Most of the pipeline is fiction

Every Monday, somewhere, a revenue leader opens the pipeline view and makes decisions from it. Here is what that view is actually made of, measured across the B2B organizations whose CRMs are connected to Stalar: more than 125,000 deals and nearly a quarter of a million contacts.

What the deals say

Of the CRMs with meaningful deal volume, the two largest had 70 and 97 percent of their open deals past the close date the deal itself claimed. A third ran at 41 percent. One, and only one, kept it in single digits, which proves the discipline is possible and shows how rare it is. A past-due open deal is a deal somebody moved into the pipeline with intent and then stopped telling the system about. It may have died, or closed, or moved. The CRM does not know, so the forecast does not either.

The missing-field problem concentrates where the volume is: in the largest connected CRM, three deals in ten had no close date at all, and one in five had no amount. The smaller CRMs kept those fields largely filled, which fits the pattern: hygiene survives at low volume and loses at scale. A deal without a date and a value is not a forecast entry. It is a bookmark.

Stack these and the working pipeline, the part a leader can actually reason about, is a minority of what the view shows. The rest is history that nobody archived.

What the contacts say

The same CRMs held close to 250,000 contacts. Not one CRM had phone numbers for even half of them. In the median CRM, roughly nine of ten contacts had no phone number. Emails varied more: some CRMs had them for nearly everyone, others for almost no one, with the median CRM missing them on about one contact in seven.

A contact you cannot call or email is a name. Half the average CRM is names.

Why pipeline hygiene programs do not fix it

The standard response to bad CRM data quality is process: mandatory fields, pipeline hygiene sessions, a Friday reminder. The data above comes from companies that have those. The upkeep still loses, because the design pits a rep’s selling time against the system’s need to be told things, every day, forever. The system loses that trade a hundred small times a week, and the misses compound quietly until the pipeline view is a work of fiction with confident formatting.

That is the actual case for agents in sales, and it has nothing to do with writing emails faster. A system that hears the call, reads the thread and updates the record does not need the rep to remember. The close date moves when the buyer says next quarter. The deal closes when it closes. The pipeline reads true because nothing depends on a human finding time to type.

If you are evaluating tools on this problem, judge them on one question: after a customer conversation, what happens to the record if the rep does nothing? For most of the stack the answer is nothing happens. We compared the categories in the AI SDR guide and the CRM comparisons.

Methodology

Figures are aggregated across the B2B organizations whose CRMs (Upsales, Salesforce, HubSpot and Attio) are connected to Stalar, measured in August 2026. No organization is identifiable and none is weighted; per-CRM framing (median, “the two largest”, “not one”) is used throughout because pooled numbers would be dominated by the largest connected CRMs. Open deals are those whose stage is not won, lost or closed; deal-level percentages are reported only for CRMs with meaningful deal volume.