Sensitivity — which assumption decides it?
About this article
Author: Dmitry Dugarev
Every number in a funnel model is a guess with a confidence level attached — a conversion rate, a price, a cost per click. Not all guesses matter equally: moving some inputs by 20% barely shifts the outcome, while moving others by the same 20% swings the whole model.
FunnelCAD moves each input to −20% and to +20% in turn and calculates the expected profit a month each time — exactly, the same number you'd see if you typed the changed value into the field. It then ranks the inputs by how much they move the expected profit. Here they are for an example funnel with an upsell: €1,100 a month of Google Ads at €2.50 a click, 5% of clicks book a free intro call, 5% of the calls buy a €900 workshop, and 30% of workshop clients add a €4,000 implementation project. With no payment fees, its expected profit is +€1,210 a month.
| Input | At −20% | At +20% |
|---|---|---|
| Cost per click (€2.50) | +€578 | −€385 |
| Clicks → intro call (5%) | −€462 | +€462 |
| Intro call → workshop (5%) | −€462 | +€462 |
| Implementation price (€4,000) | −€264 | +€264 |
| Implementation take-up (30%) | −€264 | +€264 |
| Ad budget (€1,100) | −€242 | +€242 |
| Workshop price (€900) | −€198 | +€198 |
The cost per click is lopsided: 20% cheaper clicks buy a quarter more of them for the same budget, while 20% dearer clicks buy a sixth fewer.
The homepage's example funnel has a single offer, so there both rates and the price each move profit by the same €440 either way. The ranking gets interesting once a funnel has more than one offer.
The input at the top is the assumption most worth testing first — it's where a wrong guess costs the most, and where a small improvement in the real number pays off the most.
This is different from asking "which number am I least sure about?" — an assumption you're very unsure of but that barely moves the outcome isn't worth chasing yet. Sensitivity tells you where uncertainty actually costs you money.
Hours aren't in the list: they load your week, not your profit — see capacity. Neither are the profit target, the capacity and the test length, which change no profit.
To see the list, select a scorecard: the property sheet shows it under What moves profit. Click a row to jump to that input.
About the author