Aller au contenu
Peak Atlas

Peak Chat · Experiments · planned

Two ways to talk.
One fair comparison.

Planned for Peak Chat: experiments that compare two versions of a conversation fairly, against a metric fixed in advance. Not available today.

Which question must the source answer?

Your next review briefFinding

Both variants are counted on the same metric, over the same period and with comparable visitors.

You can take the brief with you at the end of the page. The original evidence stays open.
See the planned test setup
Statut
Planned, not built
Compares
Two versions of one conversation
Today
No tests, no results
Experiments · planned
Question → method → finding

Set the rule first.
Then compare.

The sculpture is an illustration, not study evidence.
The starting point

A short hello, or straight to the question?

Whether a chat works better when it opens by asking about the visitor's project or with a brief greeting is not something to guess. The plan is to run two versions of a conversation side by side and compare them on a metric chosen beforehand, such as confirmed meetings. Experiments are meant to build on Peak Experiments. They do not exist in Peak Chat yet.

Context → question → next step

Finding & data slice

What an experiment should answer.

No experiment and no results. The page describes how a fair comparison is planned to be set up.

Observation

Interpretation

Decision
Figure 01

The evidence is still pending

No measurements. Without a real test the figure shows no variants and no differences.

Chart and alternative data table share the same checked source.

Transparency is part of the result

How conversation experiments are planned to work.

An experiment is only as honest as the rule you fix before it starts.

01 / Source briefMethod before measurement

Hypothesis

Written down in one sentence before the start

Variants
Planned: two versions of a conversation, otherwise equal
Metric
Fixed in advance, for example confirmed meetings
02 / Split

What counts as an
observation?

Planned: visitors assigned to a variant at random

Event defined reference groupMeasurement scheme · no calculated study value

03 / LimitsA result holds for your site and your time window

How data becomes orientation

Hypothesis. Result. Decision.

  1. What showed up?

    Both variants are counted on the same metric, over the same period and with comparable visitors.

    Key idea

    Finding

  2. Is the gap real?

    A small gap across few conversations can be chance. The plan is for Peak to say so openly instead of naming a winner.

    Key idea

    Interpretation

  3. What do you keep?

    The decision stays with you: adopt one variant, keep testing or drop both. The result is meant to be kept as a learning.

    Key idea

    Decision

Your quality check

Is your test set up fairly?

Tick what you have checked. Your selection stays in this view only.

00/ 03your checkpoints

0 of 3 points checked

Is your test set up fairly?
Sources & context

What this article is based on.

Planned feature, not available. No test results, no sample and no claims about better variants. Peak Experiments exists as its own app; the link to Peak Chat is planned.

  • Editorial demo

    Editorial example with made-up sample data. The Peak Chat inbox and Peak Helpdesk's guest live chat exist today; everything else is marked as planned or example.

  • Evidence pending

    No dataset and no study. The page describes a method, not measured values.

The next step

Compare without fooling yourself.

Experiments are planned. The live demo shows a simulated conversation that could be tested later.

See the live demo

Frequently asked questions

Can I test conversation variants today?

No. Experiments in Peak Chat are planned and not available yet.

Will Peak pick a winner automatically?

That is not the plan. The decision is meant to stay with you, and with too few conversations Peak is meant to say so.

What will variants be measured on?

On a metric you fix before the start, such as confirmed meetings.

More in Peak Chat