Google Ads best practices for B2B in 2026: measurement before campaign craft

Check more than twenty published sources against one another and six practices hold up, almost all of them about measurement rather than campaign craft. The evidence is what others have published, not figures from campaigns I have run. I kept a practice only when at least two independent sources back it, and in the two places where Google's documentation and independent data flatly contradict each other, I lay out both sides instead of picking one.

The sources are Google Ads Help, the Google Ads API documentation and developer blog, Google's own Accelerate pages, four independent datasets that disclose their sample sizes, practitioner reporting in Search Engine Land and PPC Land, and agencies that sell Google Ads management. When an agency recommends a tactic it also sells, I say so where the recommendation appears.

One caveat runs through the whole article: almost no page on Google Ads Help carries a publication or last-updated date, and several date only the statistic, in a footnote, never the page itself. "Google's current documentation says this" holds up, while "Google said this in 2026" usually doesn't, so I give a date only where a page states one.

The commercial case for the channel is on the Google Ads service page; this article sets out the evidence behind the practices, contradictions included.

Because in B2B the event that matters happens after the click, and a bidding system can't optimize toward something it never sees. Google Ads Help's offline conversion imports page states the mechanism plainly: "By importing offline conversions, you can measure what happens in the offline world after your ad results in a click or call to your business." Search Engine Land's coverage of B2B lead-gen automation puts it more bluntly: "Connecting your CRM to Google Ads or Microsoft Ads is essential for making automation work in lead generation. This isn't optional."

Three more sources repeat the point without adding anything on how it works. Search Engine Land's lead-quality piece recommends importing "CRM MQLs, demo completions, closed deals" back into the account, and both Directive's 2026 B2B guide (30 October 2025) and The Marketing Blender's 2026 review (2 March 2026), published by agencies that sell the implementation, make it their first recommendation.

Read Google's own uplift figure closely. Its Data Manager page claims advertisers connecting offline and app data "see a 26% average increase in incremental ROAS", footnoted to "Google Data, Global, Measurement, April 2025 - April 2026. Based on Search campaigns bidding to conversion value." The figure measures incremental ROAS on campaigns already bidding to value, not conversions in general, yet Google's own Help page restates it elsewhere as a conversion boost. Connecting the CRM so that bidding follows revenue instead of clicks is the starting point of our methodology, and that 26% isn't the reason to do it.

The import path itself. Google Ads Help's offline conversion imports page carries the deadline in its own words: "Starting June 15, 2026, offline conversions import and enhanced conversions for leads uploads will be migrated to the Data Manager API and blocked in the Google Ads API." The block isn't total, and the next sentence says who it hits: "Developer tokens that haven't sent a request between January 2026 through June 2026 will not be allowlisted for legacy access." The Google Ads API documentation states the same cut-off from the developer side, and the Google Ads developer blog confirms the API "will no longer accept new adopters of offline conversion imports, including enhanced conversions for leads", and adds that developers who had already adopted them "can continue importing using the Google Ads API while they integrate with the Data Manager API."

The destination is enhanced conversions for leads rather than the legacy click-ID import. Google's documentation describes it as "an upgraded version of offline conversion import that uses user-provided data, such as email addresses, to supplement imported offline conversion data to improve accuracy and bidding performance", and recommends upgrading through Data Manager. It also says to "include GCLIDs with your uploaded events whenever possible", which is required where no tag is collecting user-provided data. PPC Land reports the same consolidation from outside Google.

This is the only practice in the article that a platform deadline forces, and the deadline has passed. A new integration can only be built on the Data Manager API, and Google presents the legacy route as something to use while you integrate, not a place to stay, so measurement plumbing still belongs on this year's list rather than next year's. Four sources, three of them Google properties, describe the same migration and the same date.

