Performance marketing glossary: jargon, translated

But really: Asking the audience instead of arguing in a meeting. It tells you which version won, not whether the ads made money.

But really: The "you forgot something" email. Some of those people were coming back anyway, so measure it against a holdout.

But really: Google's scorecard for each auction. Money is one input. Relevance is another.

But really: Ads keep working a little after they stop running. A model that ignores this credits the wrong week.

But really: Meta's autopilot for sales. Fewer levers to pull, and you still need to check it isn't mostly reselling to people who'd buy anyway.

But really: A chatbot in a new jacket with "agent" on the label.

But really: Google answering the question on its own page, sometimes using yours. Fewer clicks can follow.

But really: Never switched off, so never tested off. Worth asking what happens when it is.

But really: The part of Meta's system that decides which ads even get considered. Any "Andromeda strategy" is an interpretation unless it cites Meta's own documentation.

But really: What a typical order is worth before costs. A bigger basket full of discounted, returned items is not a better basket.

But really: Apple opens your email for the reader. Your open rate went up. Nobody read more.

But really: A way of splitting credit. It tells you who was in the room, not who closed the deal.

But really: How long an ad is allowed to claim a sale after the fact. Each tool can set its own clock.

But really: Another site pointing at yours. Ask where the link comes from, not just how many there are.

But really: What would have happened anyway. A lift claim is only as good as this number.

But really: How much control you hand to the algorithm. It will chase whatever target you give it, including a bad one.

But really: MER's cousin, with agency fees and production costs left out of the bill. It looks a little better for exactly that reason.

But really: Visits where nothing happened. Sometimes that is a problem. Sometimes the page answered the question.

But really: Proof that people noticed you. Whether they then bought is a separate question.

But really: People looking for you versus people looking for what you sell. Only one of them needed convincing.

But really: The ROAS at which the ads pay for themselves and nothing more. Below it, you are paying to sell.

But really: CPA's grown-up sibling. It counts customers, not clicks, and includes the costs people like to leave out.

But really: Paying twice for the same customer.

But really: People who put something in the basket and left. Some come back without being chased.

But really: How long your money sits in boxes before it comes back as cash. Fast growth can make the wait longer.

But really: The customers who stopped buying. In ecommerce, "stopped" is a definition you choose, not a cancellation email.

But really: A measure of curiosity. Clicks are not customers.

But really: What the products you sold cost you. Not the ads, not the office, just the stuff.

But really: A way to stop averaging good and bad customers into one comfortable number.

But really: A messenger between your cookie banner and Google's tags. It does not replace the banner and does not make consent optional.

But really: It is what is left from a sale before you pay for the ads. That is the real ceiling on what ads can cost.

But really: The thing you wanted to happen, as counted by whoever is counting. Three tools, three counts, one order.

But really: A proper experiment, run by the company selling you the ads. Useful. Also worth reading twice.

But really: The share of visits or clicks that turned into the thing you wanted. Change the denominator and the number changes too.

But really: A second pipe for your sales data, because the browser can drop the first. Two pipes mean you need deduplication.

But really: Three numbers for how a page feels to real people. Worth fixing for buyers. Not a ranking cheat code.

But really: What one conversion cost you, according to a platform that also decides what counts as a conversion.

But really: What one click costs. The click is not the sale, so a cheap click can still be an expensive customer.

But really: The price of being shown a thousand times. Cheap attention is still attention you have to turn into sales.

But really: People are bored of your ad. The auction notices before your monthly report does.

But really: Making the shop work better for the people already in it. A button colour change without a hypothesis is guessing.

But really: Your own people, uploaded or tracked, so you can show them ads or exclude them. Your data, Meta's matching.

But really: Everything that happened before and after someone bought. Most of it never shows up in a tracking report.

But really: Sorting customers into groups you will actually treat differently. If nothing changes per group, it is a slide, not a segment.

But really: A smarter way to split credit, trained on Google's view of the path. Still splitting credit, not measuring cause.

But really: Google's social-style ads on its own surfaces. Judge them on incremental results, not on last click.

But really: How much you knocked off. Every point comes straight out of margin, before the ads are paid.

But really: A note to Gmail saying "if it doesn't really come from us, bin it". The "none" setting only watches. It doesn't protect.

But really: You sell to the customer yourself. You also pay for everything the retailer used to do.

But really: Google's shorthand for "would you trust this page?". You can't install it. You can show who wrote the page and why they know.

But really: Emails that send themselves when a customer does something. Some recover sales. Some take credit for sales that were coming anyway.

