Use this marketing experiment tracker to run small, interpretable tests and preserve what the business learns about audiences, messages, offers, channels, and conversion steps.
The tracker is suitable for organic, email, partner, referral, outbound, landing-page, and paid tests. Use the marketing attribution guide to choose a proportionate attribution rule.
A Useful Marketing Experiment
Define one uncertainty, a falsifiable hypothesis, intended audience, controlled change, baseline, exposure, budget, timebox, primary measure, guardrail, and decision rule before launch.
Experiment vs Campaign
A campaign aims to produce a result. An experiment aims to learn whether a specific change contributed to a result. A campaign may contain experiments, but routine promotional activity should not be mislabeled as causal proof.
Interpret Small Samples Carefully
Solopreneurs often lack enough volume for strong statistical conclusions. Record counts as well as percentages, avoid false precision, repeat important findings, and use qualitative evidence to explain—not overwrite—the observed behavior.
Copy the Marketing Experiment Tracker
Create one record per experiment. Keep the hypothesis and thresholds unchanged during the test; document any deviation instead of silently rewriting the plan.
Experiment identity
Experiment ID and name: [Stable identifier and descriptive name]
Owner: [Accountable person]
Start and end dates: [Fixed testing period]
Decision date: [When results will be interpreted]
Status: [Planned, running, paused, complete, or invalid]
1. Decision and hypothesis
Decision this informs: [What will change if the result supports or weakens the hypothesis]
Uncertainty: [Single important unknown]
Hypothesis: [For audience X, changing Y from A to B will change measure Z because reason R]
What would weaken it: [Observable contradictory result]
What this test cannot establish: [Causal or commercial limits]
2. Audience and eligibility
Intended audience: [Customer or account criteria]
Trigger or context: [Situation in which the message appears]
Inclusion rule: [Who counts in the test]
Exclusion rule: [Bots, employees, existing customers, duplicate contacts, or irrelevant traffic]
Recruitment or traffic source: [Where exposure comes from]
3. Variable and control
Independent variable: [One element intentionally changed]
Control or comparison: [Current version, holdout, prior baseline, or matched comparison]
Constants: [Offer, price, audience, timing, page, or other elements held stable]
Contamination risk: [Other changes that could affect the result]
Implementation check: [How correct exposure is verified]
4. Baseline and measures
Baseline period: [Dates and comparable conditions]
Baseline counts: [Numerator and denominator]
Primary measure: [Single measure tied to the hypothesis]
Secondary measures: [Diagnostic measures only]
Guardrail: [Unsubscribe, complaint, refund, support, margin, or brand-risk limit]
Data source: [Authoritative analytics, CRM, finance, or platform record]
5. Budget, exposure, and stop rule
Cash limit: [Maximum spend and currency]
Owner-time limit: [Maximum hours]
Planned exposure: [Qualified sends, visits, conversations, or accounts]
Minimum test period: [Time needed to include relevant behavior]
Early stop condition: [Ethical, legal, financial, technical, or customer-harm threshold]
6. Decision thresholds
Continue threshold: [Result supporting continued use or larger test]
Revise condition: [Result suggesting the idea but not the current execution]
Stop condition: [Result making further investment unjustified for now]
Guardrail threshold: [Maximum acceptable downside]
Pre-registered date: [When these rules were finalized]
7. Results
Actual exposure: [Count after exclusions]
Primary result: [Numerator, denominator, percentage or amount, and period]
Secondary results: [Counts and context]
Guardrail result: [Observed downside]
Cost and owner time: [Actual amounts]
Data-quality issue: [Missing events, tracking changes, sample imbalance, or contamination]
8. Interpretation and action
Decision: [Continue, repeat, revise, stop, or invalid]
Evidence supporting the interpretation: [Observed result]
Alternative explanation: [Seasonality, source mix, novelty, timing, selection, or implementation]
Commercial effect: [Revenue, gross contribution, acquisition cost, payback, or qualified pipeline where measurable]
Next experiment: [One remaining uncertainty]
Action, owner, and due date: [Operational follow-through]
Completed Marketing Experiment Tracker Example
This fictional small-sample example illustrates disciplined interpretation rather than statistical certainty.
Show the completed example
Experiment identity
Experiment ID and name: EXP-2026-10 — Referral Partner Introduction Test.
Owner: Alex Morgan
Start and end dates: 30 September 2026
Decision date: 30 September 2026
Status: Active and approved for the stated period
1. Decision and hypothesis
Decision this informs: Proceed with the bounded next step, owned by Alex Morgan, and review the evidence on 30 September 2026.
Uncertainty: Whether relevant advisers can repeatedly identify SaaS founders with an urgent, measurable activation problem.
Hypothesis: Relevant partners receiving the diagnostic message will produce more qualified introductions because the trigger is easier to recognize
What would weaken it: Fewer than five qualified conversations or no paid sprint after 15 appropriate partner asks.
What this test cannot establish: Cold-channel performance, long-term referral volume, delivery margin at scale, or downstream conversion lift.
