Marketing Analytics Glossary

Total marketing + sales spend divided by new customers acquired in the same period.

Always specify the channel, the period and what the spend includes. An ad-only CAC, without sales salaries, is a valid variant as long as the report says so. Blended CAC mixes paid and organic and hides which channel is actually working. Compare CAC to LTV: a healthy ratio is roughly 1:3 (LTV at least three times CAC). For Kazakhstan fintech, watch out for one-off bonuses - a 5,000 ₸ welcome cashback inflates first-purchase CAC but can pay back over 6 months.

The total revenue (or gross profit) you expect from a customer over their lifetime.

Measured in revenue, gross profit or contribution (revenue minus all variable costs, not only the cost of goods). The last two are usually more useful: ads are paid for out of what the business keeps. Calculate as ARPU × gross margin × average customer lifespan, or run a cohort revenue curve and project. A lifespan of 1 / churn fits an open-ended subscription that loses the same share every month; for a product with a fixed term, such as a three-month course, sum the cohort's revenue or margin month by month and divide by the cohort size. Don't use blended LTV for ad bidding decisions; segment by acquisition channel because retention differs sharply.

Revenue generated by ads divided by ad spend.

A 4× ROAS means ₸4 of revenue for every ₸1 spent. ROAS is a vanity number on its own - it doesn't account for COGS, returns, or fulfillment costs. Use POAS (Profit on Ad Spend) when shopping platforms underweight margin, or compare ROAS to your contribution-margin breakeven: revenue per sale divided by the contribution per sale. For illustration, a 45,000 ₸ course month that leaves 16,000 ₸ after variable costs breaks even on ads at a ROAS of 45,000 / 16,000 = 2.8125 (about 2.81).

Average ad spend per single conversion (lead, install, purchase).

Closely related to CAC but usually narrower: CPA per a specific conversion event vs CAC for "becoming a paying customer." In ad platform UIs CPA usually appears as "Cost / Conversion." Always pair with the conversion definition - "lead" can mean form-fill or qualified call.

Ad spend divided by the number of leads.

Example: 9,860 ₸ spent and 11 leads give 9,860 / 11 = 896 ₸ per lead. In Meta Ads Manager it appears as "Cost per result" when the campaign's result is a lead. Take the lead count from your own list, with repeat submissions and colleagues' tests removed: the ad account counts events, and you need people with phone numbers. CPL is CPA with a lead as the action. It is not what a paying customer costs: in the same example one of the 11 leads paid, so the ad-only CAC came to 9,860 ₸.

Clicks divided by impressions. On Meta, link clicks divided by impressions.

Which clicks count depends on the platform. In Google Ads they are clicks on the ad. Meta has two rates to tell apart: "CTR (all)" also counts likes, comments, shares, clicks to the Page or profile and full-screen photo or video opens, while "CTR (link click-through rate)" counts only link clicks. Use the link rate, or an ad that people like will look better than it is. Most useful for comparing creatives within the same audience. Industry CTR benchmarks are noisy - your historical baseline matters more. A sudden CTR drop with stable spend usually means audience saturation; refresh creative before bidding more.

How much you pay for one thousand ad impressions.

A signal of audience demand and creative quality (algorithms reward higher engagement with lower CPM). Programmatic and YouTube use CPM bidding by default. Watch for CPM inflation in Q4 (Black Friday → New Year) and during election cycles in markets with political ad budgets.

Ad spend divided by clicks. On Meta, by link clicks.

Example: 9,860 ₸ spent and 171 link clicks give 9,860 / 171 = 58 ₸ per click. Meta's "CPC (cost per link click)" counts only clicks that take people somewhere: a website, a lead form, an app store. "Cost per click (all)" divides the same spend by all clicks, likes and comments included, so say which one your table uses. A cheap click says nothing yet about leads: what matters is what the person did after it.

Share of customers from a cohort still active in a later period.

Measured by cohort: out of users who signed up in March, how many were still active in April, May, June. Retention shapes how long the cohort keeps generating revenue. Define "active" by the action that matters for your business - login is not the same as a paid transaction.

Share of customers at the start of a period who left during it.

