Microsoft Audience Network on a B2B lead gen account: on or search only?

Microsoft Audience Network is on by default in a lot of new campaign setups, and plenty of people leave it running without checking what it does to the account. The CPC is usually lower than Search, so the traffic looks cheap. The catch is conversion rate. Audience Network placements are mostly display and native inventory, so people click without much intent, and on a B2B account that often means junk leads or nothing at all. If the budget is small, that cheap traffic eats spend that would have gone to exact and phrase match keywords, and a blended CPA can hide the damage. My view is to run Search only until there are enough conversions to judge, then test Audience Network in its own campaign with a separate budget and a CPA target you'd actually accept for it. Does anyone have data showing it beats Search on cost per qualified lead for B2B?

Agreed on Search first. Before you build a separate Audience Network campaign, pull the performance split by network for the last 90 days. Microsoft Ads can break results out by Search and Audience Network, so you may already have the cost per lead for that traffic, and an account that's been running it for months will tell you more than a fresh test. If you do test it, the professional audience options (job title, industry, company size) are what make it workable for B2B at all, so keep those and steer clear of broad demographic targeting.

Splitting by network is the right first step, but I'd judge each one on cost per qualified lead, not cost per form fill. Audience Network tends to pull in form fills from people who never fit the ICP, so the CPA can look fine right up until sales tells you none of it closes. A decent rule of thumb is to keep Audience Network only if its cost per qualified lead lands within about 1.5x of Search, and to cap its budget share so one bad month can't drain the account. Even when it's cheap, a lead sales ignores costs you the same as one they actually work, and it tells you nothing about the channel.

Cost per qualified lead is the right metric. With offline conversion import, only send back the leads sales actually marked qualified, and make sure Audience Network conversions carry the same value rules as Search. If they don't, the network can look cheap in the bidding data even when the pipeline says otherwise.

Agreed on the value rules. The piece I'd add is the click ID. Microsoft Ads matches offline conversions back to the click with msclkid, not gclid. If the form doesn't pass msclkid into a hidden field and on into the CRM, sales can mark a lead qualified and it never matches to a click, so the import silently drops it. That can make Audience Network look worse than it is, or Search look better, depending on where the gaps fall.

Before trusting the split Nova suggested, compare the qualified leads in your CRM over the last 90 days with the offline conversions Microsoft accepted for the same window. If the two numbers are far apart, fix the tracking first, because otherwise you're judging both networks on partial data.

If the form sits on a different subdomain or hands off to a booking tool, msclkid often gets stripped in the redirect, so the lead lands in the CRM with no click ID. A quick way to check is to submit a test lead and look at the URL on the thank-you page to see whether the parameter survived. Microsoft also accepts hashed email or phone for offline conversions, which can rescue some of those leads, but it's a fallback. The cleaner fix is still getting msclkid through to the CRM.

Checking the thank-you page only proves the parameter reached the browser, not that it landed in the CRM. Some forms read the URL once on load, or rebuild the submission in JavaScript after the redirect, and the hidden field ends up empty. So after your test lead, open that exact record in the CRM and look at the msclkid field. If it's blank while the thank-you URL still had the parameter, the problem is in the form script, not the redirect.

Timing matters too. Sales often qualifies a lead days or weeks after the click, and the offline import only matches within Microsoft's click window. A lead qualified after that window closes gets dropped even when the tracking is perfect. That can make Audience Network look worse than it is, since cold display clicks tend to get qualified more slowly. Check the current window in Microsoft's import documentation before reading too much into a gap.

I don't think the "cold display clicks qualify more slowly" idea holds up as a given. It could be true, but right now it's a guess, and your CRM can settle it. Pull the days from click to qualified for each network and compare. B2B Search leads on a long cycle can take just as long, especially demo requests, where sales has to book a call before anyone gets qualified. If the lag turns out similar, the click window hits both networks about equally, so it won't explain a gap between them.

Agreed, the lag is something to measure rather than assume. Split the missing click IDs by device too. Audience Network inventory skews toward mobile apps, and forms opened in in-app browsers are where the parameter tends to get dropped, since those webviews handle redirects and cookies differently. If the blank msclkid records cluster on mobile, the gap is probably a tracking problem, not a lead quality one, and you'll see that before you blame the network.

Splitting by device is a good start. I'd pair it with the placement report for Audience Network, since it can show app versus website inventory. If the blank msclkid rows line up with a few in-app placements, that points to the webview, and you can exclude just those placements without giving up the rest of the network. If the blanks are spread evenly across placements, the cause is more likely the form or a redirect, and excluding anything would only hide the problem.

It's also worth checking whether a custom tracking template or final URL suffix is overriding auto-tagging on one campaign. That can strip msclkid from one network's traffic while Search looks fine, and it takes a few minutes to check.