The conversation happened in a coffee shop in the Northern Quarter in Manchester, where the paid lead's team is based. She works at a UK retail brand with roughly £58m of revenue, primarily direct-to-consumer with a small wholesale presence. The paid function she runs is a team of two and a small agency retainer for creative production. Her name is Lior, and she has been in the role for seven years — long enough to have gone through most of the industry's fashions and to have developed her own view on which ones apply to her account.
The decision
Priya Fenella-Moreau: When did you stop running Microsoft Ads?
Lior: We turned it off in the last week of March 2024. It had been running for perhaps eight years before that. I inherited it when I started in this role and, honestly, ran it on autopilot for the first three years. It was doing something — I could not have told you exactly what, but the reported numbers were positive and the spend was small enough that it did not warrant the attention required to interrogate it.
The trigger, in early 2024, was that our finance team was doing a bottom-up review of every marketing line item — this was a broader exercise, not a Microsoft Ads-specific one — and they came to me with a question. They said, roughly: this Microsoft Ads line has been running at a fairly consistent monthly spend for the last three years, and by our internal model of paid-media contribution margin, it is producing meaningfully less contribution than the platform-reported ROAS would suggest. Can you explain the gap?
I could not, immediately, explain the gap. So I spent three weeks investigating.
What she found
Priya: What did the investigation surface?
Lior: Three things, in the order I found them.
The first was that a substantial share of the Microsoft Ads clicks we were paying for were coming from users who were also, in the same session or in an adjacent session, clicking on our Google Ads. When I ran the analysis on the analytics side, the modal path was: Microsoft Ads click on Bing, then a Google Ads click on Google an hour later, then a purchase attributed to Google Ads by our attribution model. Microsoft Ads was, in these cases, appearing in the customer journey but not being credited for the conversion. It was, however, still costing us the click.
The second thing I found was that the reported conversions Microsoft Ads was booking were, on inspection, largely conversions the platform's modelling was crediting to it that our own analytics did not corroborate. This is a specific mechanism I hadn't previously appreciated: Microsoft's conversion modelling is trained on a smaller data set than Google's, and it is, in our category, meaningfully more generous in what it credits to itself. The reported ROAS was materially higher than the analytics-verified ROAS.
The third thing I found — and this was the one that clinched the decision — was that our incrementality on the Microsoft Ads line was very low. We ran a small holdout in one geographic region for six weeks and compared conversion volume against a matched control region where Microsoft Ads continued to run. The difference was within the noise band. The holdout region did not, on any reasonable read, produce fewer conversions than the control region.
Why she thinks this happens
Priya: What's the mechanism, as you understand it?
Lior: I have a working thesis. The thesis is that Microsoft Ads, in our category, is primarily reaching a user segment that is highly likely to also be reached by Google Ads at some point in the same purchase journey. The absolute reach of Microsoft's search inventory is much smaller than Google's, but the reach it does have is heavily overlapping with Google's reach. For a brand that is already running a robust Google Ads programme, the marginal reach that Microsoft Ads is adding is small, and the incremental conversions it is producing are, correspondingly, small.
The line item, in that specific configuration, is not producing incremental revenue. It is producing incremental credit for revenue that would have converted through Google Ads. The dashboard makes it look productive. The incrementality analysis makes it look neutral. The finance model makes it look like a cost that is not offset by revenue.
I want to be careful here. This is our brand, our category, our funnel. There will be brands for whom Microsoft Ads is genuinely incremental — brands whose Google Ads presence is weak, brands in categories where Bing has a specific audience concentration, brands with distinct enough targeting that the overlap is smaller. My argument is not that Microsoft Ads is universally worthless. My argument is that in most retail brands with mature Google Ads programmes, it is likely producing far less incremental value than the dashboard suggests, and it is worth testing.
What she did with the budget
Priya: What did the reallocated budget go on?
Lior: Perhaps 60% of it went into strengthening our Google Ads programme in ways I had been under-investing in — more creative production, a more sophisticated value-based bidding setup, a better product feed. Perhaps 25% went into a series of small tests in adjacent channels: connected TV, retail media, a small YouTube programme. The remaining 15% went into landing-page work.
The blended contribution margin per paid media pound, across the whole marketing programme, rose by about 8% in the two quarters following the reallocation. Some of that lift is attributable to the Microsoft Ads discontinuation. Some is attributable to the reallocated spend being deployed against higher-margin opportunities. Some is background market variation. I would not claim more than that. What I can claim is that the reallocation was net positive, that we have not missed Microsoft Ads in any measurable way, and that the internal conversation about paid-media honesty has been meaningfully sharper since.
"The reason most brands keep running Microsoft Ads is not that they have measured its incrementality and found it positive. The reason is that it is small enough not to warrant the attention required to measure it. That's a bad reason to spend money."
The advice
Priya: Last question. What would you say to another paid lead reading this, currently running Microsoft Ads at the same low, unexamined level you were?
Lior: I would say: run a six-week geographic holdout on it. Pick a region, turn it off there, keep it running everywhere else. Compare conversion volume between the holdout region and a matched control region. If the difference is within the noise band, you have your answer: the line is producing essentially no incremental value, and you can turn it off with confidence.
The cost of the test is one region's worth of Microsoft Ads spend for six weeks — for a typical mid-market retail brand, perhaps a few thousand pounds. The information value of the answer, in either direction, is much larger than that. If the test tells you the line is incremental, keep spending, and now defend the spend to your CFO with confidence. If the test tells you the line is not incremental, reallocate the spend and improve the overall economics of the paid function.
Almost nobody I have talked to has run this test. Almost everyone who has, in retail, has come back with the same answer we came back with. That should mean something.