Adray

Glossary

Marketing attribution glossary: 50 terms explained

Fifty terms you will meet when you measure marketing, each defined in one or two sentences in plain words. Where an article explains a term in depth, the definition links to it.

Attribution models and windows

Attribution

Giving credit for a sale or a lead, by a set rule, to the marketing touchpoints that came before it. It decides how the credit is shared, never how many sales there were.

Multi-touch attribution

Attribution that splits the credit for a sale among several touchpoints instead of giving all of it to one. Linear, time decay, position-based and data-driven are multi-touch models.

Last-click attribution

Gives all of the credit to the last touchpoint before the purchase. Simple and useful for day-to-day decisions at the bottom of the funnel, and blind to whatever created the demand.

First-click attribution

Gives all of the credit to the first touchpoint, which shows what brings new people in and ignores everything that closed the sale.

Linear attribution

Splits the credit evenly among every touchpoint in the journey.

Time-decay attribution

Gives more credit to the touchpoints closest to the purchase and less to older ones.

Position-based attribution

Favours the first and the last touchpoints, and gives less of the credit to the ones in between.

Data-driven attribution

Lets an algorithm assign the credit from patterns in the platform's own conversion data. Google Ads and GA4 now offer it and last click as their only models, and it only sees the touchpoints that platform can see.

Touchpoint

Any interaction a person has with your marketing before converting: an ad seen or clicked, a search, an email opened, a visit.

Customer journey

The sequence of touchpoints and visits a person goes through between first hearing of you and buying. Attribution models are rules for splitting the credit along it.

Attribution window

How long after an interaction a conversion still counts for it, set separately for clicks and for views, as in "7-day click, 1-day view".

Read more: What a 7-day click, 1-day view attribution window actually means

Click-through conversion

A conversion credited to an ad the person clicked, within the click window.

Read more: What a 7-day click, 1-day view attribution window actually means

View-through conversion

A conversion credited to an ad the person saw but never clicked, within the view window. It exists in the platform's report and can never appear in your site analytics.

Read more: What a 7-day click, 1-day view attribution window actually means

Self-attribution

Each ad platform reports the conversions it can credit to its own ads, under its own rules, so several platforms can claim the same sale and their totals add up to more than you sold.

Metrics

Conversion

The action a campaign is meant to produce, such as a purchase, a lead or a sign-up. For a store, the conversion that matters is a settled order.

Read more: 15 attribution metrics and the question each one answers

ROAS (return on ad spend)

Revenue attributed to a channel or campaign divided by what you spent on it. A platform's ROAS uses its own attribution; verified ROAS uses your store's settled orders.

Break-even ROAS

The ROAS at which ads pay for the products they sell and nothing more: 1 divided by your gross margin. With a 40% margin it is 2.5.

POAS (profit on ad spend)

Gross profit of the sales attributed to ads divided by ad spend. Above 1, the ads paid for themselves after the cost of the products.

MER (marketing efficiency ratio)

Total revenue divided by total marketing spend over the same period, with no attribution involved. Honest as a total, and silent about which channel produced it.

CPA (cost per acquisition)

The cost per conversion a platform reports, counting every purchase it credits to itself, from new and returning customers alike.

CAC (customer acquisition cost)

What it costs to win one customer: acquisition spend divided by the customers acquired in the same period.

New-customer CAC

Acquisition spend divided by the people whose first settled order ever falls in the period. Dividing by all customers instead makes acquisition look cheaper than it is.

CPL (cost per lead)

Ad spend divided by the leads it produced. Easy to optimize and misleading on its own: the number that matters is the cost per closed deal.

LTV (customer lifetime value)

The value a customer produces over their relationship with your store. For ad decisions, measure it as gross profit over a fixed horizon, such as 12 months.

LTV:CAC ratio

Lifetime value divided by acquisition cost. 3:1 is a common rule of thumb, not a law.

Cohort

A group of customers who share a starting point, usually the month of their first order, followed over time to see what they actually produced.

