Confident and Wrong

The industry has spent 2026 rebuilding itself around outcomes. The IAB is standardising how they are measured. WPP has dissolved its holding company structure to chase them. Omnicom has bought IPG and named data its strategic engine. None of it has yet reached the people who actually spend the money, and they are failing in two opposite directions at once. This article discusses the difference between a planner who cannot see sales and a buyer who can see the wrong ones, and why both end up confident and wrong.

---

So I sit down in a coffee shop, a planner walks me through a campaign he is proud of. Engagement rate up. View rate ahead of benchmark. Click-through better than the last two quarters. Everything on the slide is green. Then I ask what it sold, and there is a pause, and the honest ones say they do not know. The less honest ones tell me about brand lift. Neither is being evasive. They genuinely cannot answer the question, because nothing in their working week was built to answer it.

That gap gets described as a measurement problem. It is not. It is a decision problem, and the distinction matters enormously, because you can solve the first and leave the second exactly where it was. I used to be a media planner and the issue of using proxies insted of real sales, hasn’t changed in 25 years.

Consider what a paid social and creator planner at a large network agency actually has to work with. Organic performance is graded on engagement rate, which is engagements divided by views. Paid is graded on view rate for awareness work and click-through for anything mid-funnel. Reach is not benchmarked at all, because creator audiences are not comparable to one another, so everything is benchmarked within vertical and follower tier. North of ninety percent of what he runs is graded on awareness rather than conversion.

Then add the structure. Organic and paid are measured in entirely separate systems, and deliberately so. He will not use the group’s own reporting layer as a source of truth for paid when his own team is the one buying the paid media, which is a rational position and also a revealing one. So paid numbers come straight from Meta and TikTok, via UTM tracking where the client has bothered to set it up. On accounts where a second agency handles a different discipline, each side contributes slides to one shared deck, and the client is left to reconcile them into a plan.

He uses AI daily, for research and for building decks. He has never once pointed it at his own performance data, because he sits at the planning and review layer rather than the analytics layer. He is separated from the sales signal by his job description, not by his tooling.

None of this is incompetence. It is standard practice at a good agency.

The obvious objection

At this point anyone who has run performance media will object, and they will be right to. A DTC buyer in Ads Manager is not looking at engagement rate. They are looking at purchases, ROAS and cost per acquisition, refreshed continuously, sitting next to revenue in their commerce platform. For a very large share of digital buying, the claim that nobody can see sales is simply false.

In fact the standard criticism of that population for the past decade has been the exact opposite. They are accused of being welded to bottom-funnel sales metrics, over-optimising to last click, and starving everything that builds demand. Too close to the sale, not too far from it.

Both things are true, and that is the interesting part. There are two populations here, failing in opposite directions.

Upper-funnel and creator planners have a distance problem. There is no sales signal in their working week at all. They are graded on proxies, and they know they are graded on proxies.

Performance buyers have a fidelity problem. There is a sales signal, it is front and centre, and it is self-reported and partial.

Consider what that number actually is. Every platform reports the conversions it believes it caused, using its own attribution window and its own methodology. Meta marks Meta’s homework. TikTok marks TikTok’s. View-through attribution means a platform can claim a sale from an impression nobody clicked. Add up what each platform claims across a month and the total will frequently exceed the number of orders sitting in your commerce platform, which is a straightforward arithmetic impossibility and also completely normal. Anyone can run that check on their own accounts this afternoon.

Then there is what the number cannot see at all. Platform-reported conversion is pixel-bounded, and a pixel cannot be placed on Amazon or on TikTok Shop. For a brand doing meaningful volume through marketplaces, a substantial share of the sales their media actually drove is invisible to the system telling them what their media drove.

So the performance buyer is not looking at sales truth. They are looking at a precise, daily, biased, partial estimate, presented with the confidence of a bank statement.

That is a purer version of the problem than the creator planner has, not a milder one. The planner on proxies knows he is on proxies. The buyer on platform-reported ROAS believes he is on ground truth, and acts accordingly.

