Why B2B marketing attribution is degrading, and what to measure instead

Attribution is not a record of why a deal happened. It is a rule for assigning credit among the interactions your systems could observe. That distinction has always mattered, but it is harder to ignore now that consent choices, browser controls, multi-channel research and AI answers remove or model parts of the path before a buyer reaches your website or speaks to sales.

The dashboard may still allocate every conversion across paid search, organic search, direct traffic and other channels. A total of 100% only means the attribution rule distributed all the credit it was given. It does not mean the system observed 100% of the buying process, and it does not prove that the credited channel caused the conversion.

Some missing signal can be estimated, and some AI traffic can now be measured. Google Analytics models certain unobserved key events when its quality thresholds are met, ChatGPT adds a source parameter to outbound publisher links, and Google has begun testing dedicated reporting for impressions in generative AI search features. What remains unavailable is a complete, cross-platform chain connecting every influence to every opportunity.

The practical answer is not to replace last-click attribution with another single source of truth. B2B teams need three different views: what was observed, what changed because of marketing and what buyers say influenced them. Each view answers a useful question, provided the report does not pretend they are interchangeable.

What attribution can and cannot tell you

Google Analytics defines attribution as assigning credit to ads, clicks and other factors along a user path. The important verb is assigning. An attribution model applies a rule to the events available inside the system; it does not independently establish what would have happened without the marketing activity.

Consider a buyer who hears a supplier discussed by a colleague, sees its name in an AI answer, searches for the company and later completes a form. Web analytics may record the final search and form submission accurately. A last-click rule can reasonably say that organic search receives credit for the recorded conversion. It cannot conclude that organic search created the demand, because the model never observed the colleague’s recommendation, the AI answer or a comparable buyer who did not receive those influences.

This is the difference between a trace, an attribution rule and a causal estimate.

Google makes the same distinction in its own measurement documentation. Standard attributed conversions follow the campaign’s tracking settings and credit rules, while Conversion Lift compares treatment and control groups to estimate conversions that would not otherwise have occurred.

Both numbers can be useful. They should not carry the same label.

Where the visible path gets thinner

The loss of visibility does not come from one change, and it does not affect every company in the same way. The practical effect depends on the browser mix, consent implementation, sales cycle, marketing channels, analytics configuration and the amount of buying activity that happens outside company-controlled systems.

Browser and consent controls change what can be joined

Safari’s tracking prevention blocks third-party cookies by default and reduces third-party referrers to their origin. Firefox isolates third-party cookies and blocks known trackers through Enhanced Tracking Protection. These protections make cross-site identity and detailed path reconstruction less complete, although they do not make all first-party website measurement disappear.

Consent adds another boundary. With basic Google consent mode, for example, Google tags do not load before consent and send no data when consent is denied. With advanced consent mode, cookieless signals may support advertiser-specific modeling. Google’s consent documentation is explicit that the observed and modeled data depend on which implementation is used.

That means a report should distinguish events that were recorded directly from events estimated by a platform model. Combining both may produce a useful planning number, but describing the total as fully observed gives the audience the wrong level of confidence.

Buyers use sources your analytics system does not own

A B2B purchase is rarely contained within one browser history. Buyers move among supplier websites, search, video calls, in-person conversations, colleagues and third-party experts. McKinsey’s 2026 Global B2B Pulse, based on nearly 4,000 decision-makers, found an average of ten channels across the buying journey. Gartner separately reported that buyers used an average of seven information sources in a recent purchase.

Some of those interactions create a digital trace and some do not. A private message may prompt a direct visit, an analyst conversation may prompt a branded search, and a recorded webinar may be shared internally without the eventual decision-maker using the original link. Calling all of this “dark social” gives the gap a memorable name but does not make it one measurable channel.

A more useful term is unobserved influence. It describes a limitation in the company’s evidence without guessing which invisible interaction deserves the credit.

