GTM Basics · Evidence guide
How to judge a GTM claim
A correct calculation, a plausible forecast, a pattern in the data and an intervention effect are four different things.
1 · Accounting identity
Claim: “Ending monthly recurring value equals starting value plus new and expansion, minus contraction and churn, plus defined adjustments.” This is a ledger reconciliation. On a fixed scope, period and currency, with movements counted once, it must balance. If it does not, inspect the data and definitions.
What it does not say: why customers churned, what next month will be, or what a new campaign would change. A precise identity can contain badly measured inputs. Monthly recurring value is also different from recognized revenue and collected cash.
2 · Assumption
Claim: “If a new channel produces 20 qualified conversations and 25% buy, we will win five customers.” The multiplication is exact; the conversion rate is a supposition about this channel, segment and period. A historical rate from another segment does not turn it into a fact here.
Use it for: a scenario or a bounded capacity plan. Show the input, its source and the decision's sensitivity: at 10% conversion, the same 20 conversations yield two customers. Record the actual denominator and check the assumption against the next cohort before scaling.
3 · Observed relationship
Claim: “Accounts that attended a sales demo renewed more often than accounts that did not.” That describes an association in the observed accounts. More motivated buyers may both request demos and renew; sales may choose to demo the strongest accounts; product fit, size and cohort age may differ.
Use it for: finding segments or mechanisms worth investigating. Report the populations, time window, outcome, missing records and group sizes. A chart or regression can adjust for recorded differences, but unrecorded selection may remain. “Demo attendance caused retention” goes beyond this evidence.
4 · Causal finding
Claim: “Offering a demo increased 90-day activation among eligible accounts by four percentage points.” This is a causal claim only if the comparison estimates what would have happened to those accounts without the offer. For example, randomly assign eligible accounts or buying groups to a demo offer or usual process, keep the original assignment groups, and compare the pre-specified activation outcome. Measure the offer's effect; comparing people who chose to attend would reintroduce selection.
Even a randomized result has a scope: the tested offer, eligible accounts, period and outcome. Check assignment, sample size and uncertainty, spillovers between groups, missing outcomes and whether the policy was delivered. A credible observational design may also support a causal estimate, but only with a defensible comparison and explicit assumptions about confounding. Neither design makes the effect universal or proves later renewal.
Put the claim beside the decision
| Type | What it establishes | Next question |
|---|---|---|
| Identity | The ledger balances under stated definitions. | Are scope, categories and inputs complete? |
| Assumption | A conditional scenario, not an observed result. | What if the input is wrong, and how will we measure it? |
| Relationship | A pattern in these observed groups. | What selection or common cause could explain it? |
| Causal finding | An estimated effect of a specified intervention, if its comparison is credible. | How uncertain is it, and will it carry to this decision? |
Before repeating a GTM claim, write down its population, denominator, period, outcome and source. For a forecast, add its assumptions. For a causal recommendation, name the intervention, comparison and uncertainty. Then ask whether the possible effect is large enough to matter after acquisition, delivery and capacity costs. An evidence label is a starting point for judgment, not a stamp of certainty.
See these distinctions applied to fictional pricing and acquisition decisions in the worked GTM plan, or examine the typed claims in the GTM World Model.