Glossary
The terms we use on calls, in audits and in our case studies, in plain English, with what each one means for your ad account. Most are industry standard. The ones marked "our term" are names we coined. Most link to a real account where they made the difference.
No terms match. Try a shorter word.
Clicks, installs or leads that don't come from real, interested people: bots, click farms, accidental taps and paid-to-install schemes. It concentrates wherever traffic is cheapest, which is why campaigns optimized for link clicks attract so much of it. The first fix isn't a fraud tool. It's optimizing for an event that's hard to fake, like a registration, a funded account or a qualified lead.
In action: Ground News
The order people's eyes move through a feed ad, as we took it from Facebook's eye-tracking work: center, up, down, right. The image or video first, then the primary text above it, then maybe a tap on See more, then the headline under the image, then the button. Treat See more as consent. The first lines have to work alone and earn the tap; everything after the cut is written for someone who asked for more. For 4ever Mints, that's where the price argument lived.
In action: 4ever Mints
Apple's privacy framework, introduced with iOS 14.5 in April 2021, which makes apps ask permission before tracking people across other apps and websites. Most people say no. For advertisers, it cut the data Meta and others used to target and measure ads, weakened lookalike audiences, and broke a lot of landing-page conversion tracking overnight. Most of what works in paid social today, from the Conversions API to instant forms to broad targeting, is an answer to it.
In action: Jenfi
Average revenue per user: total revenue divided by the number of users over a period. On its own, it's a vanity number. Next to CAC and churn, it tells you how much you can afford to pay for a customer.
When people respond more to an ad showing who they want to be than one showing who they already are. It's the flip side of identification. Testing Canva in French-speaking Africa, our instinct was to cast local models; the best performer featured a model who read as French. Findings like this are uncomfortable and need handling with respect. But you don't get to ignore a result because you don't like it.
In action: Canva
Total revenue divided by total ad spend across every channel, instead of each platform's own attributed number. Platforms over-claim, and their claims overlap, so adding up every dashboard's ROAS gives you a number bigger than reality. Some teams call it MER, marketing efficiency ratio. We set targets on blended ROAS, then use platform data to decide where the next dollar goes.
In action: Ceremony Matcha
Brand marketing shapes how people feel about you over time. Performance marketing drives a measurable action now: a purchase, a sign-up, a lead. We're a performance marketer, which doesn't mean brand doesn't matter. On-brand creative often beats off-brand creative, as it did for Jenfi in Vietnam. It means every brand idea earns its budget in a test.
In action: Jenfi
Searches that include your company or product name. They convert extremely well because the person already wants you, which is exactly why paying for them deserves scrutiny. If you rank first organically, paid branded ads may just be buying clicks you'd have gotten free. Sometimes they're worth it, to block competitors bidding on your name or to send people to a page that converts far better than your homepage. Test it; don't assume either way.
In action: PASCO Scientific, SmarterSelect
Running ads with few or no audience restrictions and letting the platform find buyers, guided mostly by your creative. Since lookalike audiences faded, it's become the default on Meta. It only works when the creative is strong enough to pick its own audience. Before we trust it, we want to see a click-through rate around 2%.
In action: PASCO Scientific · Read: Meta & Google Ads in 2026: Creative Is the New Targeting
What you spend to win one paying customer. It's the number every other metric eventually answers to. Measure it where someone actually pays you, not at the click, the install or the sign-up. For a fintech like UGLYCASH, that meant cost per first funding, not cost per download.
In action: UGLYCASH
When paid ads take credit for conversions you'd have gotten anyway, usually from organic search, app-store search or direct traffic. The tell is what happens when you pause: if organic rises to fill the gap, paid was cannibalizing. For UGLYCASH, organic Android installs went up when Google paused, which pointed straight at paid app-store search on the brand name.
