> ## Documentation Index
> Fetch the complete documentation index at: https://docs.abtestly.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Time to decision

> An honest estimate of how much longer this test might need. Not a promise, not a forecast, a rough runway based on current traffic.

The results page shows a small **time to decision** hint under the
verdict:

> Roughly 4 more days at current traffic to reach a call.

Two useful things it does:

* Answers "how long do I need to wait" without having to think.
* Sets expectations before someone else asks.

## What the estimate does

Simple projection:

1. Look at the current confidence-interval width for the primary
   goal.
2. Estimate the sample size needed to shrink the CI to the width that
   would clear the [confidence gate](/results/reading-a-result).
3. Divide by the current visitors-per-day rate.

That is it. No forecasting model.

## What it is not

* **A guarantee.** Traffic can change, effect size can regress to the
  mean.
* **A green light to stop early.** The verdict copy is what says the
  test is done. A "3 days to decision" hint means "keep going", not
  "you can call it in 3 days regardless of the verdict".
* **Always available.** Very early tests or tests where the effect is
  tiny show "cannot estimate yet".

## When it says "cannot estimate yet"

* **Not enough data yet.** Wait a day.
* **The effect is so small the CI is not shrinking.** A near-zero
  true effect can leave the CI wide forever.
* **Traffic is very uneven.** If most of your visitors came on one
  day and traffic dropped since, dividing by average visitors-per-day
  misleads. The estimator is careful and refuses a bad guess.

## How to use it

* **Under a week?** Almost always keep going.
* **A month or more?** Consider whether the test is worth waiting for.
  Sometimes the honest answer is "the effect is too small to be worth
  detecting", and the business call is to move on.
* **"Cannot estimate yet"?** Keep going. Come back in a day.

## If you are impatient

The sequential engine (see [the engines page](/results/engines)) is
designed for peeking. Switch to it on the next test, you can look as
often as you like without inflating your false-positive rate.