GOALS.Reference:5.1 - Choose the results and their completion basis
A refund service has 100 eligible cases in its opening cohort, all due by the end of a four-week cycle. Customers want the settlement owed and a usable account of the outcome. The service also wants faster handling of new cases and two employees able to diagnose exceptions. The following amounts, terms and outcomes are stipulated for this example. The service’s competent providers supply the case criteria.
The initial KRs are “publish a FAQ”, “launch a bot” and “reduce incoming contacts”. All three could pass with every refund unpaid. The team retains useful information and automation as means and constructs two objectives with different reliance:
| Objective and mode | Key results and their joint meaning |
|---|---|
| O1, committed: By the end of week 4, settle the 100 opening-cohort refund obligations and give every affected customer a usable confirmation of the outcome. | KR1.1: All 100 cases have a completed correct payment or a valid final settlement under their applicable terms, with no unresolved amount, recipient or authorization error. KR1.2: Every customer has received an accurate confirmation through the agreed channel, identifying the actual settlement and any action still required of them. Both KRs are mandatory. A confirmation of a future answer does not satisfy KR1.1. |
| O2, aspirational: Make new eligible refund cases dependable and less costly to handle during the selected service window. | KR2.1: At least 95% of all eligible requests received in week 3 reach a correct payment or valid final settlement within five working days of receipt; every case also meets its individual contractual deadline. KR2.2: Mean manual handling effort for that same complete cohort is at most 12 minutes per request, including exceptions and its allocated checking and maintenance work. Both KRs are required for the stated improvement. Correct amounts, recipients, authorization and agreed customer access remain required conditions. |
Eligibility follows the existing service terms and is checked at receipt. Difficult eligible cases remain in their cohort. In this case the incoming requests’ contractual deadline is ten working days, so the five-day target is an improvement. Requests arriving at the end of week 3 can still receive their full five-day observation by the end of week 4. Any cases open then remain tracked to their actual contractual deadline. O1’s fixed opening cohort and O2’s new cohort are separate; improving the latter leaves the earlier obligations intact.
The preceding comparable service window had 85% completed within five working days and 20 minutes of manual effort per request. The initial O2 forecast falls short of the desired improvement: 90% within five days and 15 minutes per request, against the desired 95% and 12 minutes. The team chooses the ambitious improvement while protecting the current contractual service. Other teams plan on confirmed staffing and contributions, not on the hoped-for effort saving. O1 has funded payments, the required authorizations and a feasible processing arrangement; its initial forecast is all 100 completed by the agreed date. A threatened payment or dependency will reopen that forecast and the response.
Payment confirmations and the existing case register establish financial and communication outcomes. Existing case timestamps and handling-time data cover O2’s whole cohort. The effort calculation includes the service’s manual exception work; omitting difficult cases would change the claim. These adequate sources are reused without a new reporting system. The all-pass test is now concrete: correct final settlements plus usable confirmations establish O1; timeliness plus lower total handling effort under the retained conditions establish O2. The reverse test also matters: a correct manual settlement may satisfy the goal even if the bot is never launched. Faster work with wrong amounts fails O2’s required conditions, and lower incoming contacts alone establishes neither objective.
The separate learning goal is for both employees, by the end of week 4, to frame unfamiliar routine and exception cases in the current refund service, identify missing grounds and route a decision beyond their authority. Their case procedure remains available; an AI-generated diagnosis is withheld only in the episodes used to establish their own diagnostic contribution. The HCD provider chooses representative variations, feedback and assessment for that bounded capability. This goal is not a KR proving customer settlement: a competent bot can improve service while the employees still need the intended practice.