Choose leading indicators that test a business thesis
A leading indicator is useful only when it provides an earlier, relevant observation of the mechanism you want to test. Being published before revenue does not make a number predictive. Specify its population, expected lag and the result it is supposed to precede.
Connect the indicator to an economic step
For a subscription product, trial signups precede paid conversions, which precede recognized revenue and collection. Each transition can fail. IFRS 15 ties revenue recognition to satisfaction of performance obligations; an early customer action is therefore not automatically reported revenue. Check the issuer’s actual accounting framework and revenue policy.
Choose indicators close enough to the mechanism to be interpretable. Website visits may be early but affected by advertising, bots and free users. Paid activations might be closer to revenue, but refundable promotions and delayed service delivery still matter.
Hypothetical conversion check
Assume a fictional provider receives 2,000 trials in March versus 1,000 in February. The headline suggests acceleration. However, February produced 300 paid customers after thirty days while March produced 240. These invented numbers illustrate why equal observation ages matter.
| Cohort | Trials | Paid after 30 days | Conversion |
|---|---|---|---|
| February | 1,000 | 300 | 30% |
| March | 2,000 | 240 | 12% |
If March is only ten days old, the 12% cannot be compared fairly with February’s thirty-day rate. Wait until both cohorts mature or compare at ten days. If both are fully mature, the larger funnel still produced fewer paying users. Neither interpretation follows from trials alone.
Create an indicator specification
Record: indicator name; exact numerator and denominator; source; cohort date; age at measurement; expected lag to the target outcome; competing explanations; and the condition that would make you stop using it. Keep this specification alongside results rather than reconstructing it later.
Verify against several completed cycles when possible, including weak periods. A relationship visible only in one promotional quarter is fragile. If the company changes the signup definition, restart or bridge the series. Avoid choosing the lag after inspecting every possible lag and reporting only the most favorable one.
Leading indicators are particularly limited when disclosure is selective or the business model changes. A backlog can contain cancelable orders; bookings can include multi-year commitments; active users can include unpaid accounts. Label the contractual status instead of assuming equivalence.
The practical aim is earlier investigation, not an automatic investment signal. A false alarm log is useful: record cases where the indicator moved but the expected outcome did not, and investigate the broken transition.
A further check on the example
Specify how refunds affect paid activations. If March users pay on day twenty but can cancel for a full refund until day forty-five, a day-thirty paid count may not be comparable with a non-refundable February offer. Record both the offer terms and the maturity date. When no clean historical comparison exists, use the indicator as an operational observation with limited predictive status. Do not describe a promotional conversion rate as a durable customer-economics result before the refund window closes.
FAQ
Can management’s own KPI be used?
Yes, if its definition, coverage and revisions are available. Company origin does not make it useless, but changes in presentation require scrutiny.
How many indicators should I track?
Enough to cover the necessary transitions. Ten versions of the same signup metric do not provide ten independent views of demand.