Cold email statistical uncertainty: report a range around a small reply count

Define the decision

Three positive respondents among one hundred contacted people gives a point estimate of three percent. That exact arithmetic can look more precise than the underlying evidence. For a simple binomial model, a Wilson interval is one way to show uncertainty around the proportion; the model assumptions still matter.

Work through the procedure

1. Define a binary outcome and count each eligible independent unit once.
2. Calculate the observed proportion using the stated denominator.
3. Apply an appropriate interval method and report its confidence level and assumptions.
4. Discuss selection, clustering and measurement errors separately because an interval does not automatically include them.

Worked example

The following is a synthetic example for this procedure, not a customer result or performance benchmark.

Positive respondents: 3
Independent contacted units assumed: 100
Observed proportion: 3%
Approximate 95% Wilson interval: 1.03% to 8.45%
The range is wide. It is not evidence that the campaign's true rate is exactly 3% or that another campaign at 4% is better.

Read the result

Under the model, repeated use of a 95% confidence-interval procedure has a long-run coverage interpretation. It is not a 95% probability statement about this fixed interval containing a randomly moving true rate. Multiple people within one company or changing message treatments can undermine the simple independent-unit assumption.

Check before moving on

1. Show numerator and denominator next to the interval.
2. Use a method appropriate for zero outcomes rather than declaring zero risk or zero possible response.
3. Do not infer a significant difference by casually comparing two point estimates.

Limits and next action

NIST documents proportion interval methods; the worked arithmetic here is illustrative. Zintara's positive reply calculator does not currently compute these intervals. Obtain statistical review when the experiment has clustering, repeated monitoring or material commercial consequences.

Source: NIST: proportion confidence interval methods