In action: UGLYCASH, MOOG Audio
Meta's server-side tracking. Your server sends conversion events straight to Meta, instead of relying only on a browser pixel that ad blockers, cookie banners and iOS privacy settings can break. It's how most advertisers clawed back measurement after iOS 14.5. It helps a lot, and it's still a partial picture. That's why we check Event Match Quality, and why for lead generation we often keep the conversion on-platform with instant forms.
What you pay for one acquisition, where you decide what an acquisition is. That decision matters more than the number. Cost per install flatters a campaign, because plenty of installs never open the app. For Canva, we defined an acquisition as install, open and register, so $0.03 per registration meant $0.03 for a real user.
In action: Canva
When an ad's performance decays because the same people have seen it too many times. Frequency rises, click-through falls and costs climb. For Ceremony Matcha, image ads roughly doubled in cost per purchase once frequency reached 4. The cure is a bench of fresh creative ready before you need it, not a scramble after costs spike.
In action: Ceremony Matcha
People value something more once it feels like theirs. In the classic experiments, people handed a coffee mug asked roughly twice as much to give it up as others would pay to buy it. In ads, you borrow the effect by letting people use the product inside the ad. For Ground News, the winning creative was a real news story with the bias bar on it: the viewer experienced the app before installing it.
In action: Ground News
Meta's 0-to-10 score for how well the customer information sent with each conversion event, like an email, phone number or click ID, can be matched to a real Meta account. A low score means Meta can't connect conversions to the ads that drove them, so it optimizes half-blind and under-reports results. We treat anything under 7 as something to fix before scaling.
In action: CreativiU, Vivee Home
Our shorthand for a simple truth: most people need to see you several times, in several places, before they buy. Marketing folklore puts the number anywhere from five to seven. The exact number doesn't matter. What matters is that no single channel deserves all the credit, which is why judging each channel on last-click results, then cutting the "expensive" one, so often backfires.
Faces pull attention, and eyes looking straight out of the screen pull it hardest. Across thousands of ad variants for different clients, models making direct eye contact with the camera have consistently outperformed, and a female model has tended to draw a stronger response from men and women alike. We treat that as a starting hypothesis on every new account, not a rule.
In action: Jenfi
Growing an account by opening new audiences, markets, offers or creative angles beside what already works, instead of only raising the budget on it, which is vertical scaling. Vertical scaling runs into diminishing returns, because each extra dollar reaches people a little less ready to buy. Our rule is to win narrow first, then scale horizontally. For Food is Good, a diet app, every diet was its own silo: we won celiac and gluten-free, learned that journey and its language, then opened low FODMAP and anti-inflammatory, each with its own audiences and creative.
In action: Food is Good
The sales your ads actually caused, as opposed to the sales they took credit for. The question is simple: if you turned this channel off, what would you lose? You answer it with holdout tests, geo tests or planned pauses, not with platform dashboards. It's the grown-up version of attribution, and the antidote to cutting channels that work or keeping ones that don't.
In action: Ground News, UGLYCASH, MOOG Audio
Meta's native lead forms, which open inside Facebook or Instagram instead of sending people to your website. Their reputation for junk leads is earned, because most are built for speed. Built with purposeful friction, they became Jenfi's best lead source after iOS 14.5.
In action: Jenfi
The idea that people don't buy products; they hire them to do a job. Nobody wants design software. They want a CV that gets an interview. For Canva, every audience and every ad was built around one job, like a CV or a birthday card, and that's how cost per registration got to three cents.
In action: Canva
A rule, set before launch, for when an ad gets switched off. For Canva it was roughly: past 3,000 impressions, with click-through under 1% and cost above target, it dies. Today we use tiered kill gates, cheapest checks first: early engagement, then click-through and CPM, then spend against target, then cost per acquisition at a minimum volume. A kill rule takes opinion out of the decision, so budget stops funding losers.
In action: Canva, Mei L'ange, 4ever Mints
Whatever you offer in exchange for someone's contact details: a guide, a demo, a giveaway, a free audit. It's a bigger lever than most ad tweaks. For PASCO, leads from one giveaway cost 20 to 33% less than leads from another, with the same audience on the same platform.