But really: Whether anyone gets to see your email at all. Your email tool can't promise that. Mailbox providers decide.

But really: Stop emailing people who stopped reading. Your inbox placement for everyone else depends on it.

But really: The share of visits where something happened. "Something" can be eleven seconds of reading.

But really: Matching sales to ad clicks using hashed customer details. Hashed data is still customer data, so consent still applies.

But really: Making sure one sale counted by two pipes stays one sale.

But really: A score for how well Meta can recognise who bought. It says nothing about whether the purchase was counted once.

But really: A cookie set by the shop you are actually in. Being first-party does not make it consent-free.

But really: Data you collected yourself. Owning it does not mean you can use it however you like.

But really: How many times the same person saw the ad, on average. Averages hide the people who saw it twenty times.

But really: Often means "please also spend on awareness". Fine, if someone can show what awareness is doing.

But really: A tidy diagram of a messy process. Real customers wander in and out in any order.

But really: Google says it's still SEO. Most of what works is the unglamorous part: being worth citing.

But really: A holdout for when you cannot track individuals. Whole regions go quiet on purpose.

But really: Recording who would have seen your ad, without paying to show them a charity ad instead.

But really: What's left after paying for the product. Ads haven't been paid yet, and shipping, fees and returns may not have been either.

But really: Real sometimes. Also the most popular explanation for a campaign that can't show direct results.

But really: Switch the ads off for some people and see what happens. The most honest question in marketing.

But really: Someone with the authority to say no before the AI acts. Not someone who reads the summary afterwards.

But really: How much of the auction you turned up for. Chasing the last slice is an expensive hobby.

But really: How many times the ad was shown. Not how many people saw it, and not whether anyone looked.

But really: "We couldn't tell" is not "it didn't work". Find out which before you bin the idea or rerun the test.

But really: Did the ads cause the sale, or did they just stand next to it and take the credit?

But really: The correction between what the platform claims and what a test measured. It belongs to one account, one channel and one attribution setting.

But really: Safari's house rules for tracking. A server-side setup doesn't automatically get around them.

But really: ROAS once the ads are no longer allowed to claim sales that were coming anyway.

But really: GA4's old word for conversion. Same idea, different counter, different total.

But really: The words people type, and the words you bet on. They are rarely the same list.

But really: How loosely Google may read your keyword. Looser means more searches, and more you did not mean.

But really: Where the ad sends people. If it doesn't answer the ad's promise, you paid for the click and lost the visit.

But really: The person who scores the goal gets all the credit. The rest of the team gets a mention in nobody's report.

But really: A real phase and a handy excuse. Worth asking which one you are hearing.

But really: How much better the exposed group did than the group that was left alone.

But really: A file Google Search says it doesn't read. Harmless. Not a strategy.

But really: Find more people like the ones who already buy. The copy is only as good as the original list.

But really: Points for buying again. Your best customers join first, so the members' numbers tend to look great.

But really: A forecast. Useful, until someone needs a bigger budget and it starts getting generous.

But really: Not "how did the budget do?" but "what will the next euro do?" Budgets should follow this one.

But really: Maths on your history to estimate what each channel did. Good for the big picture. Less good for this Tuesday's budget.

But really: The dull, honest number. Nobody gets credit, so nobody can fiddle it. It also cannot tell you which channel is doing the work.

But really: Meta's own estimate of what Meta caused. A better question, asked by the same referee.

But really: The smallest effect your test can see. If the real effect is smaller, the test will just shrug.

But really: A standard plug between an AI assistant and your data. A plug gives access. It does not check the sums.

But really: Credit shared more fairly among everyone in the room. It still doesn't ask who would have bought anyway.

But really: CAC with the returning customers taken out, so selling again to existing fans stops looking like growth.

But really: A list of searches you refuse to pay for. Nobody writes it once. It needs regular attention.

But really: Make leaving easy, or the inbox will make it easy for them by calling you spam.

But really: How long you are out of pocket. Finance cares about this more than any ratio on the slide.

But really: Stopping the test when it looks like you have won. Do it often enough and you will "win" tests where nothing happened.

But really: One campaign, every Google surface, and the machine holds the wheel. You still pick the destination.

But really: ROAS that has been told about costs.

But really: The plan to replace third-party cookies that changed course on the way. Check its current status before planning around it.

But really: Your catalogue, in the format the ad platform reads. If the feed is wrong, the ads are wrong.

But really: Text hidden in the data that tries to give the AI orders. Treat everything an AI tool reads as untrusted.

But really: A diagnostic light on the dashboard. Useful for finding problems. Not a result to report.

