Outlook for Windows is still the test that matters

Every six months, someone on the team asks if we can stop testing in Outlook for Windows. It would save us hours. Customers would never notice.

Every six months, we go through the data. About 9% of our customers' total opens come from Outlook for Windows. For B2B customers, it's 24%. For some healthcare and government customers, it's over 40%.

So we keep testing. This post is how we do it, and why it's worth the trouble.

The shape of the problem

Outlook for Windows still uses the Word rendering engine for HTML email. This means it doesn't support flexbox, doesn't support CSS grid, doesn't support background images on most elements, doesn't support modern unit types, and has its own quirks with table widths, margin collapse, and font fallbacks.

If you build a template using anything modern, it'll render fine in Gmail, Apple Mail, Yahoo, Outlook for Mac, and Outlook for Web — and look broken to your enterprise prospect's inbox.

What we built

Every template render in Mailapp goes through MJML compilation. MJML emits table-based HTML that Outlook for Windows can read. Our internal extension to MJML adds: native dark mode variants, BIMI logo embedding, accessibility audits, and language direction handling.

Every template, when saved, runs in our CI against 32 email clients in parallel. Outlook for Windows 2010, 2013, 2016, 2019, 2021, 365, and Web. Gmail dark and light. Apple Mail iOS and macOS. Yahoo, AOL, ProtonMail, Hey, Spark. The works.

Each render gets pixel-diffed against the previous baseline. Significant drift on any client fails the build.

Why we don't outsource it

There are commercial services that do this — Litmus, Email on Acid, Email Comb. We use Litmus internally and recommend it for one-off checks. But for our customers' templates, we wanted the loop to be tight: change the template, see the render in the inspector, fix it, ship.

Outsourced render testing adds 30-180 seconds to the loop. That's too slow. Our integrated tester runs in 8s p95.

What we wish was different

We wish Outlook for Windows was as stagnant as people assume. It's not. Microsoft has shipped minor changes to the Word renderer over the last few years that have silently broken otherwise-working templates. Twice in 2025, we had to retro-fix templates across our entire library after a Patch Tuesday.

We wish more email tools would treat this as a priority. Most of our competitors quietly tell customers to 'design for the modern clients and accept the Outlook fallback.' That's a way of saying 'we gave up.'

We didn't, and we won't. If you're a sender whose business depends on enterprise inboxes, it matters. It's the kind of unglamorous middle that wins deals.