The dashboard looks healthy. Green numbers. ROAS clearing the target. Cost per acquisition dropping month over month. The algorithm is doing exactly what you asked it to do. It finds people who will buy inside the next seven days and shows them your ad at the moment they are about to convert.
You are watching a fraction of reality.
The rest of it is the part that drives pricing power, organic demand, and the kind of growth that compounds over years. It does not appear in any column of your Ads Manager. Not because it is not happening. Because the system was built to ignore it.
The consensus is comfortable and wrong
The prevailing logic goes like this. Digital advertising is the most measurable medium ever created. Unlike the old days of television and billboards, you can trace every dollar to a conversion. Smart Bidding and Advantage+ find buyers faster and cheaper than any human media buyer could. The data does not lie.
That logic is internally consistent. It is also missing most of the picture.
Here is why. Meta, Google and TikTok do not measure what advertising does. They measure what advertising does inside a narrow window of time. If a customer sees your brand campaign on a Tuesday and buys six weeks later, which is how brand advertising actually works, the platform records that exposure as worthless. The algorithm learns nothing from it and deprioritises similar audiences going forward.
| Platform | Default window | Maximum available | What that actually means |
|---|---|---|---|
| Meta Ads | 7-day click, 1-day engage, 1-day view (engage-through added March 2026) | 28-day click | The 28-day click is reporting only for standard ad sets. Post-conversion optimisation is the exception. It uses the 28-day click and it does feed delivery. |
| Google Ads | 30-day conversion window. The attribution model defaults to data-driven, not last click. | 90-day conversion window | The window sets how long a conversion can still be counted. The model sets how credit gets shared. Those are two different settings and people conflate them constantly. |
| TikTok Ads | 7-day click, 1-day view | 28-day click, 7-day view | The view-through maximum is seven days, not one. Most accounts never change it. |
Now hold those windows against this. Meta’s own research, run with Nielsen, Nepa and GfK across more than 3,500 campaigns, found that long-term effects make up almost 60% of Total ROI. That study covered European markets in the CPG and Tech and Durables categories, so it is not a universal law. It is still Meta’s own research division reporting that most of the value lands after Meta’s optimisation window has closed. Their scientists know. Their algorithm does not care.
This is not a conspiracy. It is an architectural mismatch. Platforms sell advertising on the basis of attributable conversions because attributable conversions are easy to prove. Short-term ROAS is clean, precise and defensible in a board meeting.
The econometric record points somewhere else. Binet and Field analysed 996 case studies in the IPA Databank, covering roughly 700 brands across 83 categories over 30 years. Two findings matter here. Campaigns planned over three years or more deliver around double the profit of campaigns planned inside a year. Emotional campaigns deliver around twice the profit of rational ones. Neither of those is a finding about brand beating activation, and Binet and Field are explicit on that point. Brand and activation work in synergy, each one making the other work harder.
What both findings have in common is time. The profit shows up outside a seven-day window, which means your bidding algorithm never sees it and never learns from it.
Adidas spent years optimising the smaller half
Between 2015 and 2019, Adidas ran the playbook every performance marketer would recognise. Heavy investment in bottom-funnel digital. Four separate attribution models running at once, none of which agreed with each other. Every budget decision routed through a ROAS dashboard. The numbers looked excellent, so the company kept moving budget toward performance and away from brand.
Then paid search went dark across Latin America. Twice. Traffic and revenue coming from SEO did not dip.
That is a narrower finding than it first appears, and it is more damning for it. The paid search spend had not been creating demand. It had been intercepting people who already intended to buy, people primed by years of accumulated brand equity, and booking their purchases as its own. The laser was taking credit for what the chandelier had built.
Adidas commissioned econometric modelling to size the problem. Only 23% of the global marketing budget went to brand building, and that 23% was driving 65% of all sales across wholesale, retail and ecommerce. The other 77%, optimised relentlessly for short-term return, was doing far less than the dashboard claimed.
You have been in this meeting. Someone presents the ROAS figures. Everything looks efficient. Nobody asks what would happen if the ads stopped, because the dashboard does not model that question. Adidas did not ask either, until an outage answered it for them.
Simon Peel, then global media director, said it out loud at the IPA’s Effectiveness Week in October 2019. An obsession with granular return tracking had led the company to prioritise digital efficiency over business effectiveness. The metrics were precise. They were measuring the wrong thing.
Adidas said it was moving toward a 60/40 brand-to-activation split and reintroduced television, out-of-home and cinema, channels the performance dashboard could not attribute but econometric modelling could measure. That was the stated target. I have not found a public source confirming they reached it, so treat it as the direction of travel rather than a finished result.