AI discovery is no longer a complete blind spot

The previous version of this argument was simpler: an AI product could recommend a supplier, the buyer could search for the brand, and the supplier would receive no information about the recommendation. That still happens, but the statement that AI discovery offers no impression or referral data is no longer accurate in 2026.

Google announced dedicated generative AI performance reports in Search Console in June 2026. The reports show impressions, pages, countries, devices and dates for appearances in AI features on Search and Discover. The initial rollout is limited to a subset of websites, so this is not yet a universal source of data.

OpenAI tells publishers that links from ChatGPT automatically include utm_source=chatgpt.com, which means a clicked source link can appear in web analytics. ChatGPT Search also uses factors such as location and, when enabled, relevant memory to improve results, according to OpenAI’s search documentation. A manual citation audit therefore needs controlled sessions and a recorded test context if the team wants to compare results over time.

These developments improve visibility, but they measure different things. An impression in a Google AI feature, a clicked ChatGPT source and a model naming a brand in a controlled test are not equivalent events. None of them independently proves that the appearance caused an opportunity.

Citation visibility is therefore a useful diagnostic measure, not a new form of revenue attribution. It can show whether the company is present in the answers buyers may see, which claims are being repeated and which third-party sources shape the response. To make that measure comparable, keep the buyer questions, location, session conditions, engine, model and date visible in the record. The separate guide on testing whether AI search recommends you explains a practical baseline method.

Use three questions, not one attribution model

Most reporting problems become easier when the team stops asking one dashboard to answer three different questions.

Observe the trace

Observed data is the operational foundation. It answers whether a campaign delivered, which pages people used, whether a form worked, how an opportunity moved and whether revenue was recorded. It is also where basic defects appear: missing campaign parameters, duplicated events, inconsistent CRM source fields, broken offline conversion imports and stages that different teams use differently.

Clean observation matters even when the path is incomplete. An incrementality study cannot rescue a CRM whose outcome data is unreliable, and a buyer survey cannot tell you whether the form was broken for a week.

Attribution can sit inside this layer as a reporting convention. Name the model, scope and lookback window, then describe the result as assigned credit. Pipeline influence reports belong here as well. They show that an account had one or more recorded interactions within a defined period, which is useful for account planning, but association with an opportunity is not the same as incremental effect.

Estimate the incremental effect

Incrementality asks what would have happened without the activity. The strongest design creates a credible comparison group that is similar to the exposed group except for the marketing treatment. In practice, that may mean a platform user holdout, a randomized geographic test or another controlled design suited to the channel.

The test must match the decision. A study measuring form submissions over two weeks may say little about qualified pipeline in a long enterprise sales cycle. Spillover between regions, insufficient conversions, simultaneous campaigns and changes in pricing or sales coverage can weaken the comparison.

Geo tests are therefore not automatically cheap or simple. Google’s own documentation says geography-based Conversion Lift typically requires a higher budget, while its research on randomized paired geo experiments explains the difficulty created by small numbers of heterogeneous regions. An inconclusive test is a valid outcome when the study did not have enough power to separate the effect from normal variation.

Learn from buyers

An open-text “How did you hear about us?” question can recover influences that event logs miss, including a colleague, community, event, analyst or AI product. Keep it open so the answer is not restricted to the channels already present in the dashboard, then categorise the responses without deleting the original wording.

This is buyer-reported influence, not self-reported causality. Recall is incomplete, respondents may name the most memorable interaction rather than the first one, and the person completing the form may not know what influenced the rest of the buying group. Interviews and win-loss research can add the context that a short form field cannot hold.

The value comes from triangulation. If buyer interviews repeatedly mention a source, referral data shows relevant visits and an incrementality test detects lift when the activity changes, the combined case is stronger than any one measure. The evidence should remain separate in the working data so its different limitations are still visible.

Choose the method that fits the decision and the data

There is no universal replacement for multi-touch attribution. The appropriate method depends on the question, the available variation, the number and timing of outcomes, and whether the team can create a credible counterfactual.

