GTM experiment brief
Run a small market test with a hypothesis, quality guardrails, and a decision rule.
Use when debating a new audience, message, channel, or offer without enough evidence for a full rollout.
Before the working session
- Choose one variable to change. Capture the current baseline and how reliable it is.
- Specify the eligible audience and exclusions before picking a channel or creative treatment.
Run the process
Write a falsifiable hypothesis
Use the form: for this audience and trigger, this offer through this channel will produce this observable buyer behavior because of this mechanism.
Design the measurement
Choose one primary result and at least one quality guardrail. Define denominator, attribution window, and data source. Measure effort or cost if scale would change economics.
Bound exposure
Set sample, calendar period, budget, owner, and any holdout or comparison. Check that the test can reach enough eligible people to inform a decision.
Record the decision before launch
Define what would justify scale, what requires a message or channel revision, and what stops the test. Review the data quality before interpreting the result.
Fictional example: A team sends one workflow-specific offer to 60 eligible accounts over four weeks. It measures qualified conversations and disqualifications, with a cap on research time per account. If only meeting volume rises, the offer does not scale.
Failure checks
- Changing audience, offer, and channel at once makes learning weak.
- A click is not a qualified opportunity.
- A short test can miss a long buying cycle; choose a leading signal deliberately.
Editable workbook and filled example
Precommit a test design and decision rules, then record outcomes without silently changing thresholds after the result.
Download editable Word workbook ↓
What you will complete
- Decision informed and accountable owner
- Eligible audience and exclusions
- Variable changed and comparison
- Primary outcome definition
- Exposure unit and attribution window
- Duration budget and review date
Record sample adequacy, instrumentation and exclusions before interpreting a rule. A minimum sample is a planning rule, not a power calculation. A before-after change is not automatically a causal treatment effect.
Fictional filled evidence record
| Measure and definition | Source and counting rule | Threshold set before test |
|---|---|---|
| Qualified follow-up request per eligible account | Deduplicated account log; fixed response window | Fictional scale at 5%; stop at 2% |
Fictional action record
| Sample and outcomes | Cost and quality | Rule result | Data limitation and next action |
|---|---|---|---|
| 240 eligible accounts; 4 outcomes | Spend $240; $60 per outcome | 1.67% rate; stop candidate | Check eligibility before a new offer test |
Decision and next review
Stop the fictional variant: 4 divided by 240 is 1.67%, below the precommitted 2% stop threshold. The $60 cost per outcome passes the $80 cap, but does not rescue the weak rate. These thresholds are teaching values, not benchmarks.
Test owner records missing data and offer changes, investigates audience fit, and proposes a new test without retroactively moving the original threshold.
The Word file includes editable response fields and blank tables. These examples are invented and do not represent a real buyer or employer.
Ready to track the result? Open the experiment tracker →