Performance marketing glossary.
Every industry uses vocabulary to sound precise and to exclude newcomers. This is the version I would give someone joining my team: 54 terms, defined plainly, each with the practical reason it matters rather than a textbook restatement.
A
A/B test
A controlled comparison in which traffic is split between two versions of an ad, page or flow, so the difference in outcome can be attributed to the change rather than to chance.
Why it matters: Most A/B tests I am shown were stopped the moment they looked good. If a result appears within hours, it is usually noise wearing the costume of a finding.
ACOSAdvertising cost of sale
Advertising spend expressed as a percentage of the revenue it generated. It is the exact reciprocal of ROAS — a 4× ROAS is a 25% ACOS.
Why it matters: Because ACOS is 1 ÷ ROAS, it is also the contribution margin you must beat to make money. That equivalence is the fastest sanity check on any target you have been handed.
Ad rank
The score an auction uses to decide whether your ad shows and in what position. It combines your bid with expected click-through rate, ad relevance and landing page experience.
Why it matters: It is why the highest bidder does not automatically win. Improving the ad is often cheaper than raising the bid, and it lowers your click price at the same time.
AOVAverage order value
The average revenue of a single order, best taken net of VAT and after discounts from your commerce platform rather than your ad account.
Why it matters: Almost every margin calculation starts here, so an AOV inflated by ignoring discount codes quietly corrupts every target derived from it.
Attribution
The set of rules deciding which touchpoint gets credit for a conversion. Last-click gives everything to the final interaction; data-driven models distribute credit across the path.
Why it matters: Attribution is a modelling choice, not a measurement. When several platforms each claim the same sale, the sum of their reported conversions exceeds the conversions that actually happened.
B
Bid strategy
The rule set governing what you pay in each auction — manual bids, or automated approaches such as target CPA, target ROAS or maximise conversions.
Why it matters: Automated bidding needs clean conversion data and enough volume to learn from. Handing it unreliable tracking produces confident optimisation towards the wrong thing.
Bounce rate
The share of visits that ended without a further interaction. Its definition has shifted across analytics platforms, and GA4 reports engagement rate instead, which is its inverse.
Why it matters: A high bounce rate is not automatically bad — a visitor who found an answer and left may be perfectly satisfied. Judge it against intent, not as a universal fault.
Brand safety
Controls preventing ads appearing beside content that would damage the advertiser, through exclusion lists, category blocks and inventory filters.
Why it matters: The cost of brand safety is reach, and over-aggressive blocking can quietly remove much of the inventory you were trying to buy.
Broad match
A keyword setting allowing the platform to show ads for queries it judges related to your term, including synonyms and adjacent intent.
Why it matters: Broad match plus a strong negative keyword list plus reliable conversion tracking works well. Broad match without those is the fastest way to donate budget.
C
CACCustomer acquisition cost
The total sales and marketing cost of acquiring one genuinely new customer, including salaries, tools and agency fees — not media alone.
Why it matters: CAC is the number your business cares about; CPA is the one you optimise daily. Confusing them makes advertising look considerably cheaper than it is.
Consent Mode
A framework in which analytics and advertising tags adjust their behaviour according to the consent a visitor has given, modelling some conversions where consent is withheld.
Why it matters: It changes what your reports contain. A conversion drop that coincides with a consent banner launch is usually a measurement change, not a performance change.
Contribution margin
What remains from an order after every cost that exists because that order happened — goods, fulfilment, packaging, payment fees, returns.
Why it matters: This single figure sets your entire advertising ceiling. Break-even ROAS is 1 ÷ contribution margin, and maximum CPA is contribution margin in currency.
Conversion
The action you have decided to count as success — a purchase, lead, booking or sign-up. Its value is entirely determined by your own definition.
Why it matters: I regularly find accounts counting newsletter sign-ups and purchases in one number. Every metric downstream of a sloppy conversion definition is unusable.
Conversion rateCVR
Conversions divided by visitors, sessions or clicks, expressed as a percentage. The denominator you choose materially changes the answer.
Why it matters: It is the cheapest lever available, because improving it costs no extra media. Doubling conversion rate has the same effect on CPA as halving your click price.
Cookie
A small file storing browser state, historically the backbone of cross-site tracking. Third-party cookies are now restricted or removed across major browsers.
Why it matters: Measurement built solely on third-party cookies has been degrading for years. First-party and server-side collection is no longer a future project.
CPACost per acquisition
Advertising cost divided by conversions produced. Also expressed as CPL for leads or CPI for app installs.