In action: PASCO Scientific
The stretch after a launch or a big edit when an ad platform is still working out who to show your ads to. Results are unstable and usually more expensive. Meta wants about 50 optimization events per ad set within a week to exit it. That's why the event you optimize for has to happen often enough: pick one that's too rare and the algorithm never finishes learning.
In action: UGLYCASH
Audiences an ad platform builds to resemble a list you give it, like your best customers. Before iOS 14.5 they were the most powerful targeting tool in paid social, and they built a lot of our early results, Canva included. They've been fading ever since. Across the 30 to 40 accounts we audited through 2023 and 2024, their effectiveness dropped sharply from around January 2024. Today, creative does most of the targeting.
In action: Canva, PASCO Scientific · Read: Meta & Google Ads in 2026: Creative Is the New Targeting
The total revenue, or better, gross margin, a customer brings in over their whole relationship with you. It sets the ceiling on what you can pay to acquire them. For subscription and repeat-purchase brands, judging ads on first-order revenue undervalues every channel. Ceremony Matcha's average customer was worth roughly twice their first order.
In action: Ceremony Matcha
Our name, borrowed from gaming, for optimizing each channel in isolation: cut whatever looks expensive today, and pile everything into whatever looks cheapest. It's one of the most common ways founders hurt their own growth. Channels feed each other, so the "cheap" one often depends on the "expensive" one. We've watched it play out at more than one company: paid gets cut, then comes back later at a higher cost.
In action: Canva
A tool like AppsFlyer, Adjust or Singular that attributes app installs and in-app events to the ads that drove them, across every ad network. When your MMP and your ad platforms disagree, figuring out why is usually the first job, because every scaling decision depends on it.
In action: Ground News
Demanding that ad spend pay for itself within 30 days. It makes sense when cash is tight. But in a subscription or repeat-purchase business, most of a customer's value arrives after month one, so the rule undervalues every channel and caps growth. Judge channels on payback period and lifetime value instead, and decide how long a payback you can afford.
In action: Ground News
A marketing-qualified lead fits your criteria on paper. A sales-qualified lead has been vetted by sales as a real opportunity. In theory, you'd optimize ads for SQLs. In practice, that only works if sales updates the CRM consistently, because the algorithm learns from whatever you feed it, mistakes included. For Jenfi, we optimized to MQLs, the cleanest signal marketing controlled.
In action: Jenfi, SmarterSelect
Conversions that happen outside the ad platform's view, like a deal closing in your CRM, sent back to the platform so it can optimize toward them. They're how lead-generation advertisers teach Meta and Google what a good lead looks like, and how anyone keeps measurement honest when cookie consent hides part of the picture. They're only as good as the data behind them.
In action: Jenfi, Ground News
How we choose the event an ad platform optimizes for: go as deep in the funnel as volume allows, and climb a rung once the next event has enough volume. Installs, then verified sign-ups, then funded accounts, for example. For UGLYCASH, optimizing for identity verification produced first fundings 43% cheaper than optimizing for funding directly, which didn't yet have the volume to learn from.
In action: UGLYCASH
How long it takes a customer's revenue, ideally gross margin, to repay what you spent to acquire them. It tells you how fast you can scale without running out of cash. A two-month payback and a twelve-month payback call for completely different budgets.
Google's fully automated campaign type, which runs across Search, Shopping, YouTube, Display, Discover and Gmail from a single campaign. It can find volume you'd never find by hand, and it also takes credit generously, branded searches included. In our experience it overstates sales by about a third, so we check it against a standalone Shopping or Search campaign and against blended numbers. We also use it as an exploration tool, to find search terms worth their own campaign.
In action: Ceremony Matcha, PASCO Scientific
Designing a form so people have to stop and think before they submit it. Instead of a one-tap, pre-filled lead form, you ask five to seven real questions. Fewer people finish, but the ones who do mean it, and because instant forms are first-party, Meta sees every step and learns to find people who complete thoughtful forms. It took Jenfi's cost per lead in Vietnam from $4 to $2 while the qualified rate rose from 30% to 40%.