First-period churn and second-period retention add up to 100%: they split the same people into those who left and those who stayed. After that the sum breaks, because retention is counted against the original cohort and churn against the customers at the start of each period. Example: 4 of 12 students leave after month one, so month-one churn is 33% and month-two retention is 67%. Nobody leaves in month two: churn for that month is 0%, while month-three retention stays at 67%. Tracked monthly for SaaS/subscription, weekly for marketplaces. Differentiate voluntary churn (canceled) from involuntary (failed payment) - the latter is fixable with dunning automation. Churn under 5% monthly is decent for B2C subscription, under 1% is excellent.

A group of users who share a starting event (signup week, first purchase month).

The single most valuable view in product analytics. A cohort retention table reveals whether changes you ship actually improve long-term behavior, separate from acquisition mix shifts. GA4, Amplitude and Mixpanel all support cohort tables out of the box.

Total revenue divided by active users in a period.

Use Monthly ARPU for SaaS, Weekly ARPU for high-frequency apps. Going up with stable users = monetisation working; going up because heavy churn dropped low-value users is a false positive - always pair with retention.

Distinct users active in the last 30 (MAU) or 1 (DAU) days.

DAU/MAU ratio is "stickiness" - how often the typical monthly user comes back. 50%+ is excellent (used roughly every other day), 20% is solid for niche tools, under 10% means a notification or content problem.

The first time a user experiences the product's core value.

Quantify it: Facebook's was "7 friends in 10 days," Slack's was "2,000 messages sent." Find yours by comparing retained vs churned cohorts and looking for the threshold. Then redesign onboarding to push every new user toward that exact action.

When the product reliably solves a real problem for a defined market.

Sean Ellis's test: ask users "How would you feel if you couldn't use this product?" If 40%+ say "very disappointed," that's the threshold. Without PMF, growth tactics paper over the gap. With PMF, retention curves flatten and word-of-mouth becomes a real channel.

The single output metric that best predicts long-term value creation.

Should reflect customer value (not just revenue) and be measurable at high frequency. Airbnb: "nights booked." Spotify: "time spent listening." Pick one - having three is having none.

Awareness, Acquisition, Activation, Retention, Revenue, Referral.

Dave McClure's framework for breaking the funnel into stages each with their own metrics. Useful for diagnosing where growth is stuck - most "marketing" problems are actually activation or retention problems further down the funnel.

Google's event-based analytics product, replacing Universal Analytics in 2023.

Every interaction is an event with parameters; sessions are derived, not primary. Standard (free) GA4 has no monthly event cap: the 10M hits a month in Google's terms applies to Universal Analytics. The limits you will hit are elsewhere. The daily BigQuery export is capped at 1M events a day, and Google pauses it if you go over (streaming export has no volume limit). Data retention (2 or 14 months) applies only to explorations and funnel reports; standard reports are unaffected. Set it to 14 months when you create the property. Server-side measurement via Measurement Protocol or sGTM narrows gaps from ad blockers.

A tag management system that lets marketers deploy tracking without engineering for each change.

Container holds tags, triggers and variables. Three tag types you'll use 95% of the time: GA4 Event, Meta Pixel, custom HTML. Always work in Preview mode and version everything - every Publish creates a restorable version.

A JavaScript array your site pushes events into so GTM can listen.

The single most important contract in any tracking setup. Define the schema before writing tags: which events fire, what parameters they carry, naming conventions (snake_case is standard). A messy dataLayer is the #1 cause of "GTM stopped working."

A documented list of every event the product emits, with parameters and triggers.

Lives in a Google Sheet or Notion table that engineering, analytics and marketing all reference. Columns: event name, trigger condition, parameters (type, example), destination (GA4, Meta, internal). Without one, you ship inconsistent events that break reports six months later.

Five URL parameters (source, medium, campaign, term, content) that label traffic.

GA4 reads them as session-level dimensions. Lowercase Latin, words joined by underscores - and never tag internal links. Common mistake: tagging the same campaign as `meta` from one team and `facebook` from another, splitting the report into two ghosts.

Share of visitors who complete a target action.

Always specify the numerator (which conversion) and the denominator (sessions, users, or qualified visitors). In a funnel the denominator can also be the people at the previous step, such as click to lead or lead to payment; then name both steps ("click to lead 6.4%") rather than writing "CR 6.4%". E-commerce industry CR averages mask huge segment differences - first-time vs returning, mobile vs desktop, paid vs organic. A useful baseline is your own moving 90-day average; meaningful improvement is a 20%+ relative lift sustained across two cohorts.

