Use base rates before writing a company growth thesis
A base rate is the frequency of an outcome in a defined comparison group. Use it to challenge how unusual your growth assumptions are, not to assign a company an automatic success probability. The comparison group often matters more than the percentage you calculate.
Define the outcome before selecting peers
Suppose the question is whether a small supplier can double recurring revenue in three years without acquisitions. Choose companies by starting conditions observable at the beginning of that window. Selecting today’s successful suppliers introduces hindsight and omits businesses that disappeared or changed strategy.
Write an eligibility rule covering business model, starting scale, geography, period and outcome definition. The business and risk sections of a 10-K help describe starting conditions. They do not provide a ready-made statistical reference class. Constructing one requires your own documented inclusion decisions.
Hypothetical reference-class exercise
Assume you identify twelve comparable suppliers from an archived starting list. Three doubled comparable recurring revenue, four did not, two were acquired before the endpoint and three have insufficient public disclosures. These numbers are invented for the exercise.
| Group | Count | Treatment |
|---|---|---|
| Verified success | 3 | Check organic definition and unchanged scope |
| Verified non-success | 4 | Retain even if the company later recovered |
| Acquired before endpoint | 2 | Mark competing outcome, not automatic success |
| Unobservable | 3 | Preserve as missing; do not quietly exclude |
Three successes among seven fully observed cases is about 43%, but that is not the result for all twelve eligible companies. With missing outcomes and acquisitions, a single percentage conceals the uncertainty. A simple sensitivity range that treats the five unresolved cases as all non-successes or all successes runs from 3/12 to 8/12. It is a missing-data illustration, not a calibrated prediction interval.
Apply the comparison carefully
Ask what makes the subject meaningfully different. A longer contract term might matter if the mechanism depends on renewal stability; a charismatic founder does not automatically justify discarding the reference class. Record the proposed difference, its observable evidence and whether comparable companies also had it.
A common error is mixing time periods with radically different financing or demand conditions. Another is measuring revenue growth in nominal currency across high-inflation and low-inflation markets. Keep windows and definitions consistent, or present separate groups.
Base rates become weak when the sample is tiny, highly selected or structurally unlike the subject. In that case their value is to expose uncertainty, not to support false numerical precision. Keep the raw company list and unresolved cases so another reader can reproduce the selection.
A further check on the example
Run an exclusion audit on the twelve-company list. For each rejected candidate, record which starting-condition rule it failed and whether you knew its later outcome when rejecting it. A rule introduced only after discovering an inconvenient failure should be treated as a sensitivity change, not silently added to the original selection. Show how the result changes under the broader list. This does not eliminate judgment, but it makes the most consequential judgment—the construction of the comparison group—available for challenge.
FAQ
Can I use an industry average as a base rate?
An average growth rate answers a different question from the frequency of doubling. Match the statistic to the outcome you are investigating.
What if the business model is new?
Use a broader mechanism-based comparison only with explicit limitations. “No good comparison” is a valid finding and should reduce confidence in a precise probability.