In action: Jenfi
Advertising in categories platforms handle with extra caution: politics and news, finance, crypto, health and more. Expect stricter policies, special ad categories, more reviews and more user reports. Polarizing creative can get an account restricted even when no rule is broken. Our rules: know the policies cold, plan for restrictions, and never circumvent one. A ban can cost you a channel for years.
In action: Ground News, UGLYCASH
Apple's privacy-preserving attribution system for iOS app installs. It tells an ad network that an install happened, delayed and aggregated, without identifying the person. Since iOS 14.5, much of iOS campaign measurement runs through it, so configuring it well is a core part of any iOS app audit.
Any result a bot or click farm can produce as easily as a person: link clicks, likes, video views, engagement, reach. Optimizing for one tells the ad platform to find whoever performs that action most cheaply, and that is often not a person. We optimize for leads you can call and sales you can see in your own systems instead. When Yizzam tested link-click optimization against purchases, bounce rate went from about 55% to 99.99%.
In action: Yizzam, PASCO Scientific · Read: Why we don't run link-click campaigns
The testing method Kevin Veitia and his mentor Michael Reiss of Advertology built on Yizzam between 2015 and 2020, later used for Canva, Ground News and Jenfi. Its stages: Study the market, Unveil a minimum viable ad, Refine targeting, Experiment with visuals, Fine-tune copy, Revise the landing page, and Expand to find the ceiling on scale. Before iOS 14.5, the audience was the biggest lever, so SUREFIRE ran one control ad across competing targeting models before testing any creative. Since then, and especially since Meta's Andromeda update, the creative does most of the targeting and the audience stage matters far less.
In action: Yizzam, Canva, Ground News, Mei L'ange, 4ever Mints, Altium, Slickplan
The point where the right message reaches the people it was written for, also called ad-to-audience fit. It's what every test we run is looking for: the right product or job, in front of the people who care about it, in words that sound like them. For Yizzam, it was art t-shirts in front of art teachers. For Canva, a CV template in front of someone writing a CV.
What we call it when one polarizing figure in an ad drives response far above anyone else. Testing Ground News ads during the 2020 election, only Trump consistently lifted installs; in the UK, Boris Johnson did the same. It isn't about politics. It's about a face nearly everyone has a strong opinion on. It carries risk, too: polarizing faces draw user reports.
In action: Ground News
Our name for a pattern we've seen for years: plain, even awkward ads beat polished ones in the feed. Polish reads as advertising, and people scroll past advertising without deciding to. A plain image breaks the rhythm of the scroll, and once you have attention you're halfway there. For 4ever Mints, a product shot on a white background, made for the online store, beat a set of magazine-quality ads from a Madison Avenue studio. The point isn't ugliness. It's clarity, and plain usually gets there first.
In action: 4ever Mints, Mei L'ange
Our conclusion after running campaigns on every inhabited continent: good creative travels much further than most teams expect. Local teams often assume a market needs its own visual language. Sometimes it does. More often, when we've tested local motifs against strong, on-brand creative, as for Jenfi in Vietnam, the on-brand creative won. Localize the language and the compliance. Test before you localize the look.
Numbers that look good and don't pay: impressions, link clicks, cheap installs, follower counts. Ground News's early Twitter campaigns reached about $0.10 a click and moved nothing that mattered. Every account we run is measured against an event that proves a real customer.
In action: Ground News
Our name for building ads as templates and changing one element at a time, like the face, the headline, the hook or the offer, so every test tells you exactly what moved the number. Change three things at once and a win teaches you nothing. It's how we found the Trump effect, and how we test models, gaze and format on every account.
In action: Ground News, Mei L'ange
Book a free 30-minute consultation and we'll tell you which ones are costing you money. Want to know what working together costs first? Our pricing is public.
Book a Free Consultation