Become an affiliateStart your program

Measurement and metrics

Minimum data threshold

Also called minimum threshold, data threshold, statistical minimum, minimum sample size, bagatelgrænse.

Short answer

A minimum data threshold is the smallest amount of data a metric must rest on before it is shown or used for a decision. Below the threshold the figure is noise: 1 sale from 3 clicks is not a 33 % conversion rate.

Definition of Minimum data threshold.

A percentage calculated on few events swings wildly. With 10 clicks, one sale more or less moves the conversion rate by 10 percentage points. With 1,000 clicks it moves it by 0.1 percentage points. The minimum data threshold is the rule that says when the figure is stable enough to be read as a property of the program rather than as chance.

The threshold is normally set on two quantities at once: a minimum for the denominator (number of clicks, impressions or visits) and a minimum for the events (number of sales or leads). Both are needed. 500 clicks and 0 sales still give no usable conversion rate, and 3 sales from 4 clicks give a figure of 75 % that nobody should act on.

The rarer the event, the larger the base it requires. A program with a true conversion rate of around 1 % needs several hundred clicks before the measured figure settles reasonably. A lead program with 20 % conversion is stable far earlier. That is why there is no universal number – there is a threshold per metric and per order of magnitude.

A documented zero is not transparency

A new program with 4 clicks and 0 sales has no conversion rate. If the figure is shown as 0 % anyway, visitors read it as a verdict on the program, even though it is purely a verdict on the amount of data. The result is that new programs are punished for being new.

The honest display is therefore not a zero but a note that the data is still too thin. It is the same principle as in an A/B test, where a result below the sample size requirement is reported as inconclusive rather than as a winner.

How it goes wrong

  • 01

    The threshold is set only on the events

    Require 3 sales but no minimum of clicks, and 3 sales from 5 clicks slip through and are shown as 60 %. Always set a threshold on both the denominator and the numerator.

  • 02

    The threshold is moved after the figure is seen

    If the requirement is lowered because a program is just two clicks short of qualifying, the threshold has stopped being a rule. Decide the threshold before the data, not after.

  • 03

    The threshold is mistaken for quality

    A program below the minimum data threshold says nothing about how good it is. It only says that nobody knows yet. Read the absence of figures as unknown, not as bad.

At Affilyflow

Minimum data threshold in practice

Public metrics on Affilyflow are only shown above a minimum data threshold of 25 clicks and 3 sales. A program with 4 clicks and 0 sales gets no conversion rate on its public page – not because the figure is hidden, but because it does not exist yet.

The threshold applies to the public surface, not to the store's own overview of its own figures. The purpose is that a visitor is never presented with a percentage calculated on so little data that it would have looked completely different yesterday.

Frequently asked questions

What is a minimum data threshold in statistics?

A minimum data threshold is a minimum requirement for the amount of data before a calculated figure is shown or used. It exists because percentages calculated on few events say more about chance than about what they are meant to measure. Below the threshold, unknown is reported rather than a figure.

How many clicks do you need before a conversion rate is usable?

It depends on how high the rate is. At a rate of around 1 %, several hundred clicks are needed before the figure settles, while a rate of 20 % is stable far earlier. A practical rule is at least 3 conversions and a click count that matches the expected rate.

Why do some affiliate programs not show their metrics?

Because the program does not yet have enough clicks and sales for the figures to mean anything. Showing a conversion rate calculated on 4 clicks would give visitors a false impression in either direction. The figures appear once the base is large enough.

Is a minimum data threshold the same as a payout threshold?

No. A minimum data threshold is about measurement: when a metric is stable enough to be shown. A payout threshold is about money: how large a balance must be before it can be paid out. The two have nothing to do with each other beyond the word threshold.

Read on