More than Google documents, according to everyone who has measured it. Nobody disputes the principle: Google's value-based bidding documentation describes it as optimizing "based on the value brought to their business" and requires that advertisers report "two or more unique values" for their conversion goals, and its high-quality leads guidance recommends value-based strategies directly. Directive, an agency that sells the setup, describes assigning tiered values across demo, MQL, SQL, opportunity and closed-won so the system learns which conversions matter.

Compare the platform's floor with everyone else's, because the practical question is which of these numbers to plan around.

Sources: Google Ads Help, no date on page; PPC Land on Demand Gen requirements; Optmyzr; Search Engine Land.

The most useful detail is inside Google itself: its Demand Gen requirement is three times its Search guidance, so the disagreement isn't simply the platform against practitioners. Optmyzr's guide (16 July 2026) also names the trap that tiered values set: the most common value-assignment error, it says, is "accidentally adding numbers together that were never meant to be added", so a lead, an SQL and a closed sale all counted as primary actions inflate the value of a single sale. It also argues that a target ROAS you can't explain in terms of margin or deal economics "is not a target. It is a guess wearing a target's clothes." In a B2B account, that makes value-based bidding something you grow into as volume builds, not a setting you switch on.

None of the published sources answers it. Google pulls one way: its high-quality leads page says the ideal conversion delay is "within 7 days of the ad interaction for timely optimization", and Search Engine Land's Performance Max piece agrees that optimization "should be tied to a meaningful online event, such as a qualified lead submission or an appointment booked". The whole premise of CRM-fed bidding pulls the other way, and the windows the platform enforces show how little room that leaves.

Source: Google Ads Help, guidelines for importing offline conversions and best practices for generating high-quality leads, no date on page.

A B2B cycle measured in months doesn't fit inside 63 days, so the closed-won deal that Directive wants you to assign a value to often lands after the import window has shut. Read together, the sources support bidding to the nearest event that actually tracks revenue (a qualified lead, a booked meeting) and keeping the later stages for reporting. Optmyzr describes the same problem from the other end: of the reasons to hold off on value-based bidding, the one it says trips up lead-gen advertisers most is "conversion lag with no offline pipeline to close it", because the algorithm then optimizes toward whatever early-stage proxy the account reports. Two independent sources describe the tension, and neither offers a way out for deals that close after the import window.

Nowhere else in this article do the platform and the independent data clash this clearly, and the independent side has the larger disclosed sample. Google's Ad Strength documentation states that "advertisers who improve Ad strength for their Responsive Search Ads and Sitelinks from 'Poor' to 'Excellent' see 15% more conversions on average", footnoted to internal data covering 15 to 20 August 2025, a six-day window with no sample size given. Its Performance Max page makes a smaller claim, 6% more conversions, footnoted to February 2024 data that is now dated.

Optmyzr's responsive search ad study (April 2026), across roughly 20,000 accounts, reports the opposite pattern and states that Ad Strength "reflects structural completeness … not how well your ad will actually perform". Compare the direction, not the absolute values, since the two datasets measure different things.

Source: Optmyzr, Google RSA performance study, April 2026.

Neither side is clean: Optmyzr sells Google Ads software and its accounts are self-selected, while Google's figure rests on six days of data and no disclosed sample. Search Engine Land's piece on tactics to drop sides with the data, warning that "chasing an 'excellent' score by stuffing headlines and descriptions can easily result in weaker performance for the sake of a vanity metric". Treat Ad Strength as a completeness checklist, which is what it measures, and never as a performance target.

The sources converge on account-level negative lists and deliberate match-type choices, plus a check on what Performance Max is taking from Search. Directive recommends account-level negatives because they "block irrelevant queries across Search campaigns and the Search inventory within Performance Max campaigns", and The Marketing Blender includes maintaining that list in the framework it uses while naming "broad match without guardrails" among what is not working. Both are agencies that sell the work.

