Survey: 80% of Marketers Cannot Explain Why Campaigns Succeed or Fail

What the On Device Poll Found About Marketers' Confidence Gap

A poll of 254 advertising professionals at MAD//Fest 2026 has put a number on a long-simmering problem: four in five marketers say they cannot fully explain why their campaigns succeeded or failed. The survey, run by Brand Lift measurement firm On Device, found that despite widespread access to dashboards and performance reports, confidence in the reasons behind an outcome remains strikingly low.

The picture becomes sharper in the follow-up findings. Nearly two-thirds of respondents (63%) said they fall back on educated guesses, such as a sales spike or a click-rate increase, to account for campaign performance. Another 16% admitted they simply move to the next campaign without knowing why the previous one worked, and 75% said they lean on delivery metrics like reach or click-through rate when they struggle to explain performance.

There was a notable contrast by measurement approach. Marketers already running Brand Lift studies were twice as likely to say they were completely confident explaining outcomes (28% versus 14%), while those not using Brand Lift were nearly sixteen times more likely to say they were essentially flying blind after a campaign sign-off (28% versus 2%). On Device CEO Alistair Hill said the findings show the industry is not measuring the right things and that AI will only help if the underlying data is strong enough.

Why Ad Measurement Still Runs on Delivery Metrics and Guesswork

The Confidence Gap Is a Metrics Problem, Not a Data Shortage

The research points to a mismatch between the amount of data marketers hold and the quality of causal insight they extract from it. Delivery metrics such as reach and click-through rate tell advertisers whether a message was seen or clicked, but they do not by themselves explain whether the campaign changed brand perception, demand or sales. The 75% who fall back on these metrics are therefore reporting activity rather than effect.

Brand Lift Users Report a Real Difference—but the Vendor Context Matters

On Device sells Brand Lift measurement, so the findings should be read with that commercial interest in mind. Still, the contrast is meaningful: Brand Lift users were far less likely to say they were flying blind and roughly twice as likely to express complete confidence. That suggests outcome-based measurement gives marketing teams a framework for answering the underlying question of whether a campaign worked, rather than simply describing what it delivered.

AI Raises the Stakes for Measurement Quality

The survey also captures an unresolved tension over AI. A majority, 57%, believe AI will make robust measurement more important, while 24% worry AI tools will prioritise speed over statistical accuracy. If AI accelerates campaign planning and budget decisions on top of weak measurement, flawed conclusions could spread faster. The 30% who say their biggest challenge is having plenty of data but no single source of truth will face that problem most directly.

Where the Pressure Will Land

Marketing teams and agencies that cannot explain cause and effect are likely to face harder questions from finance and executive stakeholders, especially as budgets come under review. The competitive advantage described by On Device is not purely technical; it is the ability to justify past spending and make a credible case for the next campaign.

How Marketing Teams Can Move Beyond Delivery Metrics

For marketing leaders, the poll offers specific warning signs to check in their own reporting processes:

• Audit whether campaign conclusions are evidence-based or assumption-based. The 63% who use educated guesses are a reminder to require the data behind every “it worked because” statement before accepting it.
• Replace the 75% delivery-metric habit as the primary proof of success. For brand campaigns, require at least one outcome metric, such as brand lift, validated incremental sales or market share movement, before sign-off.
• Tackle the 30% single-source-of-truth problem by naming one owner for campaign performance data and agreeing definitions for reach, click-through rate and brand effect across channels.
• When evaluating AI measurement tools, make statistical accuracy an explicit selection criterion. The 24% who worry about speed over accuracy should be addressed by testing whether vendors report confidence levels and data quality, not just faster outputs.
• Before presenting the next budget request, prepare a plain-language explanation of a recent campaign’s cause-and-effect. With 80% of marketers lacking confidence, the teams that can do this will stand out to finance.