Your campaign hit its KPIs. You still lost.

Your campaign hit its KPIs. Reach, engagement, cost per click, all green on the wrap report. And over the same weeks, your biggest competitor grew faster and took share off you. Your KPIs were wrong because they weren’t benchmarked correctly.

I’ve sat in hundreds of campaign kickoff meetings, agency side and client side, and the conversation about campaign KPIs nearly always went the same way. Someone googled “average engagement rate beauty Instagram”. Someone else dug out last year’s wrap report. We picked the better-looking number, added fifteen percent so it felt ambitious, and moved on to the media plan.

I’m sure you’ve been in that meeting. If the launch is three weeks out and there’s no analyst to ask, you use what you’ve got. What should concern you is how little those numbers tell you once the campaign is live.

Where most campaign benchmarks come from

These days the search is as likely to be ChatGPT as Google. Either way you get a figure from a vendor blog, averaged across brands and goals you’ll never see. It says nothing about how you should do against the four or five brands you’re actually fighting for shelf space with.

Last year’s number feels more rigorous because it’s your own. But if your category grew 15% and you grew 10%, you’d hit target and still lose share.

Then there’s the platform’s own report, the agency’s “our other clients usually see…”, a scroll through a rival’s Meta Ad Library, or a reach target back-solved from finance’s revenue number. They all have their uses. Still, the platform is marking its own homework, nobody can check the agency’s norm, the competitor scroll shows what rivals are doing rather than whether it works, and the back-solved target just multiplies borrowed guesses together.

What none of them tell you is where you stand against your competitors, or what any of it means for revenue.

What the bigger brands use

Brands with a proper insights budget have better tools, and it’s worth being fair to them.

Brand trackers at least compare you with a category norm. They’re quarterly surveys of what people say, though, so a dip shows up weeks after the campaign that caused it. Marketing mix modeling ties spend to sales across channels, and done well it’s very good. It’s also expensive, and the answer usually lands months after the decision it describes. I’m a fan of holdout tests for reading a single campaign, but they’re slow to run and tell you nothing about competitors. The one I’d most like challengers to use is share of voice against share of market, from Les Binet and Peter Field’s work for the IPA. The catch is you need decent data on what everyone else spends.

The newest arrivals are neural networks and double machine learning, which try to pull cause and effect out of ordinary campaign data with no control group. The best test of that idea I know came from Meta’s ads researchers working with Brett Gordon and Florian Zettelmeyer at Kellogg. They took 663 Facebook campaigns with real experiments behind them, so the true answer was known. Even with deep learning models and more than 5,000 features per user, double machine learning couldn’t reliably find it. For purchases, the real median lift was 6% and the model put it at 24%. Prediction and causation are different jobs, and I’ve yet to meet a CFO who’ll accept “the neural net said so”.

Three brands that hit their KPIs and still lost

Getting the benchmark wrong isn’t a rounding error. In three of the best-known cases, the KPI said success while the business went the other way.

Adidas measured its marketing largely on last-click attribution, with no econometrics and no brand tracking. On that measure digital ads were driving ecommerce, so around 77% of the budget went to performance and 23% to brand. When the company finally measured properly, brand activity turned out to be driving 65% of sales across wholesale, retail and ecommerce. Simon Peel, their global media director, told the IPA in 2019 that they had over-invested in digital as a result.

eBay’s paid search reports said ads on the word “eBay” were bringing in sales. Its economists switched them off and watched the same shoppers click the free organic link instead. The study, published in Econometrica in 2015, found brand keyword ads had no measurable short-term benefit. The KPI had been counting customers who were already on their way.

Pepsi’s Refresh Project in 2010 skipped the Super Bowl and put $20 million into good causes chosen by public vote. It drew more than 80 million votes, and at its peak more than a third of Americans had heard of it. Judged on engagement, it was a triumph. Over the same period Pepsi lost share and slipped to third behind Coke and Diet Coke, and the project was later wound down. One campaign can’t take all the blame for that, but nobody would call it a win.

In each case the benchmark came from inside the thing being measured, and nobody was looking at the market or the till.

Why this is getting harder

The 2026 CMO Survey (Duke’s Fuqua School, with Deloitte and the AMA) found 70.6% of marketing leaders moving toward short-term impact, pushed there by CEOs, boards and CFOs. The same survey found marketing’s effect on customers now lasts up to six months, longer than in 2022. So teams get judged over a few weeks on work that pays back over half a year. A KPI borrowed from last year and checked when the flight ends will miss most of it.

What I’d want from a campaign KPI

I want it measured against the competitive set: same category, same shoppers, same weeks. Competitors won’t fill in your brand survey, but they turn up in the same scanner data, search trends and ad spend reports you do.

I’d want it built on what people do. Searches, visits, views, mentions, what goes through the till. Byron Sharp and the Ehrenberg-Bass team showed that brands grow by getting more people to buy, and attitudes tend to follow. Behavior can be checked against sales.

It has to see past the channel the money went into. A TikTok push turns up in branded search a week later, or in which brand ChatGPT suggests for a vitamin C serum. So it needs to cover all marketing activity across Paid, Owned, Earned and Search, including LLM visibility.

And it has to connect to revenue while there’s still time to act: agreed before launch, read while the campaign is still running, tied to sales. Three months later, you’re reading a post-mortem.

Setting the target in the first place

This is the bit I care most about, and it’s why we built KPI setting into Maitrics.

Once you can see where you stand against your competitors before anything goes live, the target gets much easier to set. In the Brand Momentum Index, 100 is the average of your competitive set. So the goal stops being “last year plus 15%” and becomes something like getting Visibility from 92 to above 100 this quarter, with an estimate of what that’s worth in sales next to it.

The team can steer by that while the campaign runs. And when the CMO takes it into the boardroom, the CFO can see the money attached, which in my experience is what they wanted all along.

Hitting your KPIs only counts if the brands next to you on the shelf didn’t hit theirs harder.

Sources

1. The CMO Survey, 2026 key findings: cmosurvey.org
2. Duke Fuqua release on the 2026 CMO Survey: fuqua.duke.edu
3. Marketing Week: Adidas on over-investing in digital: marketingweek.com
4. Marketing Architects podcast: Simon Peel on brand driving 65% of Adidas sales: marketingarchitects.com
5. Blake, Nosko and Tadelis, eBay paid search experiments (NBER): nber.org
6. Pepsi Refresh Project: en.wikipedia.org
7. MediaPost: Why Pepsi canned the Refresh Project: mediapost.com
8. Gordon, Moakler and Zettelmeyer, “Close Enough?” (Marketing Science, 2023): pubsonline.informs.org