Research and Innovation: KPIs
Measures for a function judged on questions answered and conversions along the research to revenue chain, not on ideas produced.
Markdown. No sign-up, no email.
A function measured on ideas will produce ideas indefinitely. These measure answers, and what happened to them afterwards.
The six that matter#
| Measure | Definition | Target | How it gets gamed |
|---|---|---|---|
| Questions answered | Settled, not explored | 6 to 10 per quarter | Counting an unfinished investigation as an answer |
| Median cost per answer | Total spend divided by questions settled | Falling | Only counting cheap questions |
| Kill rate | Experiments stopped at their criterion | 40 to 60% | Reporting zero, which means nothing risky was attempted |
| Time from question to answer | Median | Under 6 weeks | Answering only easy questions |
| Negative results documented | Written up and findable | Every killed experiment | Skipping the write-up, which is where the compounding is lost |
| Transfers to product or service | Findings that became something sellable | 2 or more per year | Counting a prototype nobody productised |
Kill rate is a band, not a maximum. Below 40% means the questions were too safe to be worth asking. Above 60% sustained means the selection is poor.
Negative results documented has no acceptable shortfall. It is the output that stops repetition, and it is the first thing dropped because writing up a failure is nobody's favourite afternoon.
Chain conversion, tracked yearly#
The chain is Research to Knowledge to Product to Service to Revenue. Each arrow gets a number.
| Conversion | Question |
|---|---|
| Research to knowledge | What share of answers were written up so someone else can use them? |
| Knowledge to product | What share informed something we built? |
| Product to service | What share became a packaged, buyable offering? |
| Service to revenue | What share earned money more than once? |
Most research functions score well on the first arrow and near zero on the third, and never measure it, so the pattern persists for years while the function feels productive.
Two counter-metrics#
| Counter-metric | Catches |
|---|---|
| Experiments running past their time box | A function that has stopped killing things. The first symptom of research becoming a hobby |
| Questions asked by other functions vs by ourselves | If nobody else brings us questions, we are researching what interests us rather than what the company needs |
What is deliberately not measured#
- Ideas generated. Free, and inversely correlated with focus.
- Papers read, trends tracked, prototypes built. Activity.
- Novelty. Being first to a technology that is not yet worth its switching cost is a cost, not an achievement.
Cadence#
| Weekly | Experiments running, anything past its time box |
| Monthly | Questions answered, negative results written up |
| Quarterly | Queue review with the CEO and AI Strategy, because the queue is a statement about strategy |
| Annually | Chain conversion, and what reached revenue |