Ahead-of-the-Data Response Rule
Act on leading signals before exponential growth becomes visible
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 97%
The rule treats reported cases, hospitalizations, deaths, and similar observations as delayed traces rather than a live picture of an exponential threat. Decision-makers first estimate the lag between cause and visible outcome, then reason forward from the earlier state that produced today's data. Instead of requiring certainty, they define credible leading signals and pre-agreed action thresholds. When those thresholds are crossed, they deploy proportionate and preferably reversible interventions while continuing to update the model. The mechanism converts weak early evidence into timely containment: infer the unseen trajectory, account for compounding, act ahead of visible damage, and communicate why prevention must begin before the public can directly observe the danger.
Origin
Extracted from Masters in Business through Michael Lewis's account of the pandemic planners featured in The Premonition.
Core principles
- 01Exponential threats outrun slow decision processes
- 02Visible damage is a lagging indicator
- 03Containment requires action before certainty
- 04Trusted expertise turns weak signals into timely decisions
How to run it
- 1
Classify the Growth Pattern
Determine whether the threat can compound or spread exponentially. Do not project future conditions with a simple linear extension when each new case can generate additional cases.
Pro tip Use analogies such as fire or compound interest to make acceleration intuitive.
Watch out Mistaking exponential growth for arithmetic growth creates false confidence about available time.
- 2
Measure the Observation Lag
Estimate the delay between an initiating event and the data it eventually produces. Work backward from today's visible outcome to the earlier conditions that caused it.
Pro tip Map separate delays for detection, reporting, hospitalization, and death.
Watch out Current-looking dashboards may describe conditions from days or weeks earlier.
- 3
Infer the Hidden State
Use the lag and estimated growth rate to calculate a plausible range for the threat's present scale. Consider multiple scenarios instead of relying on one precise forecast.
Pro tip Emphasize the range of plausible consequences rather than false numerical precision.
Watch out Do not confuse incomplete evidence with evidence that the threat is absent.
- 4
Precommit to Action Thresholds
Specify which leading signals will trigger intervention before the crisis begins. Tie each threshold to a proportionate response.
Pro tip Favor interventions that can be strengthened, relaxed, or reversed as evidence improves.
Watch out Thresholds written during a crisis are vulnerable to political pressure and motivated reasoning.
- 5
Act and Update
Intervene once the threshold is crossed, then continually revise assumptions with incoming evidence. Explain that preventing visible harm is the intended outcome, not evidence that action was unnecessary.
Pro tip Track both threat indicators and the costs created by the intervention.
Watch out Waiting for visible catastrophe defeats the purpose of early containment.
In the wild
A jurisdiction records its first death from a rapidly spreading virus. Rather than treating it as one isolated event, officials account for the delay between infection and death, infer a larger pool of infections weeks earlier, and project the subsequent compounding. They activate targeted distancing and testing measures while refining their estimates.
→ The jurisdiction acts while containment remains possible instead of waiting for the death count to reveal the already-expanded outbreak.
Common mistakes
Waiting for Conclusive Data
By the time delayed outcomes conclusively prove exponential spread, the threat may already exceed containment capacity.
Projecting Linearly
Extending today's observed rate in a straight line ignores acceleration caused by compounding transmission.
Calling Prevention an Overreaction
A successful intervention suppresses the visible outcome, which can make the original warning appear unjustified.
Is it for you?
Best for
It is best for leaders managing exponential, high-consequence threats with delayed observations.
Not ideal for
It is not ideal when early intervention is irreversible and its downside exceeds the plausible threat.
From the transcript
“By the time you have the data, you're dead.”
“You've got to act before you have visible evidence to act.”
“the current data is already lagging in a few weeks behind”
From the episode
Michael Lewis on White House Pandemic Planning (Podcast)