Why it matters: It is the most directly actionable number in an account because you can hand it to a bidding strategy. Its ceiling is your contribution margin.
CPCCost per click
Advertising cost divided by clicks. Equivalent to CPM divided by ten times the click-through rate.
Why it matters: A high CPC is usually a symptom rather than a cause — an ad nobody wants to click, an audience too narrow to fill, or campaigns bidding against each other.
CPMCost per mille
The cost of one thousand impressions. The M is the Roman numeral for a thousand, not million.
Why it matters: CPM is a price, not a result. Cheap inventory is usually cheap because fewer advertisers want it, and optimising hard towards a low CPM buys attention that never converts.
Creative fatigue
The decline in response that occurs as the same audience sees the same creative repeatedly, showing first as falling click-through rate and rising frequency.
Why it matters: It is one of the most common causes of a campaign that 'stopped working' without anything having been changed. Check frequency before you touch the bids.
CROConversion rate optimisation
The practice of increasing the share of visitors who convert, through message match, page speed, friction reduction and testing.
Why it matters: In my experience the biggest gains come from removing steps and matching the promise made in the ad — not from redesigns.
CTRClick-through rate
Clicks divided by impressions, as a percentage. It measures the advertisement, not the business behind it.
Why it matters: High CTR with low conversion rate almost always means the ad promised something the landing page does not deliver. That is the ad's fault, not the page's.
D
Dayparting
Scheduling ads to run at particular hours or days, either by adjusting bids or pausing delivery entirely.
Why it matters: Worth testing where behaviour genuinely varies by hour, but automated bidding already prices time-of-day. Manual dayparting on top of it often duplicates the work badly.
Deduplication
Removing double counting when the same conversion is reported by more than one source, such as a browser tag and a server event firing for one purchase.
Why it matters: Duplicate conversions halve your apparent CPA and flatter everything downstream. A CPA that improved suddenly and inexplicably is usually this.
DSPDemand-side platform
Software for buying programmatic inventory across many exchanges from a single interface, with its own targeting, frequency and reporting controls.
Why it matters: The interface is the easy part. Inventory quality and frequency discipline decide whether programmatic works or quietly wastes budget.
Dynamic creative
Automatic assembly of ads from component headlines, images and calls to action, with the platform testing combinations.
Why it matters: It scales creative testing well and obscures creative learning badly. Keep a record of which components actually won.
F
First-party data
Information collected directly from your own customers and visitors, with consent — purchases, sign-ups, behaviour on your own properties.
Why it matters: As third-party signals degrade, this is the durable asset. It is also the part of measurement that competitors cannot replicate.
Frequency
The average number of times each reached person saw your ad. Impressions divided by reach.
Why it matters: Frequency is where budget disappears silently. The same person seeing an ad fifteen times is not fifteen opportunities; it is one person becoming irritated.
Funnel
The staged model of a customer journey from awareness through consideration to conversion, used to structure targeting, creative and measurement.
Why it matters: Useful as a planning device, misleading as a description of behaviour. Real journeys loop, pause and restart across devices.
G
GA4Google Analytics 4
Google's event-based analytics platform, in which every interaction is an event rather than a pageview within a session.
Why it matters: The event model is more flexible and less forgiving. Poor event naming early becomes very expensive to correct later.
H
Holdout
A group deliberately excluded from advertising so its behaviour can be compared with the exposed group, isolating advertising's true effect.
Why it matters: It is the closest thing to proof available in media. Platform-reported conversions describe correlation; a holdout describes cause.
I
Impression
One instance of an ad being served. Not a person, not a view, and not necessarily something anyone saw.
Why it matters: Impressions count deliveries; reach counts people. Treating the two as interchangeable overstates how many people you actually spoke to.
Incrementality
The additional outcome caused by advertising, over and above what would have happened anyway. Measured with holdouts or geo experiments.
Why it matters: Some share of every campaign's reported conversions would have occurred without it. Incrementality testing is how you find out how large that share is.
L
Landing page
The page a visitor arrives on after clicking an ad, and the point at which the promise made by that ad is either kept or broken.
Why it matters: Sending all traffic to a homepage is the most common and most expensive shortcut in paid media.
Lookalike audience
An audience built by a platform to resemble a source list of your existing customers or converters.
Why it matters: The output is only as good as the seed. A lookalike built from a list of low-value or refunded customers faithfully finds more of them.
LTVLifetime value
The total contribution a customer is expected to produce across their relationship with you, rather than on a single order.
