Negotiate Experience Scope and Decision Rights
10 min read
Lesson objectives
- Prepare a decision brief that separates sponsor demand from performance evidence, compares interventions under time, privacy and safety constraints, and specifies who decides and when a pilot stops.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
An advanced designer faces a problem beyond selecting a learning activity: several stakeholders may each be partly right. A sponsor wants results quickly, supervisors fear that training will delay work, and a data steward sees unnecessary detail in the proposed measurement plan. Averaging their requests could produce a short path that teaches nothing and exhaustive tracking without a defensible purpose. A decision brief separates four things: an observed fact, a plausible explanation, a constraint that cannot be ignored, and a decision owned by a named role. That distinction is not bureaucracy for its own sake. It makes the reason for a design visible when the result differs from the expectation.
At fictional Nawa, the quality of case handoff documentation fell from 82 complete cases in a review of 100 last quarter to 69 of 100 this quarter. The sponsor asks for a mandatory two-hour course within ten days and promises a return to 90%. Supervisors say a new handoff form added three unclear fields and freezes near shift end. The data steward objects to storing every named employee click for a year because no improvement decision requires that level of detail. These statements do not settle the cause. We know of a descriptive fall in two reviewed sets, but not yet whether case types changed or reviewers agreed on ‘complete’. There is no evidence for a promise of 90% in ten days.
Build the brief around decisions, not a requirements wish list. First, should the form interface be repaired before everyone is assigned new learning? Second, which knowledge gap remains after removing an interface barrier? Third, what minimum data would show whether the intervention is useful and safe? Compare feasible alternatives: A, a general course immediately; B, an interface test followed by short practice for handoff cases that reveal a misunderstanding; C, an interface repair alone followed by measurement. For each, state a possible benefit, risk and needed evidence. A may launch quickly but imposes time cost and confuses mechanisms. C reduces burden but can leave judgment errors untreated. B is more involved yet can separate the barriers if its steps are sequenced clearly.
This tradeoff is not a decorative scorecard. Apply eligibility conditions before ranking alternatives. Any option requiring unnecessary named tracking or teaching staff to pass an incomplete case is out of scope, however fast it looks. Then compare time, reversibility and strength of evidence. In week one a product specialist could clarify field labels and test the form with six workers across shifts, while a task expert classifies 30 records against a stated criterion. If judgment errors remain with fields visible, the learning team designs 15 minutes of practice on boundary cases. These are fictional planning figures, not evidence that six users represent everyone or that 15 minutes suffice in every organization. They make the decision bounded and testable.
In the worked case, reviewers inspect 30 current-quarter cases. In 12, the third field is not visible on a phone; another six have complete fields but a handoff decision that conflicts with the rule. The team chooses B in two stages. First make the field visible and test form completion on a phone without permanently collecting each worker's click history. Then give the six workers, or whoever shows that gap, graduated decision practice with fair access for the other shift. Monitor completeness under one definition, incorrect handoffs and failures to submit the form. If completeness improves but decision accuracy does not, do not announce complete success; investigate practice or rule clarity. If the repair blocks a valid case, stop the first stage and correct that interface fault.
The sponsor might object: ‘I need one number next week.’ An honest interim number is complete cases in a specified reviewed sample, with sample size and the proportion unclassifiable. Do not present it as a final program effect when the interface and practice changed in sequence and sampling may differ. If the sponsor wants employee rankings, ask what decision requires names. An interface repair and work-rule practice can often begin with aggregates or temporary test identifiers, subject to local access review. That is an example of data minimization in design, not a universal legal opinion. Actual legal and organizational requirements must be checked in their own context.
Equal sample sizes of 100 do not make the quarters comparable. Routine cases may have dominated before, while more cases requiring a second approval appeared now. Classify case type, complexity and shift, then apply the same completeness rule to samples from both periods. Reclassifying ten cases could change a rate without any staff behavior change, so keep the criterion version and change date. If reviewers disagree substantially, settle what ‘complete’ means before using the rate to evaluate learning. The brief may state that the apparent decline warrants investigation, rather than declare the size of a skill deficit. That is a stronger claim because it protects a decision from a precise-looking number whose meaning shifted.
The alternatives change order if specific evidence changes. If device testing shows every field clearly while most staff misinterpret the handoff rule in a separate simulation, decision practice becomes more urgent than repairing an interface fault that has not appeared. If the six ‘wrong’ cases belong to a new type that the process owner has not classified, do not author an exercise with one prescribed answer until the rule is settled. State both counterexamples in the brief. They show what makes a recommendation conditional instead of a preference. When new evidence arrives, revise the affected hypothesis and check its consequence for measurement and activity design; the entire project need not restart.
Frame data collection around a decision too. The sponsor wants employee names ‘for follow-up’, but what action follows one person's single error? For a phone-form repair, limited test records showing device type and failed step may suffice without permanent identity. If a worker needs skill support, an accountable coach may see a specific task result under restricted access rather than a public ranking. Set retention according to decision need and institutional policy, not because a year is technically convenient. A specialist checks actual local requirements before deployment. This example teaches purpose limitation and data minimization; it does not issue a legal judgment for every country.
Finally, express a stop condition as an action rather than a slogan. If an automated check blocks a valid phone submission, the product owner pauses the pilot version and inspects that case without widening data collection. If an incomplete case reaches a team that cannot handle it safely, the process owner stops the new flow until the barrier is understood. An inconclusive sample is different: it may justify a bounded observation extension or a revised comparison with stakeholder agreement, not an automatic cancellation. Distinguishing a safety stop from deferred judgment lets the team learn from a pilot without exposing people or work to an indefinite test.
Lesson summary
Prepare a decision brief that separates sponsor demand from performance evidence, compares interventions under time, privacy and safety constraints, and specifies who decides and when a pilot stops.