Three distances, two shapes

It helps to stop calling this one problem. The gap between someone buying media and the sale has three components, and each shows up differently depending on which population you are in.

Distance by metric. For the creator planner, there is no causal link to a sale at all, only engagement and view rate. For the performance buyer, there is a link, but it is self-attributed rather than incremental. Platform-reported return and incremental return are different numbers, and geo experiments routinely show how different.

Distance by structure. For the planner, two agencies each hold a third of the picture and the client reconciles the rest. For the buyer, every platform reports separately and claims the same conversions, while the channels no pixel can reach do not report at all. Neither one holds the whole picture, for different reasons.

Distance by time. Measurement is backward-looking by construction. It describes money that has already gone. On one-off campaigns the learning arrives after the conditions it describes have changed, and in neither population does anyone have a defensible answer to what the next allocation is likely to sell.

Closing one of the three changes very little. Give the creator planner a sales number and he still cannot see the half of the funnel the other agency bought. Give the performance buyer marketplace coverage and he is still looking backwards at a self-marked estimate.

The problem is not absence. It is confidence.

Here is what unites the two populations, and it is the part the industry consistently gets wrong.

Neither of them is short of data. Both have dashboards full of accurate, timely, well-presented numbers. What neither has is a number that causally connects their decision to a sale, across everywhere the brand actually sells.

That is a worse position than having no data at all. Someone who knows he is flying blind behaves cautiously. He hedges, he tests, he asks. Someone looking at a full dashboard of green metrics behaves decisively, because confidence is what releases budget. If those metrics are uncorrelated with incremental revenue, or negatively correlated with it, that decisiveness is pointed at nothing. Worse, it compounds. He optimises harder toward the metric, the metric improves, the dashboard gets greener, and revenue does whatever it was always going to do.

Frequency is the cleanest example on the planning side. Push it up and efficiency metrics improve almost mechanically. Cost per thousand falls, delivery looks tighter, the plan looks better managed. A great deal of the time you are buying the same person the same message for the fourth time and calling it a saving. On the performance side the equivalent is retargeting, which reliably reports beautifully and frequently sells to people who were going to buy anyway.

So the honest framing of where the industry stands is not that media measurement is imprecise. It is that a large number of confident, competent people are making expensive decisions using instruments that cannot detect the thing they are ultimately judged on.

Everything happening at the top of the industry right now is correct, and none of it fixes this on its own.

The IAB’s Project Eidos, announced in February and running across every one of its centres of excellence with more than forty participating companies, is the most serious attempt yet to standardise how outcomes, attribution and incrementality are measured, and to modernise marketing mix modelling on privacy-ready inputs. The IAB’s own State of Data report puts thirty-two billion dollars of media investment and productivity value on the table within two years. That is not hype, and it is overdue.

But standards harmonise vocabulary. They do not put a number in front of a buyer on a Tuesday afternoon. Consolidation is the same story. WPP’s Elevate28 plan collapses the holding company into a single company across four operating units, and Omnicom’s acquisition of IPG created the largest media operation in the industry. Both reduce the number of layers between a decision and a result, which is good, and neither changes what appears on the screen when the plan gets built. Platforms are the same again. WPP Open, with Open Intelligence connecting client data to more than 350 partners, is exactly the right architecture. A sophisticated surface with no independent sales signal running through it is a very expensive way to keep looking at self-reported numbers.

The industry has spent this year building the plumbing for outcome-based media. Far fewer people are thinking about what comes out of the tap.

You cannot make a data-driven agency successful from the boardroom unless you get the last-mile right.

What the other version looks like

Probability does not live in the strategy deck. It lives in the decision, and the decision is made by someone at the moment of the buy. So the work is narrower and less glamorous than most of the current conversation suggests. It is not better modelling. The models are fine and improving, and most serious vendors are converging on similar answers there.

I should declare an interest, because this is what we build at our venture Fospha . Several of the components below are now offered by more than one measurement provider, and I would rather describe the shape of the answer than pretend Fospha is the only one who could reach it.