Why it matters: LTV justifies paying more to acquire a customer, but only when the repeat behaviour is measured rather than assumed. Many businesses assume cohorts they have never checked.
M
MERMarketing efficiency ratio
Total revenue divided by total advertising spend across all channels, rather than per campaign.
Why it matters: It sidesteps attribution arguments entirely, which is its appeal. It also cannot tell you which channel to change, which is its limit.
MMMMarketing mix modelling
A statistical approach estimating each channel's contribution using aggregate historical data, including offline media, without user-level tracking.
Why it matters: Privacy-durable and increasingly popular, but it needs long, clean history and meaningful variation in spend to produce anything trustworthy.
N
Negative keyword
A term excluded from matching, preventing ads showing for queries you do not want.
Why it matters: The single highest-return maintenance task in search. Most wasted search budget I find is sitting in a search terms report nobody has read.
P
Pacing
Managing the rate at which budget is spent across a flight, so it neither exhausts early nor leaves money unspent.
Why it matters: Underspend is a failure that rarely gets treated as one, because nothing visibly breaks. The plan simply did not happen.
Performance MaxPMax
A Google campaign type serving across all inventory — search, display, YouTube, Discover, Gmail, Maps — with automated targeting and placement.
Why it matters: Reach and automation in exchange for control and visibility. Strong conversion data and clear exclusions matter more here than in any other campaign type.
Placement
The specific location an ad appeared — a website, app, feed position or video slot.
Why it matters: Placement reports are where programmatic waste becomes visible. Reviewing them is unglamorous and consistently worthwhile.
Programmatic
Automated buying of advertising inventory through real-time auctions across exchanges, rather than direct negotiation with publishers.
Why it matters: Efficient at reaching scale, and equally efficient at reaching nobody worth reaching if inventory quality is never inspected.
Q
Quality Score
Google's diagnostic estimate of ad quality, combining expected click-through rate, ad relevance and landing page experience.
Why it matters: A diagnostic, not a target. It points at which of the three components is weak; chasing the number itself achieves nothing.
R
Reach
The number of distinct people who saw an ad at least once, as distinct from impressions.
Why it matters: Reach and frequency together explain a campaign's delivery. Either one alone can hide a serious problem in the other.
RemarketingRetargeting
Advertising to people who have already interacted with you, using audience lists built from site visits, app use or customer data.
Why it matters: It reliably reports excellent ROAS because it advertises to people who were already going to buy. That is exactly why it needs incrementality testing.
ROASReturn on ad spend
Tracked revenue divided by advertising cost. Revenue, not profit.
Why it matters: A ROAS figure means nothing without the margin behind it. Break-even sits at 1 ÷ contribution margin, and that is the only benchmark specific to your business.
S
Server-side tracking
Sending measurement events from your own server rather than the visitor's browser, improving reliability where browser restrictions apply.
Why it matters: It recovers signal, but it does not remove the need for consent and it introduces a new place for double counting to hide.
Session
A group of interactions by one user within a time window. Definitions differ between analytics platforms.
Why it matters: Most disagreements about conversion rate turn out to be disagreements about whether the denominator was sessions or users.
Smart bidding
Google's family of automated bidding strategies using signals at auction time to price each individual auction.
Why it matters: It optimises towards whatever you told it to value. Feed it a badly defined conversion and it will pursue that error at scale, efficiently.
T
Target CPAtCPA
A bidding strategy in which the platform pursues a specified average cost per conversion.
Why it matters: Set the target from your contribution margin, not from what the account happened to achieve last month. Inherited targets are guesses wearing a suit.
U
UTM parameter
Tags appended to a URL — source, medium, campaign, content, term — so analytics can identify where a visit came from.
Why it matters: Inconsistent capitalisation and naming break reporting silently. A written taxonomy is worth more than any dashboard built on top of it.
V
View-through conversionVTC
A conversion credited to an ad that was displayed but not clicked, within a lookback window.
Why it matters: The most generous form of attribution in common use. Worth reporting separately, never worth blending into click-based results without saying so.
Viewability
Whether an impression met a standard for being seen — typically a defined portion of the ad in view for a minimum duration.
Why it matters: A served impression and a seen impression are different purchases. vCPM prices look higher because you are buying a smaller, better-defined thing.
W
Walled garden
A platform that keeps measurement inside its own ecosystem, reporting on its own performance without exposing user-level data externally.
Why it matters: Each garden marks its own homework. Reconciling their combined claims against your commerce platform is the only defence.
No term matches that search. If something is missing that you think belongs here, tell me and I will add it.