Thirteen people decide the average B2B purchase. Your CRM has objects for a lead, a contact, an account and an opportunity — and none for the group. Everything downstream inherits that gap.
"Sell to the buying group" has been settled advice for the better part of a decade. Forrester's research puts an average of thirteen people inside a purchasing decision, and 38 per cent of B2B organisations selling to groups of ten or more. Nobody in revenue operations disputes the finding. Most have a slide about it.
What almost nobody has is a place to put the group.
The CRM object graph that every commercial process runs on was designed around a different assumption: a person raises their hand, gets scored, gets routed, gets worked, becomes an opportunity. Lead, Contact, Account, Opportunity. There is no first-class object representing "the eleven people at this company currently evaluating us, and their relationship to each other". Buying groups therefore exist in slideware and in reports assembled after the fact, and nowhere in the system at the moment a decision is being made.
That gap is not philosophical. It shows up as four specific, diagnosable failures.
1. Lead-to-account matching, which is more fragile than anyone admits
Every buying-group capability depends on correctly attaching a person to an account. Native matching in most platforms leans on exact company names or clean email domains. Real inbound data does not arrive that way: company names come in abbreviated, localised, misspelt or entered as a product division; contacts arrive on subsidiary domains, regional domains or personal addresses.
When matching fails, the person does not appear as an unmatched record needing attention. They appear as a fresh, unrelated lead — which is precisely the outcome that hides a buying group rather than revealing one.
2. Deduplication, which quietly deletes the signal
This is the one worth stopping on, because it is a hygiene process, universally run, and it works directly against buying-group visibility.
Traditional deduplication treats additional people from the same account as an administrative inconvenience to be suppressed. A second person from an account arriving three weeks after the first is, in buying-group terms, one of the strongest intent signals available — the committee is forming. In dedupe terms, it is noise.
The record your data-hygiene rules are designed to suppress is the same record your buying-group strategy depends on. Most revenue operations functions are running both at once and have not noticed the contradiction.
3. Enrichment coverage, which is uneven in a patterned way
Buying groups are only visible if the people in them are visible. Single-vendor enrichment leaves consistent, predictable holes: EMEA contacts, employees of subsidiaries, technical practitioners, security roles, operations titles and smaller business units.
Read that list again as a description of a modern buying committee. Those are not incidental gaps at the edges. They are the technical evaluator, the security reviewer and the operations owner — three of the people most likely to kill a deal, and three of the people least likely to be in your database.
4. Routing, which was built for one inbound event
The standard routing pattern is unchanged from the era it was designed in: person fills in a form, person is scored, person is assigned. Run that pattern against eleven people from one account arriving over six weeks and the result is a scattering — different queues, different owners, different nurture tracks, several of them contacting the same company independently.
The account gets a coordinated evaluation from its side and an uncoordinated response from yours.
Visual 1 — Where the buying group is lost, and what it looks like when it happens
Break point | Designed to do | What it does to a buying group | Symptom in the pipeline |
|---|---|---|---|
Lead-to-account matching | Attach a person to a company record | Fails on abbreviations, subsidiaries and non-corporate domains | Committee members appear as unrelated net-new leads |
Deduplication | Keep the database clean | Suppresses the second and third person from an account | The strongest early intent signal is deleted as a duplicate |
Enrichment | Fill in missing attributes | Systematically under-covers EMEA, subsidiary, technical, security and operations roles | Deals stall on evaluators who were never in the CRM |
Routing | Assign an inbound record fast | Treats each arrival as an independent event | Several reps working one account without knowing it |
How to read it: None of these is a buying-group feature failing. All four are core data operations doing exactly what they were configured to do, against a use case they were not designed for.
Why the platform feature does not settle it
Most major platforms and several specialists now ship something described as buying-group support — an object, a container, a grouping construct that assembles related people under an account and an opportunity. These are real and useful. They are also downstream of all four failures above.
A buying-group object populated by a matching process that misses subsidiary domains contains a partial group. A buying-group object fed by a deduplication rule that suppresses second contacts contains a shrinking group. The construct inherits the quality of the data operations underneath it, and the sequencing question — fix the plumbing or buy the object — is the one that decides whether the investment produces anything.
Buyers evaluating in this category should ask vendors which of the four break points their product addresses directly, and which it assumes you have already solved.
What the published evidence supports, and what it does not
Evidence in this category is thin and almost entirely vendor-published, which is worth stating plainly. Openprise, a data-operations vendor with a commercial interest in the answer, publishes customer outcomes including a 130 per cent improvement in lead-to-account match rates at Equinix, a 48 per cent increase in contact match rates at JumpCloud alongside a tripling of addressable market within 90 days, and an improvement at Palo Alto Networks from 50–60 per cent enrichment coverage with a single provider to above 85 per cent using a multi-vendor waterfall.
Those are customer references, not independent measurement, and none of them isolates revenue impact. What they do establish is direction and rough magnitude: matching and coverage rates in large, well-resourced organisations sat well below where their owners assumed, and moved substantially when addressed. If those are the baselines at Equinix and Palo Alto Networks, the working assumption for a mid-market stack should not be better.
Five things to measure before buying anything
What proportion of inbound leads match to an existing account automatically? Measure it this quarter. It is the ceiling on every buying-group capability you might buy.
How many records did deduplication suppress in the last 90 days that belonged to accounts with open opportunities? This is usually an uncomfortable number and is rarely reported.
What is enrichment coverage by region and by function, not in aggregate? The aggregate hides exactly the roles that matter.
How many accounts had more than one person routed to more than one owner in the last quarter? That is your uncoordinated-response rate.
Which of the four break points does the vendor you are evaluating actually fix, and which does it presume?
What this changes
The buying-group conversation has been conducted at the level of strategy for years, and it has been largely correct and largely inert, because strategy does not survive contact with an object model that cannot represent it. The useful reframing for a RevOps leader is that this is not a go-to-market initiative requiring a new philosophy. It is a data-operations programme with four named, measurable failure points, three of which can be quantified from the existing system in an afternoon.
Start there. A buying-group object sitting on top of unfixed matching and aggressive deduplication will produce a tidy report of a group that was never fully in the database — which is worse than having no report at all, because it will be believed.
Sources and method. A SalesHubMedia original. Buying-committee size figures — an average of thirteen people involved in a purchasing decision, and 38 per cent of B2B organisations selling to groups of ten or more — are Forrester research as cited by Openprise, which is also the source for the four described data-operations failure modes and for the customer outcomes attributed to Equinix, JumpCloud and Palo Alto Networks. Openprise is a data-operations vendor selling into this category and those outcomes are vendor-published customer references, not independent measurement; they are reported here as directional evidence and labelled as such. Platform buying-group capabilities are referenced as market context, not endorsements. The four-break-point framework, the deduplication argument and the diagnostic measures are SalesHubMedia's own analysis. Journalism, not procurement advice. Corrections will be made openly on this article.
