How to identify the biggest leak in a COD funnel

Why is comparing stages to each other misleading?

Because different stages have naturally different achievable ceilings built into their very nature, and comparing raw percentages across them side by side ignores that reality completely. Lead-to-order conversion is often lower than order-to-confirmation conversion simply because a lead is a much lower-commitment signal than someone who has already gone ahead and placed a real order with intent behind it. Seeing a 55% lead-to-order rate sitting next to a 90% confirmed-to-shipped rate does not mean the first is "the problem" and the second is fine as it currently stands; it might mean the first is already close to what is realistically achievable for that particular stage of the funnel, while the second, despite looking impressively high on paper, still has real room to improve that nobody has bothered looking for yet.

What should each stage actually be compared against?

Its own recent history over time, or a documented benchmark for that specific transition based on past performance, not a neighboring stage sitting elsewhere in the same funnel. If your own confirmed-to-shipped rate was 96% two months ago and sits at 90% now, that is a real leak worth investigating properly even though 90% still sounds acceptable in isolation without any historical context attached to it. If lead-to-order conversion has held steady at 55% for six months despite repeated changes to messaging and follow-up timing, that may simply be close to the natural ceiling for how leads actually behave at that stage of the funnel, and effort spent pushing it further may cost considerably more than it ever returns in additional orders.

How do you calculate a stage-specific leak score?

In this example, order-to-confirmed shows the largest gap from its own baseline, at negative 6 points, even though its raw percentage of 58% is not actually the lowest figure sitting anywhere in the table above it. That gap, not the raw number sitting on its own, is what actually points at where something changed for the worse recently, rather than at where a stage simply happens to have a naturally lower ceiling to begin with.

What do you do once a stage is identified as the leak?

Ask what changed around the time the gap first opened up, since gaps of this kind rarely appear without some identifiable cause behind them. A new confirmation message template rolled out that particular week, a change in who is handling replies day to day, a shift in lead source that brought in noticeably lower-intent traffic than the store was used to seeing before. The gap analysis tells you which stage deserves closer investigation; it does not hand you the cause on its own without further digging into what actually happened. Cross-reference the timing of the gap against anything that changed operationally in that same window, since the cause is almost always something that actually happened on a specific day, not something that was quietly true all along without anyone noticing.

How often should this comparison be run?

Weekly is tight enough to catch a gap close to when it first opened, and loose enough to smooth out ordinary day-to-day noise in smaller order volumes where a handful of orders can swing a daily percentage wildly in either direction. Keep a running log of each stage's rate over time rather than comparing only the current week to the previous one in isolation from the broader trend, since a slow multi-week drift downward is easy to miss entirely if each individual week-over-week change looks small and unremarkable when viewed on its own.