2. Audience and eligibility
Intended audience: B2B SaaS consultants and fractional marketing leaders serving founder-led companies with active trials.
Trigger or context: Trial volume increased while activation remained flat, making the cost of delay visible in the latest monthly review.
Inclusion rule: Proceed only when the customer, evidence, authority, access, budget, timing, and ethical requirements are all confirmed.
Exclusion rule: Product redesign, paid acquisition, website copy, unlimited revisions, ongoing optimization, and guaranteed commercial results.
Recruitment or traffic source: Two related projects, one paid pilot, six interviews, proposal records, delivery-time data, and a dated source log.
3. Variable and control
Independent variable: A personalized introduction request using one fixed partner brief and qualification rule.
Control or comparison: Use the approved checklist, preserve evidence, resolve critical failures, and obtain written acceptance before closing the stage.
Constants: Same sender, week, follow-up timing, offer, and qualification rule
Contamination risk: Late access, inaccurate data, scope expansion, security exposure, or missed approval; pause work and escalate to Alex Morgan.
Implementation check: Every ask uses the approved wording, qualification criteria, tracked link, and one follow-up after seven days.
4. Baseline and measures
Baseline period: 1–31 October 2026
Baseline counts: Five qualified conversations and one paid sprint from 15 asks.
Primary measure: Qualified introductions within 14 days
Secondary measures: Five qualified conversations and one paid sprint from 15 asks.
Guardrail: No mass outreach, purchased data, unqualified referral reward, misleading urgency, or more than ten owner hours.
Data source: Two related projects, one paid pilot, six interviews, proposal records, delivery-time data, and a dated source log.
5. Budget, exposure, and stop rule
Cash limit: Recorded in EUR on an accrual basis, reconciled to source documents, and approved within the stated €600 direct-cost limit.
Owner-time limit: Alex Morgan
Planned exposure: 15 qualified partners during 1–21 October 2026.
Minimum test period: 1–31 October 2026
Early stop condition: Proceed only when the customer, evidence, authority, access, budget, timing, and ethical requirements are all confirmed.
6. Decision thresholds
Continue threshold: Five qualified conversations and one paid sprint from 15 asks.
Revise condition: Proceed only when the customer, evidence, authority, access, budget, timing, and ethical requirements are all confirmed.
Stop condition: Proceed only when the customer, evidence, authority, access, budget, timing, and ethical requirements are all confirmed.
Guardrail threshold: Five qualified conversations and one paid sprint from 15 asks.
Pre-registered date: 30 September 2026
7. Results
Actual exposure: 14 asks delivered, 11 acknowledged, and 7 introductions received by 31 October 2026.
Primary result: The bounded test met its decision threshold, while delivery capacity and repeatability remain the next uncertainties to test.
Secondary results: The bounded test met its decision threshold, while delivery capacity and repeatability remain the next uncertainties to test.
Guardrail result: The bounded test met its decision threshold, while delivery capacity and repeatability remain the next uncertainties to test.
Cost and owner time: Recorded in EUR on an accrual basis, reconciled to source documents, and approved within the stated €600 direct-cost limit.
Data-quality issue: Use the approved checklist, preserve evidence, resolve critical failures, and obtain written acceptance before closing the stage.
8. Interpretation and action
Decision: Whether Northstar should lead referral-partner outreach with an activation-loss diagnostic instead of a general lifecycle-email introduction
Evidence supporting the interpretation: Two related projects, one paid pilot, six interviews, proposal records, delivery-time data, and a dated source log.
Alternative explanation: Founder-written emails, generic automation templates, a generalist provider, delayed action, or internal delivery.
Commercial effect: Two proposals worth €11,800 and one signed €5,900 sprint with a €2,950 deposit.
Next experiment: Test the same qualification rule with ten new partners outside the owner’s immediate network.
Action, owner, and due date: 30 September 2026
Quality Check
- The experiment informs a stated business decision.
- One important uncertainty and one primary variable are defined.
- Audience inclusion and exclusion rules are explicit.
- Baseline and results include counts, periods, and sources.
- Primary and guardrail measures were chosen before launch.
- Cash, time, exposure, and early-stop limits are fixed.
- Thresholds were not rewritten after seeing results.
- Implementation and data-quality problems are recorded.
- Interpretation includes alternative explanations and sample limits.
- The final decision creates an owner and dated action.
Common Marketing Experiment Mistakes
Changing several variables
A new audience, offer, price, page, and message cannot reveal which change mattered.
Using percentages without counts
A large-looking rate based on a few observations can be unstable.
Stopping only when results look good
Set time, exposure, harm, and decision thresholds in advance.
Ignoring business economics
A conversion increase may still reduce margin or attract poor-fit customers.
Calling correlation proof
Channel, seasonality, source mix, timing, and selection can provide alternative explanations.
Related Guides and Templates
- Choose the channel with Marketing Channel Selection for Solopreneurs.
- Connect results to cost using customer acquisition cost.
- Interpret source credit through marketing attribution.
- Schedule supporting assets in the Content Calendar template.