The Pipeline as a System

A pipeline is a queue

Strip away the sales vocabulary and a pipeline is a queueing system. Opportunities arrive, spend time in a sequence of stages, and leave as won, lost or abandoned. Queueing systems obey a relationship proved by John Little in 1961 and revisited in his 2011 review: the average number of items in a system equals the average arrival rate multiplied by the average time each item spends in the system. The result holds under very broad conditions, regardless of the order in which items are served.

Applied to a pipeline, Little's Law says that the number of open deals you carry is the rate at which you create them times how long they stay open. That has a consequence most pipeline reviews miss. Doubling the number of open deals does not mean you will close twice as many; if the creation rate did not change, it means deals are staying twice as long. A swollen pipeline is very often a slow one, full of opportunities that should have been closed out as lost months ago.

The practical reading is that the healthy pipeline is not the biggest one. It is the one whose stages describe real buyer progress and whose deals leave, one way or the other, at a steady rate.

Exit criteria: stages the buyer has to earn

A stage is useful only if two people looking at the same deal would agree on whether it belongs there. That requires exit criteria: a short, observable statement of what must be true before a deal moves to the next stage, written in terms of what the buyer has done rather than what the seller has done.

Compare two definitions of a 'Qualified' stage. 'Had a good first call' describes the seller's impression. 'The buyer has confirmed a problem we solve, named who else is involved in the decision, and agreed to a second meeting with at least one of them' describes buyer actions anyone can check against the timeline. The second definition is harder to satisfy, which is the point.

A workable starting set for many B2B teams has five open stages, each with one or two exit criteria: Engaged (the buyer replied or met with you), Qualified (problem, stakeholders and a next meeting confirmed), Evaluating (the buyer is actively comparing, and you know their criteria), Proposal (the buyer has asked for and received pricing and terms), and Committed (the buyer has said yes and the paperwork is in motion), followed by Won and Lost. Adjust the names to your market, but keep each criterion observable.

The evidence for formality here is directional rather than precise. Jason Jordan and Robert Kelly reported in Harvard Business Review in 2015 that companies with a formal sales process generated more revenue. More recently, Salesforce's 2026 State of Sales survey (vendor data) found that 79% of high performers prioritize data hygiene, against 54% of underperformers. Neither proves that exit criteria cause growth, but both point the same way: teams that can trust their pipeline data do better than teams that cannot.

Three ways a pipeline lies

Stage inflation. Deals move forward because the seller is optimistic or because a manager asked why nothing moved. The tell is a stage whose conversion to the next stage keeps falling while its deal count rises. The cure is exit criteria, enforced in review: 'what did the buyer do that put this in Proposal?'

Zombie deals. Opportunities that have had no buyer activity for months but were never closed out, because closing them as lost feels like admitting failure. They inflate the total, depress every conversion rate, and by Little's Law make your average cycle time look longer than it is. The cure is a silence window per stage, after which a deal must be either re-engaged or closed.

Missing dates. A deal with no expected close date cannot be forecast and cannot be late, which is exactly why sellers leave the field empty. Every open deal past the earliest stage should carry a close date the seller believes, and moving that date should be a visible event rather than a quiet edit.

None of these are character flaws. They are what happens to any queue that nobody is responsible for keeping clean. Forecasting surveys are the downstream symptom: in Xactly's 2024 benchmark (a vendor survey of 405 North American sales and finance leaders), four in five said they had missed a quarterly sales forecast in the past year.

Configuring the system in the Deal Room

Kavanah's Deal Room gives each pipeline an ordered list of stages, and each stage carries four settings that map directly onto this lesson. The stage's kind is open, won or lost; won and lost stages mark deals closed and hide them from the default board, so a mid-funnel stage whose name sounds final must still be open. The default probability, from 0 to 100, is the baseline the weighted pipeline summary and the forecast use. The stale-after window is the number of days of silence after which a deal in that stage appears in the follow-up brief and in list_stale_deals. And a colour, which matters more than it sounds when you scan a board.

Stages are managed from the board's Manage stages dialog, which is admin-only because stage configuration shapes every member's board. The agent can also create and update stages with create_pipeline_stage and update_pipeline_stage. Deleting a stage is deliberately only available in the dialog.

Kavanah does not store exit criteria as a field on a stage. Write them down where the team will read them, a short Knowledge document titled 'Pipeline stage definitions', and link it from your review agenda. Then make stage changes count: a deal-stage-changed automation trigger can fire on every move into a given stage, which is how you attach a checklist task or a notification to the moment a deal claims to have met a stage's criteria.