Total revenue divided by number of orders.

Easier to grow than CR for many businesses. Levers: bundles, free-shipping thresholds, tiered cashback, post-purchase upsells. Watch median alongside mean - a long tail of large B2B orders can hide that the typical order is shrinking.

Rule for assigning credit for a conversion across multiple touchpoints.

Last-click gives all credit to the channel closest to purchase, such as branded search, which makes it misleading for budget decisions. GA4 uses data-driven attribution by default, as do most Google Ads conversion actions: it weights touches using your own conversion data, and standard GA4 has it, not only 360. Since November 2023 GA4 offers only data-driven and last click; first click, linear, time decay and position-based were removed. If you build attribution yourself, position-based (40/20/40) is a reasonable middle ground.

Gives 100% of the credit for a conversion to the last channel the user clicked through before converting.

Easy to compute and dangerous to use for budgeting: the channel that closes the sale, such as branded search, takes the credit even when other channels created the demand. GA4 has two variants. Paid and organic last click ignores direct traffic and credits the last channel clicked through; Google paid channels last click credits the last Google Ads click and falls back to the first variant when the path has none. Always run a holdback experiment for any "high-ROAS" channel before doubling its budget.

Attribution that distributes credit across all touchpoints in a journey.

Linear, time-decay, position-based and data-driven are the common variants. MTA depends on cookies - which iOS ATT and 3rd-party cookie deprecation have crippled. Pair with MMM (Marketing Mix Modeling) for the budget-allocation question.

Server-side event sending from your backend directly to Meta, bypassing the browser.

Recovers signal lost to ad blockers, iOS ATT and consent rejections. Send the same event from both pixel and CAPI with a deduplication ID; Meta merges them. Without CAPI, your Meta ad cost-per-conversion looks 30-50% worse than reality on iOS.

Conversions sent from your backend (CRM, payment) to ad platforms via API.

Used for offline conversions (loan approved, contract signed) and to bypass browser-side gaps. Google Ads has Offline Conversion Imports + Enhanced Conversions for Leads; Meta has CAPI; TikTok has Events API. For fintech KZ, S2S is essential - most loan approvals happen days after the click.

A self-hosted GTM container that runs on your servers, intercepting events before they reach platforms.

Lets you enrich, redact, deduplicate and route events server-side. Runs on Google App Engine, Cloud Run, or self-hosted Docker. Initial setup is fiddly but pays back with faster page loads, first-party cookies, and recovered iOS signal. Budget $30-100/month for App Engine.

A tiny script (often invisible 1×1 image) that an ad platform places on your site to collect events.

Meta Pixel, TikTok Pixel, LinkedIn Insight Tag and Yandex Metrika are all variations. Each platform has its own naming for events ("Purchase" vs "purchase" vs "transaction") - getting them wrong silently breaks ad optimization. Test in the platform's native debugger before going live.

Data you collect directly from your customers - email, behavior, purchases.

The strategic asset of the post-cookie era. Build a Customer Data Platform or at minimum a unified customer table in BigQuery. Hashed first-party emails feed Customer Match (Google) and Custom Audiences (Meta) for cookieless retargeting.

Apple's 2021 prompt asking users to opt into IDFA tracking. Most say no.

Without the IDFA, ad networks get iOS install attribution from Apple's SKAdNetwork 4 or AdAttributionKit (interoperable with SKAN, iOS 17.4+): anonymous, delayed postbacks, up to three of them, for conversion windows of days 0-2, 3-7 and 8-35. Server-side events (Meta CAPI, TikTok Events API) complement them. Web tracking on iOS Safari is similarly hobbled by ITP, and server-side collection holds up better than browser tags.

Google's page-experience metrics: LCP, INP, CLS.

LCP (Largest Contentful Paint) under 2.5s, INP (Interaction to Next Paint) under 200ms, CLS (Cumulative Layout Shift) under 0.1. These influence Google rankings for both desktop and mobile and correlate with conversion rate. Defer GTM, lazy-load images and self-host fonts as the three biggest wins.