Forrester’s Buyers’ Journey Survey for 2025 puts a number on something most B2B teams already suspect. A purchase decision now involves an average of thirteen people inside the buying organisation and nine from outside it, and 73% of purchases involve three or more departments.
Read that second figure again. Nine of the twenty-two people shaping the decision do not work for your customer. They are peers at other companies, consultants, people in a community or a group chat, a former colleague who used the product at their last job. There is no CRM record for them. No form fill, no cookie, no email address, no account to attach them to.
Most advice about this problem arrives at the same conclusion: connect your signals. Get sales, marketing and customer success looking at the same data, agree on what counts as buying momentum, and act on it together. That advice is correct, and this post is not an argument against it. It is an argument that it stops short, because a growing share of the research those twenty-two people do now produces no signal for anyone to connect.
What the standard advice gets right
The fragmentation is real. In most organisations sales sees intent data marketing does not, support sees product questions nobody else reads, and finance sees a procurement conversation that started weeks before anyone told marketing. Each function holds a fragment of the same account and reaches a different conclusion about it.
Fixing that is genuine work and it pays. If three teams can see the same account the same way, the organisation can respond to a live opportunity instead of discovering it afterwards. We wrote about the structural version of this problem in the five capabilities a B2B marketing system needs, and about the measurement version in full-funnel attribution for B2B.
So connect the signals. Then notice what is still missing.
The part connected signals cannot fix
Forrester’s own description of how these groups behave includes a detail worth sitting with: buyers consult colleagues, peers, communities and AI tools long before they ever talk to a provider.
Three of those four leave you nothing. A conversation between two heads of engineering at a conference is invisible to every dashboard ever built. A recommendation in a private community is invisible. And an AI assistant answering “what are the options for X and which one suits a company like ours” is invisible in a new and more final way, because the person got a complete answer and had no reason to visit anybody’s website at all.
This is the part that is genuinely new. Search at least left a trace. Someone searched, clicked, landed, and you had a referrer and a query. An assistant returns a synthesised answer, names two or three vendors, and the session ends there. If your company was named, you will not know. If it was not, you will not know that either. There is no signal to connect, because no signal was created.
So a team can do the integration work perfectly, unify every system, agree on every definition, and still be blind to the conversation that decided the shortlist.
Three questions that replace “where did the lead come from”
This is Brilad’s framing rather than an industry standard, but it is the sequence we use with clients, and it holds up because it does not depend on being able to track anything.
First: when an assistant is asked our category question, are we named? Not whether we rank, whether we are named. This is directly checkable. Write down the fifteen or twenty questions a buyer in your category would actually ask, and run them against ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot. Record who gets named and what gets cited. You now know something concrete about a stage of the journey you previously could not observe.
Second: when we are named, what did the assistant read? The cited source is the most useful piece of information in the whole exercise. If an assistant cites your own documentation, your content is doing the work. If it cites a comparison site, a directory, a review platform or an analyst note, then the third-party layer is doing the work and that is where the next investment goes. If it cites nothing of yours, you have your answer about priorities.
Third: is what it says about us correct? An assistant that recommends you confidently and describes a product you discontinued, a market you left or a price you no longer charge is costing you deals in a way no dashboard will ever flag. Accuracy is a visibility metric, not a branding one.
What to do about the nine people outside the account
You cannot market to people you cannot identify. You can, however, be present in the places they and their assistants look.
- Write the answer, not the brochure. A page that opens by claiming market leadership gives an assistant nothing to quote. A page that opens by stating what the product does, who it suits, what it costs and where it falls short gives it a paragraph it can lift. The second kind gets cited.
- Answer the questions your support queue already receives. Those are the questions the nine outsiders are being asked by the thirteen insiders, phrased in the words real people use. We covered the method in using customer support data to plan content.
- Invest in the third-party layer. Comparison pages, directories, review platforms, community threads and analyst mentions are what an assistant reaches for when it does not trust any single vendor’s own site, which is most of the time for a “which should we choose” question. Accurate, complete, current listings are unglamorous and they do more than another blog post.
- Make the entity unambiguous. One company name, one spelling, consistent descriptions across every profile you hold, linked from your own site. An assistant that cannot work out whether two mentions refer to the same company tends to cite neither.
- Say the limitations out loud. Publishing who the product is not for is the single most quotable thing most B2B companies refuse to do. It also pre-qualifies the nine outsiders, who are frequently the ones asked whether you are a sensible fit.
How to tell whether it is working
The honest answer is that you will not get clean attribution for this, and you should stop waiting for it. What you can get is a repeatable measurement that does not depend on referrers.
Run the same question set against the same assistants on a fixed schedule, monthly or quarterly, and record four things each time: whether you were named, in what position, which source was cited, and which competitors appeared alongside you. Keep the question set fixed, because the comparison between runs is the whole value. Changing the questions to flatter the results is the easiest way to make this exercise worthless.
Then connect that to the thing you can still see. Pipeline created in accounts where nobody filled in a form first. Enquiries that arrive already knowing the price and two of the competitors. Sales calls that open with a question nobody asked a year ago. These are soft indicators and they are not a substitute for the direct check above, but a team that is being cited in assistants usually notices the change in the first conversation before it sees it anywhere else.
The shorter version
Connecting your signals is necessary, and most organisations should do it. It will not show you the part of the buying journey that now happens inside an assistant, because that part produces no signal at all. The response is not a better dashboard. It is to be the source the assistant reads, to be corroborated by the third-party pages it trusts, and to check what it says about you on a schedule rather than waiting for a referrer that is never going to arrive.
Twenty-two people are deciding. Nine of them you will never meet. The question is whether the answer they are handed has your name in it, and whether that answer is right.
Source: Forrester, “Three Realities About B2B Buying Networks”, citing Forrester’s Buyers’ Journey Survey, 2025.