Find the Step Where the Funnel Leaks

15 July 2026 · 2 min read

blog contents
  1. Percentages, not counts
  2. The worst step is rarely the last one
  3. Ask why the survivors survived
  4. Fix one step, then measure again

read

Before buying more traffic, find out where the traffic you already have is going. Measure the share of people surviving each step rather than the totals, and one step will usually stand out as far worse than the rest. Fixing that step multiplies everything upstream of it. Adding traffic to a funnel with a bad step buys more of the same loss at a higher price.

Percentages, not counts

Absolute numbers always shrink down a funnel, so they tell you nothing. The share surviving each step tells you where the loss is unusual.

Write the steps down and put a percentage next to each one. The bad step is usually obvious the first time anyone does this, which is why it is worth doing before anything else.

The worst step is rarely the last one

Attention lands on checkout because that is where money is visible. The larger loss is usually earlier, at the point where a visitor decides this is not for them.

That decision often happens before anything is clicked, which is why the step will not appear as an event unless you look for it deliberately.

Ask why the survivors survived

Numbers show where people leave. They do not say why. The fastest route to why is asking a handful of recent customers what nearly stopped them.

Five conversations will explain a number that months of dashboards will not, and they cost an afternoon.

  • Which step has the worst survival rate
  • Whether it is worse on phones than on desktops
  • Whether it is worse for one traffic source
  • What recent customers say nearly stopped them

Fix one step, then measure again

Changing several steps at once produces a number that moved for reasons nobody can identify, which teaches you nothing for next time.

One change, one cycle, one measurement. Slower in the moment and much faster over a quarter.

questions

How many steps should a funnel have?

Few enough that each one is a real decision by the visitor. If a step exists only because the software has a page for it, it is a leak with no purpose.

What if traffic is too low to measure?

Then talk to people instead of counting them. Under a few hundred visits per step, conversations produce better decisions than percentages built on noise.

Is mobile always worse?

Usually somewhat worse, and the gap is the useful part. A gap far larger than the norm points at a specific problem on small screens rather than at buyer behaviour.

More on this track

  1. 5 Growth Hacks That Actually Work Most growth hacks are a story about somebody else's business. These five survive contact with a spreadsheet: fix the worst funnel step before buying more traffic, sell to the people who already bought, put a price on a lead before spending, test channels on a schedule, and make the first ninety seconds of the product work. 29 August 2026 · 3 min read
  2. Cheap Leads Are the Expensive Ones Optimising for cost per lead produces more leads and fewer customers. The metric is easy to move, so it moves: broaden the targeting, soften the offer, add a giveaway, and the number falls while the sales team quietly drowns. A lead is not a result, it is a claim on someone's attention. Measure the step where money changes hands, and let the earlier numbers land wherever they land. 19 August 2026 · 3 min read
  3. Why Platform Reports Never Match Every advertising platform reports on its own conversions, using its own window and its own rules for who deserves credit. Add the reports together and the total will exceed what your bank actually received. This is not fraud and it is not a bug, it is each platform answering a question about itself. The fix is not a better report, it is one internal number that everything else gets compared against. 1 August 2026 · 3 min read