New customer

Someone whose first settled order in your store's entire history falls in the period. Ad platforms use their own definitions, so their counts differ from yours.

Read more: Why your platforms and your store disagree about who is a new customer

Verified revenue

Revenue from your store's own orders, net of cancellations and refunds, as opposed to the revenue an ad platform claims.

Read more: How much of what Meta claims can you actually verify?

Settled order

An order that was paid and was not cancelled or refunded in full. The unit every verified metric should count.

Tracking and data

Pixel (tag)

A snippet of code on your site that reports visits and actions to an ad platform or an analytics tool from the visitor's browser.

Read more: What the Conversions API actually fixes, and what it does not

Conversions API (CAPI)

A way to send conversions to an ad platform from your server instead of the browser, so they survive ad blockers and closed tabs. It improves what the platform knows about a sale; it does not verify the platform's claim.

Read more: What the Conversions API actually fixes, and what it does not

Server-side tracking

Sending events from your own server rather than the visitor's browser, either through the platforms' conversion APIs or through a tag manager container on your own subdomain, known as server-side tagging.

Click ID (gclid, fbclid)

An identifier the ad platform adds to the landing URL when someone clicks an ad: gclid from Google, fbclid from Meta. Keeping it with the visit, the lead or the order is what ties that result to the click.

UTM parameters

Tags added to a link, such as utm_source, utm_medium and utm_campaign, that tell your analytics which campaign sent the visit. They describe the link; they do not prove the click caused the sale.

Event ID

A unique identifier sent with a conversion, such as the order number, so a platform that receives the same purchase from the browser and from the server counts it once.

Read more: How to tell if the same purchase is being counted twice

Deduplication

Recognizing that two reports of an event are the same one and counting it once. Without it, the same purchase is counted twice: once from the browser and once from the server.

Read more: How to tell if the same purchase is being counted twice

Hashing (SHA-256)

Turning an email or a phone number into a fixed string that cannot be read back, so a platform can match it to its own users. It must be normalized and hashed exactly as the platform expects, or it is accepted and matches nobody.

Read more: What the Conversions API actually fixes, and what it does not

Enhanced conversions

Google's feature that adds hashed customer data (email, phone, name, address) to the conversions you already track, so Google can match more of them to signed-in users who engaged with your ads.

Event match quality

Meta's rating of how well it can recognise the people in the events you send. Read it per event, not per account: the account average is dominated by page views, which match easily.

Read more: 9 reports you already pay for and almost nobody opens

Identity resolution

Linking the visits, devices and orders that belong to the same person, for example through a login or the email given at checkout. When it fails, one buyer looks like two customers.

Read more: Why your platforms and your store disagree about who is a new customer

Direct traffic

Not a channel but a label: what analytics records when it could not read a source, so it absorbs stripped redirects, untagged email, in-app browsers and privacy tools.

Read more: Why GA4 disagrees with your ad platforms and your store

Measuring cause

Incrementality

The sales a campaign caused that would not have happened without it. It is measured with experiments, not with attribution.

Holdout test

An experiment that holds a group of people back from the ads and compares what they buy with a group that saw them.

Geo test

An experiment that runs a campaign in some regions and not in comparable others, then compares sales by region.

MMM (marketing mix modeling)

A statistical model that estimates how much each channel contributes to total sales over time, from aggregate data. Google's Meridian and Meta's Robyn are open-source examples.

Leads

Lead

A person who left their contact details, usually through a form, a call or a message. For a business that sells through leads, the result to measure is the closed deal, not the form.

Offline conversion import

Uploading sales that closed outside your website, for example in a CRM, back to the ad platform with the click ID of the original lead, so it can learn which clicks became customers.

AI and your data

MCP (Model Context Protocol)

An open standard that lets an AI assistant such as Claude or ChatGPT connect to outside data and tools. For marketing, it means asking questions of your campaign data from the assistant itself.