The mental model: a pipeline is a manufacturing line, not a list

Most founders treat their pipeline as a to-do list of named deals they hope will close. That framing is why pipelines stall. The better model is a factory line: leads enter at one end, move through fixed stations, and revenue comes out the other. The only thing that matters is how fast and how reliably the line moves units through. This is throughput thinking, and it changes every decision you make.

The number that captures it is pipeline velocity: (open opportunities × average deal value × win rate) ÷ sales cycle length. Look at the four levers. You can add more deals, raise your average deal value, lift your win rate, or shorten your cycle. Three of those four are improvements to a station on the line, not to your effort or hustle. AI is good at exactly this — taking friction out of a station so the same input produces more output. An agent that drafts a follow-up the moment a call ends shortens the cycle. An agent that flags a stalling deal lifts the win rate by catching it before it dies. An agent that auto-logs activity keeps the data clean enough that your win-rate number is real.

The second half of the model is pipeline coverage. If your target this quarter is £100k and you need a 3x floor to hit it reliably, you need £300k of qualified pipeline in the line — with 4x to 5x the safer target for most B2B. Coverage tells you whether the front of the line is fed; velocity tells you how fast the line runs. Get both honest and your forecast stops being a hope and becomes a calculation.

A solo founder who internalised this reframe — a two-person SaaS team in Edinburgh targeting HR directors — stopped chasing new logo conversations and spent six weeks fixing their SQL-to-opportunity conversion from 28% to 44%. Revenue grew 31% that quarter without a single new lead entering the top. The line was already fed; the station was broken.