Defender missed 221 high-severity emails per 1,000 users in Microsoft's own benchmark

A 55.4% lead means the next-closest gateway missed about 496 high-severity threats per 1,000 users, since 221 divided by 0.446 comes to 495.5 ^[17]. First to second place is a spread of roughly 275 missed high-severity messages per 1,000 users across the three months ^[18]. Microsoft identifies that competitor only as "the next-closest SEG vendor" ^[16].

Microsoft counts misses instead of catches, on the argument that catch totals reflect differences in threat volume and exposure across vendor environments ^[4]. The absolute figure is the one a mail administrator has to staff for. At 221 per 1,000, a 10,000-seat tenant took 2,210 missed high-severity messages ^[19]. May through July is 92 days, so that is about 24 a day from the product that scored best in the comparison ^[20]. Defender caught 92% of post-delivery malicious messages on average during the period, so 8% of that category stayed put ^[10][21]. The two percentages use different denominators and do not multiply.

On layering, the report says: "Layered security adds the most value in promotional and bulk filtering and works; gains for spam and malicious email remain comparatively modest" ^[12]. The malicious catch credited to third-party tools more than doubled quarter on quarter and remains under a third of a percent ^[23]. Microsoft gives those figures as bare percentages and does not say what population they are drawn from ^[22].

The vendor that sells the pre-delivery layer is the one reporting that the added layer contributes little against malicious mail. Microsoft has already built product on it: the new Promotions folder in Outlook comes from the observation, across multiple benchmarking periods, that third-party tools delivered their greatest incremental benefit in promotional and bulk email ^[13]. A buyer can take the 0.30% into a renewal conversation, and the SEG vendor across the table can point out that Microsoft ran the test and sells the alternative ^[24].

Microsoft research reports a roughly two-thirds reduction in false negatives and a nearly one-fifth reduction in false positives for Defender customers over four consecutive weeks, from a redesigned machine learning and AI model stack; that measurement sits outside the quarterly benchmark ^[14]. The company also shipped prompt injection protection that detects and isolates malicious AI instructions in email before delivery, aimed at Copilot, agents and other AI systems that read and act on inbox content ^[15].