Microsoft Audience Network Test and Containment

The first change, and why: In almost every Microsoft account I take over, the fix is switching the Audience Network off inside search campaigns. Left on, it is not a channel choice anybody made. It is search budget spent on MSN, Outlook.com and the Edge new tab at a cost per acquisition flattered by view-through, and in the accounts I have inherited it has been somewhere between fifteen and forty percent of Microsoft spend. Off first, then tested properly as its own line. What a real test looks like: Separate campaigns with their own budget and their own conversion goal, view-through excluded from the optimisation signal even where it stays visible in reporting. A holdout at region level, since Microsoft's geographic granularity supports that and it is the only way to know whether the conversions were incremental. A fixed test length agreed in advance, and a decision at the end of it rather than an extension. I will tell you when a test failed. Creative, which is where this channel is usually lost: Native placements sit inside a news feed and an inbox. An ad that reads like a banner is ignored there regardless of targeting. Copy has to work as a headline first and an image second, and the image has to survive the 1.91:1 and 1:1 crops plus the wide banner without the product falling out of frame. I write the ad text and specify the crops; your designer or a freelancer produces them. If nobody is making new images, this channel is not ready for you. Audiences: UET remarketing lists, customer match seeded from your own data, and in-market segments. Similar audiences were retired across the industry and there is no lookalike hiding in the interface; the replacement is a properly seeded customer match plus in-market intent, which behaves differently and needs different expectations. Placement exclusion lists get maintained monthly, and I will say plainly that brand safety controls here are coarser than a demand-side platform's. If adjacency is a board-level topic for your brand, buy this inventory somewhere with proper controls or do not buy it. For publishers specifically: MSN traffic arrives curious and leaves quickly. A subscription business measuring this at the click, or at first pageview, is measuring nothing. I set the counted event at second session or trial start, and the campaign is judged on that even though it makes the early numbers look worse than the platform's own report. For food and pet brands: Below roughly 35 EUR average order value, native display does not clear acquisition cost in my experience, and I will say so before we start rather than after a quarter of testing. Repeat purchase rate matters more than first order margin here, so I want subscription or reorder data before recommending any budget. Not included: Image production, video, demand-side platform buying, Microsoft Invest and Curate, and any promise about view-through-attributed volume. Who this is not for: Advertisers with no creative budget. Accounts where Microsoft search itself is not yet clean, since audience spend on top of a broken search account only adds a second unreliable number. Brands who need a channel to work within a month; this one produces an honest answer in about eight weeks and the answer is sometimes no. What you get either way: A written read on whether this inventory does anything for you, with the holdout numbers attached. Roughly a third of the tests I run end in a recommendation to stop, and the value of the engagement is that the decision gets made on evidence instead of surviving as a line item nobody questions.