Easy metrics flatter everyone. Brand lift is the leading indicator that counts. Confidence decides who measures what matters, more than budget ever does.

Hill's route in was close to an accident. Doing a business degree, he changed onto a marketing course 25 years ago because he fancied someone studying it, then found the subject fascinating. His dissertation on measuring the effectiveness of digital advertising set everything since: he realised click-through rates would be useful for performance and would have no bearing on brand, and that digital needed a way of being compared against other media. That conclusion, he says, is basically his business.

A lot of the metrics people use are just easy metrics. At Comscore, research showed an inverse correlation between clicking on ads and branding impact: the people who click are odd, and there are not many of them. Using those numbers to judge the business impact of brand advertising is completely mad. A client recently boasted about twice the click-through rate of any other campaign, and his reaction was, why did you even mention that? Perhaps there was a different audience, perhaps an offer, perhaps a ton of fraud.

The industry reaches for easy things rather than harder things that make a difference. That depends on the sophistication of the brand and, more often, the confidence of the individual, because doing the hard thing is rarely the cheap or easy option. More confidence in the role tends to lead to the slightly more complicated choices that get the best result.

Brand lift is simple at heart. You use passive technology to understand who was exposed to an ad, because consumers have no idea which ads they have seen. Then you build an exposed group and an unexposed group and balance them so the only difference is the advertising. Smart brands have correlated brand metrics to the business outcome they want, so a 1% lift in awareness or consideration maps to sales over three to six months. Hill treats those numbers as a leading indicator of a brand's future success.

There is a straightforward answer and one rarely said aloud. Very often the metrics are used to pat each other on the back and say great stuff, and recommendations go unused even when the answer is clear. They ask why and do not get an answer.

The mission is to measure and enhance: a score, then a recommendation, or now data fed via APIs into machine systems. The frustration of 15 years is watching someone delete a couple of points from a slide, keep one, and plan the next campaign completely differently. The change now is that data is consumed by machines rather than humans, with models built so the next campaign is planned around it automatically.

The value is in the combinations that make a meaningful difference. One soft drinks brand found that on a hot day a radio ad in the morning and a display ad later in the day worked together and sold a lot more. On a recent CTV campaign a client found the best time of day, channel, genre and frequency, trained an algorithm to show ads at that combination, and made it dramatically more effective. Everyone talks about cutting wastage, and this is using evidence and machines to do it.

Hill is not a fraud company on the ad side, but he is deep in survey fraud, which most people do not know about. A very large proportion of market research is bots, often people in low income countries spoofing their location to farm surveys, from 10% to 40% of responses. Because his panels pay consumers, keeping fraudulent respondents out is existential, or you end up showing what a Nigerian thinks about an American campaign. So he uses ID checks, facial recognition and web history.

Others have tracked down racks and racks of phones taking surveys, and it is very much organised crime, hard to know who is really in control. When his team put a new defence in place, within a day there was a YouTube video on how to beat it, watched by a thousand people, advertising a Telegram channel with the details.

Brand advertising nudges people slightly further along the journey. So a 30% jump in awareness should be run through with a fine tooth comb, and anyone consistently reporting very high increases in effectiveness should be treated with suspicion. When his own team sees a huge lift, they assume the data is wrong, and nine times out of ten they are right. Very occasionally there is a genuine ten percentage point increase, the one in a thousand campaign.

AI falls into two camps: the business model and the way the work gets made. The most sophisticated clients build algorithms to target ads, letting machines crunch the combinations of channel, time of day, frequency and audience into outcomes based planning, working backwards from the brand result. Hill believes the future of the measurement industry is feeding data into those systems. On the work itself, automation now turns 3,000 media plans a year into measurement plans at the press of a button, where each once took from 20 minutes to a day.

He takes the vibe coding hype with a pinch of salt. A prototype is relatively straightforward, but building it into a workflow and a system, and creating change around it, is hard and always has been, and much harder at scale. Yet the real-time optimisation people imagine is happening: a client is building an optimise button that ingests his data, and it needs a different setup for each brand and each objective, since awareness for a CPG product is nothing like consideration for a car brand.

There is no real doubt now in most of the marketing world that brand advertising makes performance advertising better. So why do people panic and pour everything into performance? Confidence. If the person in the job is unconfident about their survival, they do it, and if the agency is unconfident about its position with the brand, so do they. On the work he admires, Hill points to HubSpot's advertising in the Uber app at Cannes as a brilliant piece aimed squarely at people like him.

His advice to anyone starting out is to use Excel, or any data tool, really well: learn to split data, gain insight and tell stories with it. Split data across enough parameters and you can tell a story about how the world works. The people who do that well are utterly brilliant, and there are not many of them. The book he keeps returning to is Measure What Matters by John Doerr, which has been fundamental to how he runs the business.