Your first week on a project for an IELTS prep centre in Almaty. On a call the owner says "the offer isn't landing", the manager complains about cold leads, and the media buyer replies that "on Meta's attribution it all pays off". You write the words down to sort out later. They are arguing about words rather than numbers: each of them means something different by them.
Below are the words a newcomer hears in the first month. Metrics are not here: CPM, CPL, CAC, ROAS, LTV, retention and churn are calculated on the same centre's numbers in the metrics glossary. Each word gets a short definition, an example and the lesson that covers it in depth. The numbers are illustrative and match the centre's Meta test.
Who we sell to and what we offer
Segment. A group of people with a shared job to be done who need their own offer. A schoolchild whose parents are preparing them for IELTS "for the future" and a school leaver who needs a score by a university's application deadline abroad are different segments. The second has an exam date, the first does not, and the parents decide for them. Splitting by age or district makes sense only if it changes something in the course or the ads.
Target audience. The segment you decided to start with and spend money and effort on. The centre picks applicants to universities abroad who already have an exam date, using the STP rules. The audience setting in the ad account is a separate thing that holds only Almaty and a minimum age; the ad itself picks the right people out of the feed.
Job to be done. Whatever the person is really buying for. A student who needs a score to get into university has to be ready by her exam date; "learn English" is your wording, not hers. You find the job by asking customers what they have already done, not what they would do. That is the idea behind Jobs To Be Done.
Offer. What the person gets, for how much and on what terms if they respond now. For the centre: a free trial lesson and level test, a three-month course at 45,000 ₸ a month, and a WhatsApp message from the manager within an hour of the lead. "Quality preparation with experienced teachers" is not an offer: no terms, no price. Broader than the offer is the value proposition: what the person gets, why they should believe it and what to do next.
Lead magnet. Something free and useful in exchange for contact details: a checklist, a test, a webinar recording. The centre gave away a checklist, "12 speaking mistakes", and the leads came cheap, but people came for the file, not the course. You don't sell the course in the first message to someone who took a lead magnet; you warm them up first with emails or retargeting.
Positioning. How the chosen people should remember you against the other options: the centre next door, a private tutor, preparing on YouTube. Not "the best centre in Almaty" but "IELTS preparation planned back from your exam date, not page by page through a textbook". The wording holds only while competitors don't offer the same, so you check it against their ads.
From the ad to the payment
Touchpoint. Any contact between a person and your business that a system noticed: an ad impression, a click, an email, a visit from search. There can be several before a lead, and how the credit is split between them is decided by attribution.
Funnel. The customer's path broken into stages, each showing how many people reached it. Models such as AIDA describe stages with words like "attention" and "interest"; in a report a stage is something you can count. The centre's test had 12,000 impressions, 171 clicks, 11 leads, five qualified and three at the trial lesson. You look along that chain for the stage that loses the most people, and that is called funnel analysis.
Lead. A person who left contact details and is waiting for a reply: submitted the form, filled in an instant form on Instagram, wrote on WhatsApp. Russian speakers say both «заявка» and «лид»; they mean the same thing. You count leads from your own lead list, without repeat submissions or colleagues' tests: for the centre that is 11 people, not the number of events in the ad account.
Qualified lead. A lead that meets a rule written down before counting. For the centre that is someone who replied on WhatsApp, is sitting the exam within six months and can attend on the centre's schedule, and five of the 11 leads qualify. Until the rule exists, the manager's "cold leads" and the media buyer's "decent leads" can't be compared. In B2B qualification has two stages: MQL, where marketing accepts the lead as a fit, and SQL, where sales takes it on.
Conversion. In an ad account it is the action the campaign counts as its result: a lead, a purchase, a registration. In a report it is the share of people who took the next step, written with both steps as the metrics glossary does.
CRM. A system where every lead has a card: where it came from, what the person replied, the next step and how it ended. Only the CRM shows who paid, so payments and revenue are counted from it, not from the ad account. While there are a dozen leads a week, a spreadsheet with the same columns is enough; when the manager can no longer remember who was promised what, you move to a CRM system.
Advertising
Creative. The image or video with the ad text. The centre's test has three creatives with one offer, each opening differently: with the problem ("pauses in speaking cost you points"), with the result and the price, and with what a lesson looks like. You compare them by what people did after the click, not by CTR: a performance creative is expected to bring a lead, not a click.
Landing page. The page the ad leads to, with one main action: at the centre it is booking a trial lesson. The first screen has to continue the ad: the person tapped an ad for a trial lesson and level test and should see them on the page straight away. Changing it one hypothesis at a time and checking whether it helped is the job of CRO and A/B testing.