A word on the brand value numbers, since this is an article about measurement honesty. Different valuation firms use different methods and their figures are not interchangeable. Brand Finance put adidas at €16.6 billion in its Europe 500 report in July 2025, up 23% year on year. Interbrand’s series moves differently, dipping around 6% in 2024 before reaching $17.4 billion in 2025. Splice those together and you can manufacture almost any story you want. None of them is a clean before-and-after for a media split, so I am not going to present one as if it were.
The tracking data is cleaner. Adidas’s YouGov Brand Index composite score in the US sat flat for five years, from 2019 through 2023, then rose seven points to 37.1.
Airbnb turned it into doctrine
If Adidas stumbled into the discovery, Airbnb built a policy out of it.
Before the pandemic, Airbnb was deep in the growth playbook. Heavy search engine marketing and pay-per-click, chasing measurable short-term return. When COVID hit in 2020 the company cut marketing hard, and on the Q4 2020 earnings call it reported keeping roughly 95% of its traffic. Worth being precise here, because Airbnb’s own framing tends to compress it. Spend did not go to zero. It went to $482 million. The brand was already carrying the load, and a large share of the performance budget had been buying credit rather than customers.
Brian Chesky gave the comparison its now-famous form on Lenny’s Podcast in November 2023. Performance marketing is a laser. It is precise, targeted, narrow, and it does not compound. Brand marketing is a chandelier. It lights the whole room and the advantage accumulates.
The “Made Possible by Hosts” campaign launched on 18 February 2021 in five countries, later expanding to seven. It was brand building in close to its purest form, built on broad reach and emotional storytelling with no direct-response offer, though host recruitment was an explicit objective of the campaign rather than an afterthought.
Now the part that usually gets reported wrong.
Airbnb did not spend less money on marketing. Sales and marketing spend in FY2023 was $1.763 billion, above the $1.621 billion it spent in 2019, and up around 16% on 2022. What halved was marketing as a share of revenue, from 33.7% in 2019 to 17.8% in 2023. That is the figure Chesky and CFO Dave Stephenson actually talk about, and it is the more interesting one. The business roughly doubled the revenue it got from every marketing dollar.
FY2023 net income was $4.8 billion, though that figure was inflated by a one-time deferred tax release, so the underlying picture is less dramatic than the headline. Revenue reached $3.7 billion in Q3 2024, up 10% year on year.
The structural shift shows up in the traffic mix. Across 2020 to 2022, roughly 90% of Airbnb’s traffic arrived direct or unpaid. On the app specifically, 58% of nights booked came through it by Q3 2024. Tracksuit data across 2024 put Airbnb at 86% brand awareness in the US accommodation category, level with Hilton, a company founded in 1919 with an 89-year head start. Worth noting the match is on awareness alone. On preference, Hilton still leads at 22% against Airbnb’s 17%.
Airbnb did not get here by finding a better seven-day algorithm. They got here by refusing to let a seven-day window decide the strategy.
The discounting treadmill eats margin too
ASOS is the cautionary version, and it is worth telling accurately because the popular version welds two unrelated events together.
On 17 December 2018, ASOS issued a profit warning and the share price fell around 40% in a day. CEO Nick Beighton attributed it to unprecedented levels of discounting across the market. That is the relevant lesson for anyone running a permanent promotional calendar to hit this week’s return target. Discounting to force conversions works until the market discounts back, and then it is just margin erosion with extra steps.
The 68% fall in pretax profit that often gets attached to that quote is a different event. It came in the FY2019 results on 16 October 2019, and ASOS blamed warehouse and international expansion problems in Atlanta and Berlin rather than discounting. The share price rose 28% the day those results landed, which tells you how differently the market read the two situations.
The wider industry drew its own conclusion. In McKinsey and Business of Fashion’s State of Fashion 2024 report, published in November 2023, 71% of fashion executives said they intended to increase brand marketing spend in the year ahead.
"So I should turn off Smart Bidding?" No - you should stop asking it the wrong question.
Here is the objection you are already forming. This sounds like an argument for switching off automated optimisation and going back to manual campaign management. It is not.
The algorithm is extraordinary at what it does. The problem is not the machine. It is the objective you have handed it. Run your whole budget through conversion-optimised campaigns and you are asking the algorithm to find the slice of your market that is ready to buy today. It will find them with unnerving precision. What it cannot do, structurally, is build the mental availability that turns everyone else into a future buyer.
