. A campaign that has run out of new people to reach needs a wider audience or a different placement, and no amount of extra spend fixes it from inside the same setup.

Response covers whether the people who saw the ad actually moved. Two numbers measure different stretches.

1. Click-through rate (CTR) — the share of impressions that got clicked. Ten thousand impressions and 120 clicks is 1.2 percent. This judges whether the ad itself belongs in that placement.
2. Conversion rate — the share of arrivals who went on to the intended action. A thousand visitors and thirty signups is 3 percent. This judges the page they landed on and the product, not the ad.

They're read together because responsibility divides between them. A low click rate points at the creative and the format; a healthy click rate with a low conversion rate points away from the ad and at the page people arrive on. Copy that lifts only the click rate usually drags conversion down with it. Multiplying the two gives the odds that one impression ends in a signup, and the real decision about raising budget comes from that product together with the cost numbers below it. A campaign at 1.2 percent and 3 percent multiplies out to 0.036 percent; one at 3 percent and 0.6 percent multiplies out to 0.018 percent, so the first is twice as good despite the worse-looking click rate.

Profitability covers whether the people who arrived returned more than they cost. Five numbers make one sentence.

1. Return on ad spend (ROAS) — how much revenue each unit of ad spend produced. Spend $2,500 and take in $7,500 and that is 300 percent. It fits products where the payment happens on the spot.
2. Customer acquisition cost (CAC) — what one customer cost to bring in. Add ad spend and the labor behind it, then divide by the customers who arrived in that period.
3. Lifetime value (LTV) — the estimate of everything one customer will pay before leaving. A customer paying $16 a month who stays fourteen months on average is worth $224.
4. LTV to CAC ratio — lifetime value divided by acquisition cost. A $224 customer acquired for $56 is a 4x, and the rough industry line is that anything under about 3x doesn't justify scaling the spend.
5. Payback period — how many months it takes to earn back what you spent. Spend $56 and collect $16 a month and you are square in 3.5 months.

This is where reading ROAS alone breaks down for a subscription product. Only $16 arrives in the month someone signs up, so first-month ROAS comes in under 30 percent, while counting all fourteen months makes it a 4x business. In exchange, the company has to hold separate cash to acquire the next customer while that money trickles in. That's what the payback period is for. At 3.5 months, a dollar put in comes back around three times a year; at the same multiple with a 14-month payback, it doesn't come back even once. The multiple says whether the business works, and the payback period says how fast it can grow. Companies spending ahead on raised money watch both for exactly that reason.

"Click rate is a fine 1.8 percent but conversion is 0.4. We should be looking at the landing page, not the ad."

People are arriving, so the problem sits outside the ad. It moves the work from creative to the page.

"Reach is flat and impressions are up. We're showing the same people over and over."

Frequency is climbing, which warns that more budget will buy repeat views rather than new people.

"A 300 percent ROAS isn't the end of it. We sell subscriptions, so read the payback period."

The money doesn't arrive at once, so a single revenue ratio can't settle the question.

The quarter Pagedeck pushed ad spend to about $15,000 a month, Ravi fixed an order for reading the dashboard top to bottom and walked it every Monday. In week four clicks were up 30 percent over the previous week while signups hadn't moved. Working down the steps, impressions had doubled and reach had risen 8 percent. The campaign hadn't met new people; the same people had seen the ad six times each, and the extra clicks came from repeat viewers finally tapping once.

The next week Ravi widened the audience and changed the creative. Frequency fell from 6 to 2.4, clicks actually dropped, and signups went from 40 a week to 58. When Grace asked in the quarterly review why clicks had fallen, the answer Ravi gave wasn't one metric but the whole staircase. Sofia added the detail that settled it: the customers arriving now mentioned where they had seen the product before anything else on the call, and the previous quarter's customers hadn't.

After that Ravi wrote the weekly report in step order only. Putting the good numbers first and burying the bad ones lower down, he argued, leaves the reader unable to tell where the blockage is. With the order fixed, the question in the meeting changed from how performance was to which step was blocked.

The most common error is reading a high click rate as a good ad. It's the first thing on the dashboard and it visibly moves, so the instinct is natural. But the number measures how much curiosity the ad created, not what it promised. "For anyone stuck writing meeting notes after hours" will always out-click "$9 a month to keep your team's documents in order." Switching to the first line often takes the click rate from 1.2 percent to 3 percent while conversion falls from 3 percent to 0.6. Leave it there and the quarter ends with more clicks, more spend, and the same number of signups — and because the report says the click rate improved, it usually goes unnoticed until someone counts signups separately months later.

Reading the platform's own numbers as actual results is the other frequent error. The conversion count sits right there in the ad account, so there seems to be nothing to doubt. But each platform counts anyone who passed through its ads as its own. A person who saw a search ad and a feed ad and then signed up gets written down once by each. Some platforms count people who only viewed an ad without clicking, as long as they sign up within a few days. Add up the conversions across platform dashboards and the total exceeds real signups. At Pagedeck two platforms reported 74 a week between them while the product database gained 51 accounts. Miss that gap, conclude both channels are efficient, and raise both budgets, and spend doubles while signups rise 20 percent. The basis for the decision has to be the record on the product side. This doesn't mean throwing the platform numbers away: they're useful for comparing campaigns inside one platform, and they stop being trustworthy the moment they're summed across channels.