Reach and frequency. Reach is how many accounts saw the ad at least once, and Meta estimates it rather than counting one by one. Frequency is how many times on average each of them saw the ad, impressions divided by reach. The centre's retargeting reached 2,300 accounts in two weeks at a frequency of 7.6: almost eight impressions each on average.
Retargeting. Ads for people who have already been on your site or profile. The centre shows them to people who opened the course page and did not submit the form, and excludes current students. Without that exclusion Arman, who paid for the course two weeks ago, will keep seeing the trial lesson invitation on Instagram.
Performance and awareness. A performance campaign works with people who are already searching or choosing, and it is judged by leads and their cost within a few days. An awareness campaign works with people who will need the product later, and its return takes weeks and months: reach first, then recall, then orders. The centre's lead test is performance; measuring it by reach makes no sense, any more than judging an awareness video by cost per lead (two rulers for two jobs).
Who brought the customer
UTM tags. Labels at the end of a link: where the visit came from, what kind of traffic it is and which campaign it belongs to. They tell GA4 and your lead list that a lead came from Instagram on the ielts_trial campaign. A team writes tags to one tagging scheme; otherwise one platform splits into three rows in the report.
Pixel. An ad platform's code on your site that tells the platform what a visitor did. The centre sends the Lead event through the Meta pixel when the server has accepted the form, and Meta uses those events to look for similar people. If Lead fires on a button tap, the campaign learns from people who like tapping buttons (pixels and conversions).
Attribution. The rule for dividing a lead or a sale between touchpoints. Every system has its own. A student saw the centre's ad on Instagram in the morning without tapping it, then found the centre in search in the evening and submitted the form. Meta will record the lead as its own, while GA4 knows nothing about the impression in the feed and gives it to search. Attribution divides credit; it doesn't tell you whether the person would have bought without the ad (attribution models).
Attribution window. The period after a click or an impression during which a system still credits the result to the ad. In Meta the default for new ad sets is 7-day click or 1-day view. If someone clicked, vanished and submitted the form ten days later, the lead lands in the CRM and stays out of Meta's count. That is one source of the gap between the ad account and the CRM.
End-to-end analytics. A report that puts spend, leads, payments and revenue side by side for every campaign. It is built as a chain: spend from the ad account, the click with UTM tags, the lead in the CRM, the payment. The ad account knows what was spent and the CRM knows who paid, and "which ads paid off" can only be answered once they are linked.
Money and time
Everything counted in tenge and percentages lives in the metrics glossary, and short definitions of CAC, LTV, ROAS and retention with formulas are in the Goodlabs glossary. The key distinction from there is the cost of a lead against the cost of a customer: a lead cost the centre 896 ₸, while a student cost 9,860 ₸ in ad spend.
Cost per lead and per student
Test ad spend: 9,860 ₸
CPL
11 leads
9,860 / 11 ≈ 896 ₸
Ad-only CAC
1 new student
9,860 / 1 = 9,860 ₸
One student paying for the first month after the test is an assumption.
Unit economics. A calculation per unit of the business: a student, an order, a subscription. How much the unit brings in after variable costs, and how much you can spend to get it. A student leaves the centre 16,000 ₸ a month out of 45,000 ₸, and the owner is willing to spend up to 24,000 ₸ to acquire one. From those two figures unit economics works out how many months a student takes to repay the acquisition cost. What remains of a sale after variable costs is called contribution throughout the course.
Cohort. A group of people with a shared starting point, such as students who first paid for the course in the same month. Cohorts are compared by age, not by calendar: the August students' first month against the October students' first month. Otherwise revenue per October student will look lower simply because their second month has not started yet (cohort analysis).
Exercise: your project's glossary
Take your own project or the test site from the Start track and write one line for each:
Now read the lines as the centre's owner would. "Women 25-45, Almaty" is an ad account setting, not a segment: there is no job in it, and it doesn't say who you leave out. Compare it with "applying to a university abroad, exam date already set". An offer with no price or terms, such as "a convenient schedule", won't tell the person what happens after the form or give the manager a way to open the conversation.
Every funnel stage needs a source. If the "leads" stage has none, within a week it will fill with events from the ad account instead of people from your list, and leads will grow without new customers. A qualified lead rule written after the count gets bent to fit the result, so you write it before the first lead. And you will need the attribution window the moment the client asks why the numbers in the ad account and the CRM don't match. Six filled-in lines make a ready crib sheet for your first call with the client, where you can check each line against what the word means to them.
Video: marketing metrics
The video is about metrics rather than the words in this lesson, so it pairs well with the metrics glossary.