That idea comes from Byron Sharp and the Ehrenberg-Bass Institute. Mental availability is the probability that a buyer thinks of your brand in a buying situation, and Sharp’s argument is that you build it across all category buyers, not just the ones currently in market. You may have seen the 95:5 rule quoted alongside this. That figure comes from John Dawes at Ehrenberg-Bass, it was derived from B2B interpurchase cycles, and Dawes himself calls it a heuristic rather than a measurement. The underlying principle travels. The specific number does not.
The fix is architectural, not tactical. You split the budget before the algorithm ever touches it.
Binet and Field’s 60/40 framework, meaning 60% brand building and 40% activation, splits the advertising and media budget rather than total marketing spend. Binet is explicit that it is an average across the databank and not an iron rule.
The category-specific numbers are where most articles go wrong, mine included until I checked them. The real figures come from Effectiveness in Context (IPA, 2018) and the LinkedIn B2B Institute’s 2019 work, and several of them run opposite to what performance marketers assume.
What the research actually recommends, by context
Brand building versus activation, as a share of the advertising and media budget.
Brand building
Activation
| Context | Brand : activation | Split |
|---|---|---|
| Financial services | 80 : 20 | |
| Large brands | 76 : 24 | |
| Online and subscription | 74 : 26 | |
| High innovation | 72 : 28 | |
| Premium | 64 : 36 | |
| New brands | 63 : 37 | |
| Other sectors | 51 : 49 | |
| B2BThe only commercial context that tilts toward activation. | 46 : 54 | |
| Not-for-profit | 44 : 56 |
The dashed line marks the 60/40 baseline. Six of the nine contexts sit above it.
Recommended activation spend never exceeds 56% in any context studied.
Source: Binet & Field, Effectiveness in Context, IPA 2018, and the LinkedIn B2B Institute, 2019. The 60/40 framework splits the advertising and media budget, not total marketing spend, and Binet is explicit that it is an average across the databank rather than a rule.
Read that table slowly if you sell online. Businesses with high online search and online sales sit at 74:26, which is more brand than the 60/40 baseline, not less. Subscription and membership businesses sit in the same place. B2B is the only commercial context in the set that tilts toward activation, at 46:54, and it is the exact opposite of the “long sales cycle so go heavier on brand” argument you hear at conferences. Only not-for-profit sits further that way, at 44:56. Across every context studied, recommended activation spend never went above 56%.
So if you are running an ecommerce brand at 20:80 because activation is where the tracked revenue shows up, the research is not on your side. It is pointing the other way.
The structural requirements then differ by platform. On Meta, brand building means Reach and Video Views objectives with broad targeting and high-emotion creative. Meta’s own creative signal is quality ranking, scored as above average, average or below average against ads competing for the same audience. Branding needs to land inside the first three seconds. Facebook’s research with Nielsen in March 2015 found that exposures under three seconds still produced a 47% lift in ad recall, a 32% lift in awareness and a 44% lift in purchase intent, with measurable value transferring in under a second.
On TikTok, it means storytelling creator content rather than tactical creator content. A Dentsu marketing mix modelling study across 15 Nordic brands, published in February 2025, found storytelling UGC returned 70% higher ROI than tactical UGC. Note the comparison is between two kinds of creator content, not between creator content and everything else.
On YouTube, it means narrative content built on the ABCDs framework, which stands for Attention, Branding, Connection and Direction. Google and Kantar’s April 2021 analysis put it at a 30% lift in short-term sales likelihood and a 17% lift in long-term brand contribution.
The critical move happens before any of that. Protect the brand budget from the ROAS conversation. A brand campaign measured on a seven-day window will lose every budget meeting it ever enters. That is not evidence the brand campaign failed. It is evidence that a seven-day ruler cannot measure a six-month distance.
The dashboard hasn't changed. What you see in it should.
The green numbers are still there. The ROAS still clears the target. The algorithm is still doing precisely what you asked it to do, finding people who will convert inside seven days and serving them an ad at the right moment.
You now know what Adidas found when the ads accidentally stopped, what Airbnb found when they deliberately cut them back, and what the platforms’ own research keeps confirming. Those green numbers are a partial report. The rest is happening past the edge of the attribution window, in the slow accumulation of memory and trust and preference that no pixel tracks and no seven-day optimisation loop can value.
Your dashboard is not broken. It is measuring the wrong time horizon. The businesses that work that out first do not just grow faster. They become the ones everyone else’s algorithm is trying to intercept.
Your marketing, looked at properly
Thirty minutes on your current setup — what’s working, what’s quietly leaking budget, and what I’d fix first. You’ll leave with a clearer picture whether we work together or not.
Got something specific bugging you? Flag it when you book and I’ll have it looked at before we talk.
