Microsoft Advertising is now live inside HubSpot, and a change of lifecycle stage can be sent back to it as a conversion. LinkedIn and Google already take the same kind of feed. A lead definition written to settle arguments with sales is now telling ad platforms whom to buy more of.
On 30 September Microsoft told advertisers that "Microsoft Advertising is now live in HubSpot." Most of the coverage focused on audiences: CRM lists, LinkedIn profile targeting across Bing, Copilot and Outlook. The more consequential detail sits in a HubSpot help article updated on 4 October. It explains how to "use lifecycle stage changes as a trigger to create an event that sends a conversion signal," which Microsoft then uses to "measure conversions and optimize ad delivery."
Put plainly, when a contact moves from lead to marketing-qualified lead in HubSpot, Microsoft's delivery system can now learn from it. HubSpot already syncs ad conversion events to Google, Meta, LinkedIn and TikTok. Microsoft completes the set of networks most B2B paid budgets run through.
The platforms have been asking for this for months. Since LinkedIn's August API release, its Conversions API has accepted marketing qualified lead and sales qualified lead as separate conversion types, used by its qualified-leads optimisation goal "for deeper funnel campaign optimization." September's release added a 180-day attribution window for both. Then on 13 October Google announced a "Data Strength Uplift" metric in Google Ads, meant to show advertisers "the conversion gains driven by your data setup."
Each change is reasonable on its own terms. Taken together, they mean the MQL is no longer an internal arrangement.
A threshold written for a different argument
In most B2B organisations the MQL is a negotiated number. Marketing commits to a volume; sales agrees to work what crosses the line; the score that defines the line gets adjusted whenever either side loses patience. It is lowered in a thin quarter and raised after a bad month of complaints from reps. Nobody pretends it is a scientific measure of buyer intent. Its job is to stop an argument.
That was tolerable while the consequences stayed inside the building. Once the stage is synced to an ad platform it becomes a training label. The bidder does not know the threshold was cut in March to rescue a quarter. It learns that the people who crossed the line are the people to find more of, and it goes looking.
Lead scoring used to decide whom sales would call. Synced to an ad platform, it also decides whom marketing pays to reach.
Where the signal goes wrong
The most common failure is silent retraining. A scoring change made in the marketing automation platform alters bidding behaviour with no change logged in the ad account. Cost per lead moves, the paid media team blames creative or seasonality, and the person who edited the score never hears about it. HubSpot's own documentation adds a wrinkle: only stage changes that happen after a conversion event is created are counted. History does not carry over, so every redefinition lands inside a data series the platform treats as continuous.
Then there is the cheapest trip-wire. If two content downloads and a pricing-page visit are enough to make someone an MQL, an optimising system will become very good at finding people who download content and look at pricing pages. Students, competitors and consultants doing research all qualify. Goodhart's law did not need a budget attached to cause trouble. Now it has one.
The third problem is structural. A sales-qualified lead is the more honest label, but it arrives weeks later and in far smaller numbers. When LinkedIn launched qualified-leads optimisation in 2025, it suggested sending at least five qualified leads every two weeks. Plenty of enterprise programmes fall short of that on a single platform once they move from MQL to SQL. The 180-day window lets long cycles earn credit eventually; it does not make a learning system that hears about a good lead six weeks late learn any faster. Teams that cannot feed the deeper stage fall back to the shallower one, which returns them to the trip-wire.
Figure 1
Stage sent to the platform | Volume and delay | What the bidder learns to find | Main risk |
|---|---|---|---|
Form fill or raw lead | Highest volume, near real time | People who complete forms | Optimises toward form-fillers, whatever their fit |
Marketing-qualified lead | Moderate volume, hours to days | People who trip your scoring model | Inherits every compromise in the score; retrains when the score changes |
Sales-qualified or sales-accepted lead | Low volume, days to weeks | People sales chose to work | Too few events to learn from on one platform; reflects rep capacity as well as buyer quality |
Opportunity or later stage | Very low volume, weeks to months | Closest proxy for pipeline | Often too sparse to optimise on; best used to check the shallower signals |
Choosing which lifecycle stage to send. Editorial guidance, not platform documentation. What each network accepts differs: LinkedIn names MQL and SQL types explicitly; the HubSpot–Microsoft sync triggers on any lifecycle stage change. Volumes and delays are indicative and vary by programme.
One more caution about comparisons. Teams that send MQLs to LinkedIn, SQLs to Microsoft and form fills to Google, then rank the three by cost per conversion, are comparing three different things. HubSpot notes that Microsoft "may calculate conversions differently" from HubSpot itself, so even an identical stage can produce different counts in the two systems.
Consent travels with the label
Sending a lifecycle stage tells an ad platform something about a named person: that your company has judged them a likely buyer. HubSpot's setup guidance asks advertisers to confirm that each contact whose data they share "has given permission." Microsoft says the consumer data behind its new LinkedIn company lists excludes users in the EEA, the UK and Switzerland. For any programme with European contacts in the CRM, the lawful-basis question belongs to the privacy team. It should not be settled by whoever in paid media happens to click connect.
What marketing operations now owns
The lifecycle model has become an input to media spend, and it needs the controls that implies. Most of them cost nothing to put in place.
Before a lifecycle-stage sync goes live
Put stage definitions and scoring rules under change control, with paid media named as a stakeholder who is told before any change ships.
Keep one log that records scoring changes and media changes together, so a moving cost per lead can be traced to its cause.
Decide, per platform and in writing, which stage is sent and why, based on the monthly volume that platform actually produces.
Report cost per conversion only across platforms receiving the same stage; label everything else as not comparable.
When switching to a deeper stage, keep a small campaign on the old signal for a few weeks as a control.
Route EU and UK contacts through privacy review before they are included in any sync.
None of this argues against sending CRM stages to ad platforms. Optimising on form fills is worse, and the platforms are right that they cannot find better buyers without being told who the better buyers were. The problem is that most lead-scoring models were built to manage a relationship with sales and are now being asked to brief a buying machine.
The scoring model was always a statement about which buyers marketing valued. Now it also places the media orders.
Sources and notes. Microsoft Advertising, "Less busywork, more growth: What's new in Microsoft Advertising this September," 30 September 2026 (HubSpot integration live; LinkedIn company lists raised to 10,000; underlying consumer data excludes EEA, UK and Switzerland). HubSpot Knowledge Base, "Create and sync ad conversion events for your Microsoft Advertising account," last updated 4 October 2026 (lifecycle stage changes as conversion triggers; only changes after event creation counted; consent requirement; Microsoft "may calculate conversions differently"; also lists Google Ads, Meta, LinkedIn and TikTok syncs). Search Engine Land, Anu Adegbola, 1 October 2026; PPC Land, 6 October 2026. LinkedIn Marketing API recent changes, versions 202608 (MQL and SQL conversion types) and 202609 (180-day attribution window), page dated 16 September 2026. LinkedIn qualified-leads optimisation guidance (at least five qualified leads every two weeks) as reported by Social Media Today, 22 April 2025; current guidance may differ. Google, "Build your data strength with October's Demand Gen drop," 13 October 2026. Figure 1 is editorial analysis. MarketingHubMedia has no commercial relationship with any firm named. Journalism, not procurement or legal advice. Corrections welcome.
