Blog
Dated takes and operating notes — a reaction to something we noticed, a platform change, how we run things internally. For durable methodology that doesn't expire, see Learn.
2026-10-22
After the first second, only one Meta placement changes as your ad plays
We put a grid and a running clock on a test ad in the Ads Manager previews and measured what sits on top of a 9:16 video in each placement, second by second.
2026-10-21
Top of page rate is not top impression share
Two Google Ads columns sound alike and divide by different things. Know which one you are looking at before you read a rival's row as pressure.
2026-10-16
Exclude 'flowers' and your ad can still show for 'red flower'
Your keywords reach plurals and synonyms whether you want them to or not. Your negative keywords reach neither. How we write negatives around the gap.
2026-10-15
A keyword below average on ad relevance can still score 8
Google publishes three Quality Score components and a list of fixes for each, but not which to fix first. Our table of how the components combine suggests an order.
2026-10-14
Before you copy Google's 50% Display setting, check you want Display
Google Ads is adding channel priorities and budget caps across Search, Shopping, Performance Max and Demand Gen, and Search campaigns with Display turned on lose their current spend setup in February 2027. Google says one setting can keep things as they are. Our view is that most Search campaigns should look at the network report before keeping anything.
2026-10-13
Data Strength Uplift counts Google's gain, not yet yours
Google is launching a metric that reports the conversion gains driven by your data setup, inside Google Ads. In our view it is a useful number and a self-graded one. Here is what the announcement says, what a better data setup changes in the Conversions column, and the check we run against real sales before believing it.
2026-10-12
A tighter Target CPA can buy fewer conversions
The target is the one input Target CPA bidding takes from you, and the one most often set from hope. Here is what Google says the strategy does with it, where we start the number, and how we tell a target that is holding a campaign back from one that is working.
2026-10-11
Meta wants your ad sets combined. Keep the comparison you split them for
In Meta's own opportunity score example, fixing audience fragmentation may be worth 35 points, and Meta's reasons for it are real. But most split ad sets were built to answer a question, and combining them deletes the answer unless you rebuild it somewhere else.
2026-10-10
Meta gave one ad most of the spend. That is not a test result
Meta delivers the ad most likely to be cheapest per optimisation event for each person, so an uneven split between your ads is the system's forecast, not a finding. What to read instead, and when to cut the number of ads.
2026-10-09
Your outranking share can rise without you beating anyone
Google's outranking share includes how often your ad showed when a rival's did not. Read it next to overlap rate and position above rate before you call it progress.
2026-10-09
What is a good top impression share? The bands we work to, and the four columns that say why you miss them
Google defines top impression share and tells you how it is calculated. It does not tell you what a good one looks like, so every account manager carries a number in their head. Here are ours, why brand is held higher than non-brand, and how to read the four loss columns before you change anything.
2026-10-08
How to average impression share across days without getting the wrong answer
Export a month of daily impression share, take the average, and you have a number that looks like a measurement and is not one. Google defines the metric as a ratio, and a ratio has to be rebuilt from its parts before it can be summarised. Here is how we do it, and which days we refuse to count.
2026-10-07
How to optimise Google Ads ads when you have too few conversions to judge
The textbook ad test waits for enough conversions to call a winner. On a small account that can mean waiting until the copy is out of date. On its 'Optimise' ad rotation setting, Google may already be favouring some of your ads over others, auction by auction, and the result shows in how often each one serves. Read that first, use weaker signals in a known order, and say which one you decided on.
2026-10-07
Reddit lead generation ads with Max campaigns: worth testing on a small budget?
Reddit has put its automated Max campaigns to work on lead generation, with in-market audiences and a route from the form into your CRM. Most of the figures in the announcement are Reddit measuring Reddit. Whether it is worth your money is a question only your CRM can answer, and the test is cheap to design that way from the start.
