Guide · C76
Bound a learning-value heuristic
Which learning device merits study before deciding without confusing it with EVSI?
Direct answer
The heuristic subtracts cost and delay from a declared avoidable loss weighted by assumed reliability; it does not model possible test outcomes.
01Reject, investigate further or provisionally keep a learning device with its assumptions made explicit.
02The rule, the evidence and the next review date stay visible.
How to use this
Move from the question to the deliverable.
Which learning device merits study before deciding without confusing it with EVSI?
Reject, investigate further or provisionally keep a learning device with its assumptions made explicit.
Learning-value heuristic
Original chart · SQ-76
Bound a learning-value heuristic
Information is valuable through the decision it can change. This framing heuristic compares a stated avoidable loss with the cost, delay and assumed reliability of the device, without claiming to calculate EVSI.
Fictitious decision between a national launch, regional pilot or deferral for an offer with uncertain adoption.
- 01Model without information
Calculate the best decision under current probabilities.
- 02Identify switches
Locate outcomes capable of changing the selected option.
- 03Bound
Test a range of avoidable loss and device reliability.
- 04Compare
Subtract cost, delay and value lost during learning.
Net heuristic of four learning devices
The proxy uses a stated avoidable loss of €140k. The regional pilot offers the best trade-off in this scenario alone.
Net heuristic = stated avoidable loss × assumed reliability − test cost − delay cost| Device | Assumed reliability | Cost | Delay cost | Net heuristic | Decision |
|---|---|---|---|---|---|
| No measurement | 0% | €0k | €0k | €0k | Decide now |
| Interviews only | 35% | €12k | €8k | €29k | Insufficient alone |
| Quantitative study | 60% | €45k | €18k | €21k | Too costly |
| Regional pilot | 75% | €52k | €20k | €33k | Selected |
| Controlled national test | 90% | €105k | €32k | −€11k | Rejected |
01The regional pilot maximises the heuristic in the stated scenario without proving that it maximises expected information value.
02The national test destroys €11k of value because its additional precision does not offset its cost.
03For the pilot, indifference with an immediate decision occurs at €53k of delay cost; above that, deciding now is preferable again.
Fictitious case. This indicator is a framing heuristic, not EVSI. Expected information value requires possible test outcomes, their probabilities and the optimal decision after each outcome.
Coherence controls
Minimum conditions before validation.
The status, the evidence and the owner must be available for review.
The status, the evidence and the owner must be available for review.
The status, the evidence and the owner must be available for review.
The status, the evidence and the owner must be available for review.
Usable deliverable
Learning-value heuristic
Explore the decision further with an interactive control, or start from a structured template to produce your deliverable.