A marketing mix model can estimate effects across channels without reconstructing individual paths, which makes it useful in a more privacy-constrained environment. It is still a model rather than an automatic truth machine. Google’s Meridian documentation states that the goal is causal estimation, but causal quality is difficult to validate without well-designed experiments. Meridian therefore supports using relevant experiment results to calibrate channel priors, while warning that experiments and models may define different effects and time periods.

For a smaller B2B company with low conversion volume, a full MMM or platform lift study may not be feasible. That does not justify returning to false precision. The sensible approach is to improve observed data, collect buyer evidence consistently, test a decision at the highest practical level and state where the result remains directional.

A practical measurement reset

The first reset does not require a new dashboard. It requires clearer labels, cleaner inputs and one decision worth testing.

1. 01Label the evidence already in the report

   Mark every important number as observed, attributed, modeled, experimental or buyer-reported. If the team cannot identify which one it is, the metric is not ready for a leadership claim.

2. 02Audit the path from event to revenue

   Check campaign parameters, referral handling, consent states, key events, CRM source fields, duplicate records, opportunity stages and offline outcomes. Record where modeled data enters the total.

3. 03Add one open-text source question

   Keep the original answer, classify it afterwards and review the patterns alongside interviews and sales evidence. Do not force every answer into the existing channel taxonomy.

4. 04Measure AI discovery as visibility

   Separate AI feature impressions, clicked referrals and controlled citation results. Track each over time without merging them into a fictional AI-sourced revenue number.

5. 05Choose one investment decision to test

   Write the hypothesis, counterfactual, outcome, test window and conditions that could invalidate the result before choosing a platform lift study, geo test or model.

This sequence also exposes a common problem: the company may have more measurement tools than decision discipline. A test is valuable when leadership has already agreed what it would change if the result is positive, negative or inconclusive.

How to report without a 100% pie chart

A useful leadership report does not need to allocate every deal. It needs to show what the company knows, how it knows it and which decision follows.

The resulting board statement might read like this: paid search produced the recorded conversions shown under the selected attribution rule; the lift study estimates a smaller incremental effect within the reported interval; buyer interviews frequently mention two untracked research sources; and the team cannot assign the remaining influence at person level. Each clause describes a different kind of evidence without asking them to sum to 100%.

What good looks like

Good marketing measurement does not reconstruct every buyer journey. It makes better decisions with incomplete evidence.

The observed record should be clean enough to run the business. Attribution should be named as a credit rule rather than disguised as causality. Modeled values should be labelled as estimates. Incrementality work should show its counterfactual and uncertainty. Buyer research should retain the language and context that event data cannot capture.

AI discovery has not made measurement impossible, and new platform reporting is recovering some of the signal. It has made the boundary more obvious: a clicked source, a search impression, a buyer’s recollection and an incremental effect are four different facts.

The most credible report may leave part of demand unallocated. That is not a measurement failure. It is evidence that the team understands where its data ends.

Questions B2B teams ask

Is marketing attribution still useful?

Yes, when it answers a limited question about observed interactions. Attribution can describe which recorded touchpoints received credit under a stated model. It should not be presented as proof that those touchpoints caused the sale or as a complete account of how the buyer formed a preference.

What is the difference between attribution and incrementality?

Attribution assigns credit among interactions that were observed. Incrementality estimates what changed because marketing happened by comparing the observed outcome with a credible counterfactual, often through a randomised test, holdout or well-designed quasi-experiment.

Should B2B companies use self-reported attribution?

Use a short open-text question as one source of evidence, not as the single source of truth. Buyer recall is incomplete, but the language people use can reveal private recommendations, events, communities and other influences that clickstream data cannot see. Preserve the original answer before categorising it.

Can AI search traffic be measured?

Some of it can. Referral parameters, first-party analytics and newer reports from search platforms can reveal visits, impressions or citations on supported surfaces. They still do not provide one complete, person-level path from every AI answer exposure to a closed B2B deal.