# How TractionScout measures growth

TractionScout measures changes toward the outcome selected by the project
owner. Every useful comparison needs a defined metric, unit, population,
source, and time window. Missing observations remain missing; completing a
task or seeing more impressions does not establish more activated users or
revenue.

The project chain, metric rows, Details evidence, and Scout use shared
measurement definitions. Use those definitions when asking a question or
evaluating a change rather than creating a second calculation with the same
label.

## Read a number with its definition

| Measurement | What must stay attached |
| --- | --- |
| Flow, such as completed signups | Recorded completions in the selected period, deduplication and source coverage |
| Stock, such as an account balance | The observation date; daily balances are not added together |
| Rate, such as activation | Numerator, eligible denominator, identity scope and completion horizon |
| Distribution, such as time to value | The measured observations, aggregation and population |
| Currency | Currency, revenue meaning and qualification; no silent currency conversion |

Inspect the source references and dates in Details. Partial imports, stale
connections, unfinished reporting days, and unmeasured outcomes are different
limitations. A recorded zero is a value; an absent reading is not zero.

## Preserve the owner's outcome

The selected objective declares what progress means and its desired
direction. Supporting metrics explain a mechanism or a problem. They do not
replace the objective because they are easier to observe.

For example, search impressions may help explain reach when the objective
is activation. They do not prove that more visitors activated. If activation
is not measured, investigate its completion event and identity coverage
before claiming progress.

## Compare compatible periods

Use matching source scope, metric meaning, filters, unit, and population.
The chosen reporting range and the comparison's completed observation
buckets may differ while a period is still open; preserve the displayed
coverage and provisional status instead of treating them as a final result.

A positive comparable baseline supports a descriptive percentage change.
A zero baseline supports absolute change but no finite growth percentage.
A missing or incompatible baseline supports neither a fabricated trend nor
an infinite percentage. Rates should retain their denominators; a move from
20% to 25% is five percentage points.

Do not average project growth percentages to make a business total. Even
compatible additive metrics need disjoint populations to represent a
deduplicated total. Two sources may describe the same people, and different
currency or attribution scopes must stay separate.

## A funnel needs more than separate counts

An ordered funnel requires recorded links between the same subjects and
the defined steps, within a declared completion window. Independent daily
signup, trial, and Stripe totals cannot establish which signups paid.
Neither current account state nor an elapsed trial proves a completed
conversion or lost access.

Compare entrants who have had the same time to complete. Preserve unknown
identity and unclassified source coverage. When the evidence cannot support
a join, report the separate observations and explain the measurement gap.
See the [trial activation workflow](/workflows/posthog-stripe-readout.html)
for a worked, explicitly fictional cohort.

## Separate observation from explanation

Scout can state that a metric changed and investigate plausible causes.
A page rewrite, launch, audience change, pricing change, or instrumentation
change may overlap the same period. A before-and-after association does not
isolate the effect of one action.

Review the existing change or experiment record with its baseline, outcome,
measurement window, coverage, and verdict. Preserve an inconclusive result
when evidence is insufficient. A randomized experiment needs its own valid
design, assignment, adequate observations, and evaluation before a causal
claim is appropriate.

## Keep a reproducible record

Save the source report or receipt, definition, filters, observation dates,
baseline, actual change, and intended follow-up. Use the existing project
evidence and change paths. Record what would support or contradict the
hypothesis before reading the result; do not invent a target from the sample
calculator.

Start with [Getting started](/docs/getting-started.html),
[acquisition channel activation](/docs/acquisition-channel-activation.html),
or [Search Console clicks](/workflows/search-console-content-experiment.html).
Measurement informs a decision; publishing, spending, and changing the
objective still follow their existing permissions.


Maintained by TractionScout · Reviewed 2026-10-02 · Updated 2026-10-02