2026-10-06
How Target ROAS treats a low-value conversion goal, and what removing it really costs
Teams hold on to a cheap sign-up goal on a value bidding campaign because they expect taking it out to wreck the numbers. On a value bidder it is a small slice of what the strategy chases. What removing it costs is mostly a period of adjustment, plus the count it carried, and both are questions of timing and checking, not of value.
2026-10-05
Search demand vs buying intent: the two curves behind seasonality
"It's seasonal" usually means one curve. There are two: how many people look for the product, and how many of the people who sign up go on to buy. They do not peak in the same months, and they call for different responses.
2026-10-04
Was it demand or was it us? How to tell seasonality from campaign changes
A good month gets credited to the season by default. Three checks, all of them readable in an afternoon, say whether the season earned it or something you changed did.
2026-10-03
Conversions vs All conversions in Google Ads, and how to rebuild a campaign's goal history
The two columns look like near-duplicates. Read side by side, month by month and action by action, they are a record of what a campaign was counting, for months nobody wrote down.
2026-10-02
How to see conversions by individual action in Google Ads (and why the cost column goes blank)
When a campaign's goal holds several actions, the Conversions figure adds them together. The segment that splits them takes cost away. Two reports, joined, put it back.
2026-10-01
Should Google Ads optimise for purchases or sign-ups?
The usual answer is that the bidder needs more signal, so give it a shallower goal. Sometimes that is right. But the cheaper cost per sale that follows can come from cheaper clicks rather than better ones, and one division tells you which.
2026-09-30
How to judge a bid change without fooling yourself
Most bid changes are judged on the first week, against the wrong target, by someone who does not know whether the clock was reset. Here is what the platforms themselves say about when a change can be read.
2026-07-24
Why we write every ad change down twice
Riibon keeps a change log and a decision log for every material account edit, on purpose, because a single note is never good at both jobs at once.
2026-07-21
What "AI-native agency" actually means in practice, not as a slogan
Every agency now claims to be AI-native. Here's a definition specific enough to fail, plus three questions that tell you whether a given agency actually clears it.
2026-07-17
How to stop an AI's marketing "insight" from becoming permanent bad advice
A single wrong AI answer is a bug. A wrong AI conclusion that quietly becomes a standing assumption is a much harder problem to catch.
2026-07-14
How we stress-test an AI-generated ad recommendation before anyone acts on it
An AI system can produce a plausible-sounding recommendation in seconds. The discipline that matters is what happens between that output and a change to a live account.
2026-07-10
The One Punctuation Habit That Gives Away AI-Written Marketing Copy
The em dash has quietly become the most talked-about tell for AI-generated text, and if your ad copy leans on it out of habit, readers can feel it before they can name it.
2024-05-31
Refer and Grow: How Dropbox Achieved Explosive Growth Through Smart Referrals
Dropbox turned free storage into a two-sided referral incentive, built the invite flow into onboarding itself, and is still cited as one of the clearest examples of product-led referral growth.
2024-05-31
Creating Hype and Engaging Directly: Tesla's Marketing Masterclass
Tesla built a major brand with almost no traditional advertising, relying instead on high-profile product launches, direct social media engagement, and customers who do the advocacy for it.
2024-05-31
Roasting the Competition: How Wendy's Became the Queen of Social Media Engagement
Wendy's built one of the most-followed brand voices on social media with sharp, consistent humour aimed at competitors, trending culture, and its own followers.
2024-05-31
Unlocking Growth: How Headspace Used the Freemium Model to Boost Downloads and Subscriptions
Headspace built a low-friction free tier that showcased real value, then used content marketing and influencer partnerships to convert that free audience into paying subscribers.
2024-05-31
Optimising User Experience: How Booking.com Boosted Conversions with A/B Testing
Booking.com runs constant, iterative A/B tests across headlines, layout, and personalisation, and treats every product change as a hypothesis to prove with data rather than an assumption.
2024-05-31
Tap to Shop: How H&M Revolutionised Social Commerce with Shoppable Posts
H&M turned its Instagram and Facebook feeds into a checkout by tagging products directly in its posts, and paired the feature with influencer content and an omnichannel purchase flow.