The 'You Need Clean Data Before You Use AI' Argument Just Got Demolished. Most Brands Are Going to Hate the Implication.
AdExchanger argues that the data-cleaning prerequisite for AI is overstated. Agentic AI handles messy inputs the way a good analyst does. The implication: stop waiting for perfect data and start measuring more often.
Marketers have been waiting for the data to be ready. The data was never going to be ready. The brands that stop waiting will compound learnings against the ones who keep waiting.
The data quality conversation in marketing has been stuck for years. Brands assume they need clean, complete data before they can deploy AI usefully. Garbage in, garbage out. The argument is intuitive and the implication is paralysing because clean, complete marketing data is rare. AdExchanger's data-driven thinking column this week pushes back hard. Agentic AI is good at exactly the kind of messy input that has stopped marketers from measuring more often.
The parallel offered is the human analyst. A good analyst confronts inconsistent formatting, missing columns, dashboards from different platforms, signal loss, aging data and decayed audience segments every working day. The analyst does not refuse to work. They interpret, infer, triangulate and produce a useful answer with appropriate caveats. Agentic AI does the same thing at scale and at lower marginal cost.
The deeper point is about measurement velocity. MMM, incrementality testing and attribution are expensive, slow and manual when done by humans on imperfect data. Most brands cannot afford to do them more than once a year. Agentic AI changes the unit economics. More frequent measurement on imperfect data is more useful than annual measurement on perfect data because each measurement is a course-correction opportunity.
Why it matters
Australian marketing teams routinely defer AI projects because the analytics environment is not what the consultants said it should be. GA4 has gaps. The CRM is out of date. The CDP is half-built. The platform metrics do not reconcile to each other. The argument that AI needs clean inputs has been the polite way to defer the project for another quarter.
The imperfect-data thesis reframes the question. The risk is not that imperfect data plus AI produces a wrong answer. The risk is that no measurement at all means no course correction. Imperfect-data measurement at monthly cadence beats perfect-data measurement at annual cadence for almost every marketing decision a business has to make in the next 12 months.
For brands that have been holding back on AI in measurement, the practical move is to start with what exists, document the limitations, and run more often. That is closer to the actual practice of every working analyst than the theoretical clean-data approach the consulting decks have been selling.
Frequent measurement on imperfect data reduces risk through continuous course-correction. Annual measurement on perfect data does not
What to do about it
Stop deferring AI measurement projects until the data is clean. Begin with the data you have and document the limitations alongside the outputs. The limitations become the next quarter's data project.
Run marketing mix or incrementality measurement at monthly cadence rather than annual. The frequency of insight matters more than the precision of any single measurement.
Use agentic AI to handle the cleanup. Letting Claude, ChatGPT or a purpose-built tool reconcile inconsistent metrics across platforms is faster than hiring an analyst to do the same job.
Report measurement outputs with confidence bands rather than point estimates. A range of likely outcomes is more useful for decision-making than a single false-precision number.
Reinvest the time saved by AI-assisted measurement into action rather than into more measurement. The point of measuring is to change something. If your measurement programme grows without your action programme growing, the AI is not paying its way.
Clean data is a long-term project. AI measurement is a now project. The brands that get those two timelines in the right order will outpace the brands that keep them confused.