The Evidence Debt Ledger: Track What Your Growth Team Is Assuming

Growth teams accumulate more than technical debt. They also accumulate evidence debt: decisions, reports and automations that depend on assumptions nobody has verified recently. A target audience is treated as proven because it was written in an old brief. A dashboard calls a form submission a qualified lead. A channel receives credit because its tracking is easier to observe than the rest of the journey.

Evidence debt is not the presence of uncertainty. Every decision contains uncertainty. Debt appears when an uncertain claim becomes infrastructure without an owner, an expiry date or a plan to test it.

Recognize the units of evidence debt

The basic unit is a claim used to justify an action. Examples include:

• “This audience has the highest buying intent.”
• “This conversion event predicts sales quality.”
• “The sales team follows up within the agreed window.”
• “This campaign is incremental rather than capturing existing demand.”
• “The CRM source field is reliable enough for budget allocation.”
• “Customers choose us because of this feature.”

Each claim may be true, partly true or outdated. The problem is that its status is invisible while the team continues to build on it.

Create an evidence debt ledger

Field	Purpose
Claim	States the assumption in testable language
Decision affected	Shows what could change if the claim is wrong
Current evidence	Links to the strongest available support
Evidence class	Observed, inferred, modeled, reported or unknown
Last reviewed	Shows when the claim was examined
Risk if wrong	Describes financial, customer or operational exposure
Owner	Names the role responsible for the next review
Next test	Defines the smallest useful way to reduce uncertainty
Expiry or trigger	Prevents an old conclusion from remaining permanent
Status	Accepted, bounded, testing, overdue or retired

The ledger should link to evidence, not copy entire reports. A future reviewer needs to find the source, understand the method and see the dates.

Score debt by decision exposure

Do not prioritize the claims that are easiest to test. Prioritize the claims that can cause the most damage when wrong.

Use four questions:

1. Impact: How much spend, revenue, customer experience or risk depends on this claim?
2. Dependence: How many workflows and decisions reuse it?
3. Uncertainty: How weak, indirect or outdated is the evidence?
4. Detectability: Would the team notice quickly if the claim failed?

A rough high/medium/low scale is usually sufficient. Multiplying invented precision across subjective scores can make the ledger look scientific without improving the priority.

Distinguish five evidence classes

• Observed: directly recorded under known conditions.
• Inferred: concluded through a documented business rule.
• Modeled: estimated by a statistical or platform method.
• Reported: provided by a person, customer or external party.
• Unknown: evidence is absent or insufficient.

None of these classes is automatically “bad.” Customer-reported reasons can be valuable. Modeled results can support planning. The class tells the decision-maker what kind of claim is being made and what verification would strengthen it.

Set an evidence budget for large decisions

Before increasing exposure, decide which debts must be reduced. For example, a team may permit a small channel pilot with partial source linkage but require reliable qualified-outcome feedback before scaling. The evidence requirement should rise with the irreversibility and cost of the decision.

Decision	Acceptable evidence posture
Small reversible pilot	Explicit assumptions, bounded spend, primary tracking tested
Meaningful budget increase	Downstream quality evidence and stable measurement coverage
Automated optimization	Reliable event definition, monitoring and stop-rule
Strategic channel commitment	Multiple periods or methods, economic evidence and known limitations

Use expiry dates carefully

Evidence does not become false on a calendar date, but the environment can change. Set review triggers based on what threatens the claim:

• a material product or pricing change;
• a new market or audience;
• a platform or tracking migration;
• a sales-process change;
• a sustained change in outcome distribution;
• the passage of an agreed period for a volatile assumption.

When a trigger occurs, the status becomes “review required.” Do not silently reset the date.

Run a monthly debt review

1. Inspect overdue high-exposure claims.
2. Check whether new projects created unregistered assumptions.
3. Close debts that have sufficient evidence or no longer affect a decision.
4. Choose a small number of tests for the next period.
5. Update dependent dashboards, briefs or automation when a claim changes.
6. Record what the team deliberately accepts without further testing.

The last step matters. Some uncertainty is too expensive to remove. A documented acceptance with a bounded risk is more honest than a permanent “research later” task.

Starter ledger prompts

• Which conversion events are treated as business outcomes?
• Which audience beliefs determine most of the media budget?
• Which source fields are trusted without coverage reporting?
• Which sales explanations are based on anecdotes rather than coded outcomes?
• Which automation assumes a field is complete or correct?
• Which “best practice” entered the process without a local test?
• Which decision would be embarrassing to explain if its underlying claim were false?

An evidence debt ledger does not require the team to prove everything. It prevents hidden assumptions from acquiring the authority of facts. That makes experiments more useful, dashboards more honest and scaling decisions easier to reverse when the underlying evidence changes.