Marketing metrics: 12 terms from CPM to churn with tenge examples

A week after an Instagram test, the media buyer messages the owner of an IELTS prep centre: "CPL 896 ₸, half the plan." The owner replies with one question: "And how much did a student cost me?" CPL does not answer that. Mix the two numbers up on your first project and the client will soon decide you are measuring something other than what he pays for.

Below are twelve metrics you cannot do without on a first project. All of them are calculated for one business: the centre's Meta test from week 2 and the centre's economics from the scaling calculation. The numbers are illustrative. Check the definitions against the Goodlabs glossary, which adds industry benchmarks to the formulas.

The starting numbers

We added the payment ourselves: there was none during the five days of the test. In the centre's history one in three people who come to the trial lesson buys the course, so we assume that a week later one of the three paid for the first month.

Impressions and clicks: CPM, CPC, CTR

CPM = spend / impressions × 1,000. For the centre that is 9,860 / 12,000 × 1,000 = 822 ₸ per thousand impressions. In Ads Manager this is the CPM (cost per 1,000 impressions) column. CPM shows what the audience's attention costs, and you need it to compare months, placements, or the price of a post in someone else's Telegram channel with Instagram ads. Impressions are not people: one person may have seen the ad three times.

CPC = spend / clicks. 9,860 / 171 = 58 ₸. Meta calculates CPC (cost per link click) from link clicks only: to a website, a form, an app store. A cheap click says nothing yet about the lead: what matters is what the person did next.

CTR = clicks / impressions × 100%. 171 / 12,000 = 1.4%. Ads Manager has two kinds of clicks. Clicks (all) also counts likes, comments and opening a photo full screen, while Link clicks counts only click-throughs. Meta calculates CTR (link click-through rate) from link clicks, and that is the one to use in your table, or an ad that collects likes will look better than it is.

CTR is handy for comparing creatives on the same audience, but it is too early to pick a winner by it. In the centre's test the best CTR, 1.6%, belonged to the "Problem" ad, and each person it brought to the trial lesson cost 4,900 ₸. The "Result and price" ad had a CTR of 1.3%, and each person who came cost 1,650 ₸. With one and two people at the trial you cannot name a winner anyway, but CTR pointed to the other ad.

From click to payment: CR, CPL, CPA

CR (conversion rate) = actions / people at the previous step × 100%. It is the share of people who took the next step, and without naming both steps the number means nothing. The centre has three conversion rates in a row:

• click to lead: 11 / 171 = 6.4%;
• lead to trial lesson: 3 / 11 = 27%;
• trial lesson to payment: 1 / 3 = 33%.

The glossary puts website visitors in the denominator, Google Ads uses interactions with the ad, and we divided by clicks. GA4 may show fewer sessions than there were clicks, because some people close the tab before the page loads. So a report should not say "CR 6.4%" but "click to lead 6.4%".

CPL = spend / leads. 9,860 / 11 = 896 ₸. In Ads Manager this is Cost per result when the campaign's result is a lead. Take the leads from your own list, where you have already removed repeat submissions and colleagues' tests. The ad account counts events and, when data is missing, fills in part of the results with modelling, while you need people with phone numbers.

CPA = spend / chosen actions. Always name the action. CPA per lead is just CPL, and per person who came to the trial lesson the centre gets 9,860 / 3 = 3,287 ₸. The owner is willing to pay up to 8,000 ₸ for a person at the trial lesson, so on this number the test passed with room to spare. Write a bare "CPA 3,287 ₸" in a report and the client will read it as the cost of a student, then ask a month later why students cost three times as much.

The student and the money: CAC, AOV, ROAS

CAC = acquisition costs / new customers. Counting ads only, the centre has 9,860 / 1 = 9,860 ₸ per student against an allowed 24,000 ₸. The glossary also puts sales costs in the numerator. Say the manager spent half a day on the 11 leads, which is 5,000 ₸ of salary. Then CAC = (9,860 + 5,000) / 1 = 14,860 ₸. In a report, label which of the two CACs you calculated.

One student proves nothing yet. If a second person from the trial had paid, CAC would have halved to 4,930 ₸. All one payment can tell you is that the cost of a student has not gone over the limit yet.

