Find the one bottleneck and fix only that

Once the machine is running, the instinct is to improve all of it at once — better copy, faster response, a slicker booking flow, a tighter sequence, all in the same week. Resist it. A solo founder optimising the whole funnel at once is a solo founder making many small changes, measuring none of them cleanly, and learning nothing. The discipline that compounds is the opposite: find the single worst stage, fix only that, re-measure, repeat. The general method is find your one growth bottleneck; the idea of a single binding bottleneck is what makes the whole loop work.

Instrument every stage

You cannot fix what you have not measured, so the first move is to instrument the funnel stage by stage. A lead is captured; some fraction become an MQL; some of those become an SQL; some become real opportunities; some close. Each transition has a conversion rate, and writing those rates down — even roughly — turns a vague sense that 'sales is slow' into a specific picture of where deals fall out. Most founders carry a story in their head about where the problem is, and most of the time the story is wrong. The instrument tells you the truth. The funnel metrics that actually move revenue is the shortlist worth tracking.

Compare to a benchmark, then attack the worst stage

With the stages instrumented, the question becomes which one is furthest from where it should be. Compare each stage to a reasonable benchmark for your motion — and where you have no clean external number, compare it to itself over time and to what the stage plausibly could be. The point of the comparison is not the precise figure; it is to surface the one stage dragging hardest. Then fix only that stage. If qualified leads are not turning into opportunities, the problem is not your capture page and you should not touch it this cycle. Attack the worst stage, leave the rest alone, and you get a clean read on whether the fix worked. Watching sales velocity and keeping enough pipeline coverage tells you whether the whole machine is healthy while you work one stage at a time.

The weekly rhythm

Make this a rhythm rather than a project. A fifteen-minute pipeline review every week — ideally with an AI assist that pulls the stage conversions and flags the worst one — keeps the machine improving without eating your time. Recalculate the stages, identify this week's single bottleneck, make one change, and stop. Next week you measure whether it moved and pick the next one. Run that loop for a quarter and you will have attacked twelve bottlenecks deliberately and measured every one, which beats a year of changing everything at once and never knowing what helped.