But really: How many different people saw the ad. An estimate, not a headcount.

But really: You pick the conversations. Reddit finds the people who were recently active in them.

But really: Reddit hears about your sales from your server, not just from the browser. Send the same sale both ways without a shared ID and Reddit counts it twice.

But really: Your catalogue, turned into Reddit ads by Reddit. Only as good as the catalogue and the purchase events behind it.

But really: Reddit's autopilot. Useful when your tracking is good. When it isn't, it learns from the wrong sales.

But really: Reddit's tag on your site. It tells Reddit who bought after an ad. It does not tell Reddit, or you, whether the ad caused it.

But really: How many customers came back. The period you choose decides whether the number looks good.

But really: "You are probably running low." Time it from when your customers actually reorder, not from what the label says.

But really: Ads for people who already know you. Some would have bought anyway. That part is not the ad's work.

But really: Whether customers come back. Acquisition is easier to brag about. Retention is harder to fake.

But really: ROAS after the parcels come back. A platform counts the sale when it happens, not when it is returned.

But really: How much of what you sold came back. Ads take the credit on day one. Returns arrive weeks later.

But really: The agency is paid from your revenue. Which revenue, counted where and from what baseline, decides whether that is fair.

But really: Who bought lately, who buys often, who spends a lot. It describes behaviour. It does not predict who will be happy.

But really: Revenue per euro of ads, as the ad platform tells it. The platform also marks its own homework.

But really: ROI asks whether you made money. ROAS asks how much revenue showed up. Finance hears the first when you say the second.

But really: How much evidence you collected. Small tests produce confident-sounding noise.

But really: Spend more and hope the return holds. Returns tend to fall as spend grows, so ask where it stops.

But really: Why someone searched, not just what they typed. "Running shoes" can mean "teach me" or "take my card".

But really: Making your pages easy for search engines to understand and worth clicking. Slow, and nobody owns the results page.

But really: The page you get after searching. More things competing for the click than a plain list of links.

But really: Moving your measurement in-house. It can recover some lost data. It does not replace consent or fix a bad strategy.

But really: One visit, roughly. One person can have many, and the counting rules are not the same in every tool.

But really: Product ads built from your feed, not from keywords. The feed is your ad copy now.

But really: Google sets the bid for every auction. It optimises what you feed it, so feed it the right numbers.

But really: Bought from you? You may be able to email them about similar things. Signed up but never bought? Different rules, and it depends on the country.

But really: How many people pressed "spam". It is a small number with a big effect on where your next campaign lands.

But really: Boosting a real TikTok post instead of making an ad that looks like one. The creator has to say yes first.

But really: Ads don't respect borders. People commute, and so does your targeting.

But really: A reason to believe the result is not a fluke. Not a reason to believe it matters.

But really: Labels that tell machines what is on the page. They help with rich results. They are not a secret AI switch.

But really: Sometimes real. More often the word that turns up when each channel's numbers don't add up.

But really: A made-up control region built from real ones. It's only as good as how well it tracked you before the test.

But really: A snippet that sends data somewhere when something happens. No trigger, no data.

But really: You name a price per conversion and Google tries to buy at that average. It will not tell you whether the conversion was a new customer.

But really: You tell Google the ROAS you want, and it bids to hit it on average. If the conversion values are wrong, it hits the wrong target very efficiently.

But really: A cookie from someone other than the site you are on. That is what lets one company recognise the same browser on many sites.

But really: If you paid for it, label it. TikTok can quietly stop showing it if you don't.

But really: GMV is gross sales, and the campaign also counts organic and affiliate orders. Good for the headline. Check the margin.

But really: TikTok's tag on your site. Without it TikTok optimises blind. With it, TikTok still marks its own homework.

But really: A shop inside the app. Different checkout, different rules, different numbers from your own site.

But really: Creators sell your products for a cut. The cut comes out of your margin, not TikTok's.

But really: TikTok's autopilot, with a dial. Turn it up when your conversion data is clean. Otherwise it optimises toward noise.

But really: A small informant on your site that helps the ad platform grade its own results.

But really: Ads that look like a customer made them. Often a creator did, for a fee. Fine, as long as nobody pretends otherwise.

But really: Name tags for traffic. Without a naming rule they turn into a mess very quickly.

But really: The ad was on screen some time before someone bought. That's a coincidence until a test says otherwise.

But really: The platform owns the pitch, the ball and the scoreboard.

But really: The "we miss you" emails. Define "gone" for your product first, or you will chase people who were about to buy anyway.