AOV (average order value) = revenue / number of payments. The centre sells the course month by month, every payment is 45,000 ₸, and the average order value is 45,000 ₸ too. But it depends on what you count as an order. If the centre lets people pay for three months at once, such a student makes one payment of 135,000 ₸ instead of three of 45,000 ₸. The average order value goes up, and the centre receives exactly the same money. For courses and subscriptions it is more useful to look at what a student brings in over the whole time, which is LTV.

ROAS = revenue from ads / ad spend. The student paid for the first month: 45,000 / 9,860 = 4.6, or 456%. Google Ads, in its description of the Target ROAS strategy, calculates it the same way: conversion value divided by spend.

But revenue is not what the centre keeps. Of the 45,000 ₸ that is 16,000 ₸, or 36%. Ads pay for themselves when ROAS is at least 45,000 / 16,000 = 2.81: the glossary describes this breakeven point as one divided by the contribution margin share. The centre's ROAS is 4.6, so in the first month the student leaves 16,000 − 9,860 = 6,140 ₸ on top of the ad cost. If he stays for all three months, the same spend brings in 135,000 ₸ of revenue and ROAS becomes 13.7. That is why a ROAS figure always comes with a note on which revenue and which period it uses.

The picture shows the same thing on two campaigns with the same budget. Campaign A has the higher ROAS, but variable costs eat almost all of its revenue, and after advertising it is in the red.

The video for this lesson also covers ROMI. People calculate it in different ways; one option is (contribution − marketing costs) / marketing costs. For the centre's first month that gives (16,000 − 9,860) / 9,860 = 62%. Agree with the client on which formula you use before the first report.

What a student brings in: LTV, retention, churn

For these three metrics one student is not enough; you need a group of people who started at the same time (a cohort). Say 12 students paid for their first month, and one in three left after it, as in the scaling calculation: four left and eight finished the course.

Retention = paying in month N / cohort size. In the second month that is 8 / 12 = 67%, in the third also 67%. Active here means someone who paid, not someone who is still in the group chat. Write down what counts as active before you calculate.

Churn = left during the month / students at the start of the month. For the first month 4 / 12 = 33%, for the second 0 / 8 = 0%. First-month churn and second-month retention add up to 100%: they are the same 12 people, split into those who left and those who stayed. After that it no longer works: third-month retention, 67%, accounts for everyone the cohort has lost since the start, while second-month churn, 0%, counts only those who left during that month. What adds up to 100% with retention is cumulative churn: 4 / 12 = 33%.

LTV is what a student brings in over the whole time. The glossary gives the formula ARPU × margin × 1 / churn. It is meant for an open-ended subscription where the same share leaves every month. The course lasts three months and all four leavers went after the first one, so it is simpler to add up month by month. Contribution: four students left 16,000 ₸ each, eight left 48,000 ₸ each, on average (4 × 16,000 + 8 × 48,000) / 12 = 37,333 ₸ per student. Revenue: (4 × 45,000 + 8 × 135,000) / 12 = 105,000 ₸.

Ad decisions are made on contribution, not revenue: advertising is paid for out of what the centre keeps. The owner spends up to half of the contribution on ads. While everyone was assumed to finish the course, that was 24,000 ₸ per student. With the dropouts, half the contribution is 18,667 ₸, and the allowed cost per lead falls from 2,000 to 1,556 ₸, as in that same calculation. The test's ad-only CAC of 9,860 ₸ fits under this limit as well. On real data, groups like this are built through cohort analysis.

All 12 metrics in one table

Task: a table of 12 metrics

This is a week 0 task on the Start track, and you will come back to the table in week 4, when you calculate customer payback.

• Write down all 12 metrics: the name, the formula and, in one sentence, the question it answers.
• Add an example in tenge to each. If you have your own test, use its numbers. If not, take a business near you and make up numbers, labelling them as illustrative.
• For CR, CPA and ROAS, note which steps, which action and which revenue you used.
• Check every definition against this lesson and the Goodlabs glossary.
• Recalculate every example from scratch without looking at your first calculation.

The table is done when all 12 metrics have a formula and an example in tenge, and each one has been checked against the lesson and the glossary.

When you check, look for five mistakes. The first is CTR based on Clicks (all): an ad people like looks better than the rest. The second is a CPA with no action named, which the client will read as the cost of a buyer. In ROAS it is easy to put revenue for the whole course when the student has paid only for the first month. A CAC based on one or two buyers cannot be presented as the norm, because the very next buyer will move it noticeably. And monthly churn does not equal 100% minus retention: retention is cumulative, while churn covers a single month.

Video: a glossary of metrics