The Trust Gap in B2B Software Research: What Buyers Are Actually Looking For

Ask any operations leader who has recently led a software evaluation what they found most frustrating, and the answer is nearly always a variation on the same thing: the available research was too shallow, too obviously vendor-influenced, or too outdated to be useful. They did the research anyway, because there was no alternative, but with a persistent background anxiety that they were missing something.

That distance — between the research buyers can reach and the research they actually need — is the trust gap. It is worth being precise about what causes it, because the usual diagnosis ("content marketing ruined everything") is too vague to act on. The gap has specific, observable mechanics.

The Gap Is Visible in the Search Results

You do not need survey data to see the problem. Search for the category terms a buyer would actually type and look at who owns the first page.

In September 2026 we ran that check on the queries our own market uses. For "intent data providers", the entire first page was "best of" roundups — published by Cognism, Clay, Autobound, Close and Single Grain. Not one of those is a neutral party. They are tool vendors and agencies with a commercial interest in which answer a reader walks away with. For "b2b lead generation company", the same pattern: roundups from FullEnrich, MO Agency, SaveMyLeads and BSS Commerce.

Not a single vendor's own product page ranked in either case. That is worth sitting with. The pages buyers actually read when they research a category are almost never written by the companies being evaluated, and almost always written by someone adjacent who is selling something else.

This is not automatically dishonest. A tool vendor can write a fair roundup. But the reader has no way to tell from the page whether they are reading an assessment or a placement, and the incentive structure is not disclosed. The buyer's skepticism is a rational response to an information environment where the provenance of the recommendation is hidden.

The Currency Problem Is Worse Than It Looks

The second failure is staleness, and it is more widespread than the datelines suggest.

The clearest example we found was the query "first party vs third party intent data" — a genuine question a buyer would ask before committing budget. Roughly seven of the nine results on the first page came from a single domain, as near-duplicate pages: "first party vs third party intent data explained", "first party intent data vs third party 2026", "First-Party vs. Third-Party Intent Data 2026", and one page carrying 2027 in its title.

Those are not seven pieces of research. They are one piece of thin content, reshaped and re-dated to occupy more of the page. The 2027 dateline is the tell: the recency signal is a production decision, not a statement about when anyone looked at anything.

A buyer reading that page cannot distinguish it from research. It is formatted like research, dated like research, and ranks where research would rank. The cost is not just a wasted click — it is that the buyer now holds a belief about first-party versus third-party intent data that came from nowhere in particular.

What Makes Research Trustworthy

Independence, structurally enforced. The requirement is not that a publisher promises neutrality — everyone promises neutrality. It is that the separation between paid placement and editorial judgment is written down, published, and observable by a reader on the page. Buyers have sophisticated pattern recognition for violations of this principle, and their default posture is skepticism. The only thing that moves them is being able to check.

Depth that a vendor could not have produced. Surface-level research does not merely under-serve buyers; it actively harms them by conferring false confidence. The working test is whether the research tells a buyer something they could not have learned from the vendor's own marketing. If a comparison table restates two feature lists, it has added nothing but the appearance of diligence.

Currency you can interrogate. Software categories move. A date at the top of a page is not a currency signal unless the reader can tell what was checked and when. "Last updated" on a page where nothing was re-examined is worse than no date at all, because it converts staleness into apparent freshness.

Specificity about fit. Generic assessments are not useful. "Platform A is better than Platform B" is almost never true as a general statement. What a buyer needs is which kinds of organisation, with which operational constraints, in which buying context, are better served by one option — and, critically, where the alternative wins.

The Research Buyers Actually Use

When you ask buyers how they really decide, rather than how they describe the process in a procurement document, a different picture emerges. Three things recur.

They lean on peer networks first. A colleague two companies over who has run the same implementation outranks any published assessment, because the peer has no stake in the answer and can be asked follow-up questions. Published research competes with that standard, not with other published research.

They treat formal research as a starting point, never a conclusion. Its job is to generate a shortlist and a set of questions, not to deliver a verdict. Research that positions itself as the answer is solving a problem the buyer does not have.

They are hunting for blind spots. The most valuable thing a piece of research can do is surface the consideration a buyer had not thought of — the migration cost nobody mentions, the seat-count threshold where pricing changes shape, the integration that technically exists but is maintained by one person. Buyers are trying to find out what they do not know, which is exactly what a feature comparison cannot tell them.

What We Do About It, and Where We Are Exposed

Our model is to own the comparison surface rather than rent it. The six properties in the network are ours, which changes two things.

First, the separation between paid and editorial is ours to enforce and ours to publish. Paid content is labelled every time, and the editorial policy that governs it is public whether or not it helps us sell.

Second, it means we can see what buyers actually compare rather than infer it. When someone puts two vendors side by side on one of our pages, that is a recorded event with a timestamp, not a modelled probability. It is also the raw material for research that a vendor could not produce about itself: which pairs get compared, at which company sizes, in which sequence.

The honest counterweight: owning the surface is also the obvious conflict. We sell demand programs to some of the vendors covered on those pages. That is precisely the structure a sceptical buyer should interrogate, which is why the policy is published and the labelling is not optional. A reader who does not trust us should be able to find the mechanism and judge it, rather than take a paragraph like this one at face value.

We also publish audience sizes as bands rather than precise counts, because a number like "247,000 finance buyers" invites scrutiny it cannot survive, and the width of an honest band is itself a signal. We would rather be checkable than impressive.

The Harder Model

Building research that meets this standard means investing in domain expertise, enforcing independence structurally rather than rhetorically, and choosing depth over volume. It is slower and more expensive than the alternative, which is why the search results look the way they do — seven versions of one page will always be cheaper than one version of seven pages.

The bet is that the cheap approach has exhausted its returns. Buyers have learned to discount it, which means the next unit of shallow content earns less trust than the last. A buying committee that cannot find research it believes does not stop buying software; it falls back on peer networks and vendor demos, and the published research simply stops participating in the decision.

That is the gap worth closing, and it is not closed by producing more.