Map it against the three distances and it gets concrete.

Metric closes with a signal tied to sales and independent of the platforms being graded, expressed at the level the buyer actually works at. Not a channel-level number handed down once a quarter, but a read at ad and creative level, which is the unit anyone can act on. Independent matters more than accurate here. A number produced by the seller is not evidence, however precise it is.

Structure closes when paid, organic and marketplace sit on a single basis in one place. Amazon and TikTok Shop cannot carry a pixel, which is why so many measurement stacks are blind to exactly the destinations where a growing share of sales now happen. One person holding the whole picture is worth more than three people each holding an accurate third of it.

Time closes twice. Once with cadence, because a sales outcome that lands every morning is an input where a quarterly refresh is a post-mortem, and a post-mortem cannot change a decision. And once with forecasting, because a statistical budget planner estimates what an allocation is likely to sell before the money is committed, which lets two plans be compared on expected return rather than expected delivery. You cannot be probabilistic about the past.

And then the part that is easiest to underrate. All of it has to arrive where the buy already happens, inside the existing workflow rather than in one more dashboard, and ideally it should execute back into the ad platforms rather than terminating in a recommendation. A signal that changes a slide has not changed anything.

One caveat, because it is the honest one. Data alone does not change behaviour when the contract rewards something else. An agency planner is measured against whatever the client wrote into the scope, and if that scope specifies delivery and efficiency, a perfect incremental sales signal will not move them, because they are not paid on it. This is exactly why the shift the holding companies are announcing, from selling time and materials to selling outcomes, matters more than any measurement announcement made this year. The signal and the commercial model have to arrive together. Neither one works alone.

What changes when they do is not the reporting cycle. It is what the word good means. Media stops being judged on how efficiently it was delivered, or on what the seller claims it sold, and starts being judged on what it actually returned. That reframing is worth more than any improvement in model accuracy, because it changes behaviour at the only point in the chain where behaviour moves money.

The industry has decided that media should be bought on outcomes. It has not yet given most of the people who buy media any way to see one.

---

Sources

All industry claims in this piece are drawn from public sources.

IAB

– [IAB announces Project Eidos](https://www.iab.com/news/iab-announces-project-eidos/) — initiative scope, focus areas, participating companies, announced 2 February 2026

– [IAB State of Data 2026: The AI-Powered Measurement Transformation](https://www.iab.com/insights/2026-state-of-data-report/) — the $32bn figure

– [2026 IAB Measurement Leadership Summit recap](https://www.iab.com/blog/2026-iab-measurement-leadership-summit-event-recap/) — triangulation framework, agent-ready standards

– [Guidelines for Incremental Measurement in Commerce Media](https://www.iab.com/guidelines/guidelines-for-incremental-measurement-in-commerce-media/) — IAB and IAB Europe

Omnicom and IPG

– [Omnicom completes acquisition of IPG](https://www.campaignlive.com/article/omnicom-completes-acquisition-ipg-creating-worlds-largest-ad-holding-company/1941216) — Campaign

– [Omnicom’s new structure: creativity, media and tech plans](https://adage.com/agencies/aa-omnicom-ipg-post-merger-structure-media-tech-creative/) — Ad Age

WPP

– [Strategy Update and 2025 Preliminary Results](https://www.wpp.com/en/news/2026/02/strategy-update-and-2025-preliminary-results) — Elevate28, single-company transition, four operating units

– [WPP Open](https://www.wpp.com/en/wpp-open) — the agentic marketing platform and Open Intelligence

– [WPP news and press releases](https://www.wpp.com/en/news) — partner expansions, Open Pro, Agent Hub

– [Why WPP’s new CEO is looking to AI](https://diginomica.com/why-wpps-new-ceo-looking-ai-deliver-new-golden-age-modern-marketing) — Cindy Rose on the AI roadmap and platform partnerships

The planner described in this article is a composite drawn from several conversations with practitioners at mid-market and network agencies during 2026.