In short
One campaign spent €60 in a week and brought five leads, so a lead costs €12. Another spent €3,600 and brought 300 leads. Also €12. In a plain table these are two equal rows.
But you cannot trust them equally. If the first campaign had brought one more lead, a lead would cost €10. One fewer, €15. With five leads, the real price is known only roughly: somewhere between €5 and €37. In the second campaign, one lead more or less hardly moves the price: it lies between about €11 and €13.50. Cutting the first campaign’s budget because of its number means deciding on chance.
One formula
CPL — cost per lead
Spend ÷ Leads
the spend is known exactly, only the number of leads is chance
What the cell shows
Three kinds of marks
KO shows whether you can trust a number where the number is: in the table cell and on the metric card.
- Enough data. The number looks as usual, with no mark.
- Little data. A dot next to the number. Its tooltip explains in words: “Rough number: too little data, only 5. The real value is most likely between €5 and €37”.
- No events. A tilde before the number, in a muted colour. This is not a measured value but a limit: a cost cannot be lower, a conversion rate is almost certainly not higher.
The marks appear in the Plan vs actual, Breakdown, Metrics and Entities sections. You do not need to turn them on: they are shown by default.
No leads: a limit instead of a dash
A campaign spent €60 and brought no leads. The cost per lead cannot be calculated, because there is nothing to divide by. Many reports show a dash here. But a dash looks the same as a row where no money was spent at all, and the campaign that used up its budget gets lost among the empty rows.
KO shows ~€60 with the tooltip “It will not get cheaper: €60 spent over the period with no events”. If even one lead had come, it would have cost the full €60. Since none came, a lead costs at least that much. The table sorts by this number, so the campaign stays near the top, with the expensive ones.
Conversion rates work in a similar way. A hundred visits and no leads: KO does not show “0%”. Zero leads out of a hundred visits can happen even when the real conversion rate is about 3%, just by bad luck. So the cell shows ~3.7%: the conversion rate is almost certainly not higher than that. With ten visits and no leads, the limit is already 28%: we know almost nothing about the conversion rate.
How KO works out the range
The range is the “from–to” within which the real value lies with 95% probability. KO works it out differently for conversion rates and for costs, because the chance is in a different place.
A conversion rate is the share of visits that became leads, or of leads that became sales. What is down to chance here is who exactly left a lead: out of a hundred visits it could have been two people or five. KO works out the range of a conversion rate with the Wilson method. It also works at the edges: with zero leads it gives a sensible upper limit, while the usual “plus or minus” formula gives a meaningless zero.
A cost is a cost per lead or per sale, and also the average order value. The money here is known exactly: it has already been spent or received. Only the number of events is down to chance: under the same conditions, three leads could have come, or ten. KO works out the largest and the smallest number of leads that could have come and divides the spend by each of them. This gives the low and the high end of the price.
Money to money means ROAS, the share of spend in revenue, the margin. These formulas have only sums of money and no events, so there is no chance in the number of events. They never get a mark, and a zero in the denominator gives a dash.
KO adds the dot when the range is too wide compared with the number itself. The threshold is the same for all projects and is not in the settings: a marketer has nothing to base the choice on.
Which metrics get no mark
A metric gets a mark when it is calculated with one division: spend divided by leads, leads divided by visits, in percent too. ROMI also has a subtraction in its formula, and combined metrics add up several parts. They have no single “divided by”, so there is no honest way to work out the range. An invented range is worse than none, because it looks like knowledge. So such numbers have no mark, even when there is little data behind them.
If a metric is built from other metrics, their uncertainty carries over to the result. For example, the ratio of LTV to acquisition cost gets a dot when there is little data for LTV or few sales.
Three display modes
In a section, the Analysis menu has a Confidence item with three modes:
- Do not show: no marks, and a dash again in place of the tilde. Useful for screenshots and for people who find the marks distracting.
- Mark rough numbers: the tilde and the dot, with the range in the tooltip. This is the default.
- Show range: the cell shows the range in place of the number, for example €5–37. On the Plan vs actual chart, a pale band runs along the line. Where there is enough data, the band stays close to the line. Where there is little, it widens, and you can see that a “drop” fits inside what chance alone could give.
The range is there to compare. In one campaign a lead costs €12 (five leads), in another €20 (seven leads). By the numbers, the first is almost half the price. In range mode, “€5–37” and “€10–50” stand side by side: the ranges almost match, and this data shows no real difference between the campaigns. With tooltips you cannot compare this: you have to hover over each one in turn and keep both numbers in your head.
What else a mark changes
- Highlighting. A number with a tilde is left out of the colour scale. It is a limit, not a measurement, and colouring it as a record would be wrong.
- Export. A number with a tilde goes to the file as an empty cell, and the sheet notes say how many such cells there were. In someone else’s spreadsheet, “60” without the tilde and the tooltip would look like a fact.
- Sorting. A number with a tilde sorts by its own value, like the other numbers.
Summary
- KO works out whether you can trust a number and shows it in the cell: a dot means little data, a tilde means there were no events and the number is a limit.
- €60 spent with no leads shows “~€60”, not a dash. A hundred visits with no leads shows “~3.7%”, not 0%.
- The range is at 95% probability. Conversion rates use the Wilson method, costs use how many events could have come. ROAS, ROMI and combined metrics get no mark.
- “Show range” puts the range in place of the number and a band on the chart, so you can compare two uncertain numbers by eye.
- Numbers with a tilde are not coloured by highlighting and do not go to an export as numbers.
Updated