The best evidence on match types is Adalysis's match-type study of 16,825 campaigns, which splits results by bid method instead of lumping all advertisers together. For lead generation on maximize-conversions bidding it finds exact match "captures 23% of impressions but 32% of conversions", and adds: "Phrase match has a higher conversion rate than broad match and by far the worst CPA." That phrase-match result runs against intuition, and it's why the study is worth reading in full rather than in summary.

Then there's overlap. Optmyzr's cannibalization analysis of 503 accounts (February 2025) found keyword overlap between Search and Performance Max in 91.45% of them, but its own conclusion is calmer than that headline: where the overlap was exact, 74.92% of cases showed no significant difference in conversion rate, Search performed better in 18.91% and Performance Max in 6.17%. Overlap is close to universal and usually harmless, so measure it in your own account before treating it as a reason to switch Performance Max off. Google can't even agree with itself on how long a new Performance Max campaign needs before major changes: its Help page on Performance Max for lead generation says at least one to two weeks, or up to six for complex setups or low conversion volume, while the Accelerate version of the same guidance still says at least six weeks.

The migration, then the feedback loop, then bidding, then everything else. A platform change sets that order, not preference: since 15 June 2026 the Google Ads API has refused offline imports from new integrations and from developer tokens that sent none in the first half of the year, and the integrations Google allowlisted can keep importing while they move to the Data Manager API. Google Ads Help's own advice is to use that API "to avoid interruption", so the practical move is to migrate now rather than wait: every other practice here depends on the signal the import carries. The CRM feedback loop comes next because value-based bidding means nothing without it, and value-based bidding comes third because it needs a volume most B2B accounts take time to reach.

Ad Strength and match-type work come last, and Ad Strength arguably doesn't belong on the list at all. Where a practice rests on a single publisher, I've said so, and where Google and independent datasets disagree, both appear with their sample sizes so you can weigh them yourself.

The costs these practices are meant to move are in our Google Ads B2B benchmarks article, which works from WordStream's 2026 benchmark report and its closest B2B proxy verticals. For the demand-creation side of a B2B stack, the matching synthesis is our LinkedIn Ads best practices article, and the LinkedIn Ads service page makes the commercial case for running the two together.

ChatGPT is the newest conversational surface, and our ChatGPT Ads Europe launch analysis separates what's documented from the performance data that doesn't exist yet.

What changed for Google Ads offline conversion imports in 2026?

Since 15 June 2026, offline conversion imports and enhanced conversions for leads have moved to the Data Manager API, and the Google Ads API only accepts them from developer tokens that were already importing and got allowlisted for legacy access (Google Ads Help 2026). That access is there to cover the move, not to replace it, so migrate your integration to the Data Manager API now rather than wait.

How many conversions does value-based bidding need in B2B?

The sources disagree by a factor of three. Google documents 15 conversions in 30 days for Target ROAS, while Optmyzr puts the practical floor at 30 to 50 a month and Search Engine Land at about 30 per campaign, and Google's own Demand Gen requirement is 50 in 35 days. Plan around the higher figure and build volume toward it.

Should a B2B account bid to closed-won revenue?

Not directly. Google recommends an optimization event within 7 days of the click and won't import enhanced conversions for leads more than 63 days after the last click, which is shorter than most B2B cycles. Bid to the nearest event that correlates with revenue and report on the rest.

Does improving Ad Strength improve performance?

Google says yes; independent data says the opposite. Google reports 15% more conversions moving from Poor to Excellent on responsive search ads and sitelinks, while an Optmyzr study of roughly 20,000 accounts finds Excellent ads carry the worst cost per acquisition (Optmyzr 2026). Treat it as a completeness score, not a target.

Is Performance Max stealing traffic from Search campaigns?

It overlaps with Search almost everywhere, usually without doing harm. Optmyzr found keyword overlap in 91.45% of 503 accounts, but no significant conversion-rate difference in 74.92% of exact-overlap cases (Optmyzr 2025). Measure the difference in your own account before restructuring.