“Improve staff training” is a direction, but it leaves a practice meeting with an awkward question: what would count as improvement? Publishing a new folder, holding a session and writing a checklist are visible activities. None of them, on its own, tells you whether a member of staff can find and use the information needed for a particular administrative task.

Objectives and key results, or OKRs, offer a way to separate the change you want from the evidence you will examine. A first trial can stay small: one administrative objective, a few clearly defined results and a manageable set of tasks. The useful discipline is agreeing what the evidence means before the team starts collecting it.

1. Separate the objective, the result and the task

An objective describes the intended change and why it matters. For example: “Make approved onboarding information easier for new reception staff to find and interpret.” It names a problem worth addressing without assuming that a longer manual or a new software tool is the solution.

A key result specifies the evidence you will use to assess progress, together with a timeframe and an agreed target. A task is work you plan to do to influence that result. Creating an index is a task. Checking whether staff can locate the correct current document for a defined administrative question is a possible source of evidence. The distinction keeps the team from declaring the objective achieved merely because the index was published.

Use a small number of results that the team can explain and maintain. There is no need to impose the same number on every project or create a separate scorecard for every person. Include enough evidence to examine the intended change, while recognising that the time spent collecting data is also work.

Where you do not yet know the starting position, make that uncertainty explicit. A short measurement exercise may be the first task. Do not invent a baseline or choose an impressive percentage to make the plan look complete. Set the target after considering the actual starting position, available resources, scope and review period.

2. Write down exactly what the measure includes

For each proposed result, define the question, the source of information, the starting value, the target, the observation period and the person responsible for checking it. If it is a rate, specify both the numerator and denominator. “Most staff can find the guidance” is open to several interpretations; an agreed set of tasks and a clear definition of a successful attempt give the team something more concrete to inspect.

Decide what counts before the review. Does success mean opening a document, or finding the current approved version for the question asked? Is assistance allowed? Are repeated attempts counted separately? What happens when the document is unavailable? These are part of the definition, not details to decide after seeing an inconvenient result.

Keep missing observations visible. A staff member who was unavailable has not demonstrated either success or failure. If you report a success rate for observed attempts, report the number of eligible observations, the number actually observed and the number missing alongside it. With a small group, retain the raw counts so a percentage does not conceal how little information is available.

Compare like with like. Changing the questions, access arrangements, observation period or participants can change what a result means. Record those changes instead of presenting a before-and-after figure as if the conditions were identical. A local trial can identify a problem worth fixing without establishing that all employees will have the same experience.

3. Match responsibility with influence and resources

Ask the team what it can realistically influence. Reception staff may identify unclear instructions, but a manager may control document approval, access permissions and training time. Put those dependencies in the plan. Assigning the result to reception does not transfer the manager's responsibility for providing those approvals, access and time.

Invite staff to suggest measures and explain what those measures would miss. Someone who uses the information may notice that opening the correct page is easy while understanding the next action is not. Another person may point out that the proposed test occurs during the busiest part of the shift. These observations belong in the design discussion, before the target is agreed.

Choose an owner for the objective and owners for the practical actions. Confirm the time available, the decisions each person can make and where additional approval is needed. A result that depends on another team should identify that contribution rather than treating the final number as the performance of one employee.

Explain how the information will be used and who will see it. Do not announce that the exercise is unrelated to appraisal or pay unless the practice has actually adopted that arrangement. If its purpose or consequences are unclear, resolve that before collecting individual observations. Share an appropriate summary with the people doing the work; broad visibility does not require public patient records or individual staff scores.

4. Try an administrative objective with a fictional onboarding example

Consider a fictional practice whose reception onboarding folder contains several shortcuts and document versions. This is a worked planning example, not a report of real staff performance. The proposed objective is to make a defined set of approved administrative instructions easier to locate and explain. The exercise excludes clinical instructions and does not assess readiness to provide care.

The manager and staff first agree the scope. For illustration, it includes where to find the current induction timetable, how to report a technical access problem and where to locate the practice's administrative training-request instructions. The document owner confirms which version is current and whether the instructions accurately describe the practice's process. A search test cannot determine that content is correct merely because a file is easy to find.

One proposed result concerns locating the correct current document. Define a successful attempt as reaching the approved version for an agreed task under the stated test conditions. Record successful attempts over observed attempts, with the number of planned attempts not observed shown separately. Use the same task set for the baseline and review, and document any unavoidable changes.

A second check concerns understanding the next administrative step. Agree beforehand what an acceptable explanation must include, based on the approved process. After finding the document, the participant explains whom to contact or what action the instructions call for. Record whether the explanation matches the approved process and note the specific ambiguity when it does not. This is a limited check of those instructions, not a general score of staff competence.

The team leaves the baseline and target fields visibly unfilled until it has the necessary observations and has agreed a feasible target for the review period. Until the baseline, target and review date are agreed, this is a preliminary measurement plan rather than a completed OKR. The examples here supply definitions, not benchmark percentages or invented starting results. If too few comparable observations are available, report the individual task findings and the limitation instead of manufacturing a trend.

Possible tasks include replacing obsolete shortcuts, adding a clear version label and showing where the current index lives. Assign each task an owner and a review point. Keep the task list separate from the result: completing all three tasks does not automatically establish that staff can find or interpret the instructions.

5. Check the quality of the result as well as the count

In the fictional exercise, the correct-document retrieval rate alone is not enough. Check that the opened document is correct, current and accessible to the intended role. Ask whether the instructions give a usable next step and a route for questions. Record confusing wording or access problems even when the participant eventually reaches the right answer.

Use suitable dummy questions and an agreed practice environment. Do not submit real training requests, change personnel records or trigger messages simply to demonstrate the route. Arrange observation so it does not leave patient-facing work uncovered. Explain what will be recorded, avoid unnecessary personal detail and give participants a way to flag a problem with the exercise.

A numerical result and a qualitative check answer different questions. The count may show how many observed attempts met the definition; the comments may explain where a page name or instruction caused difficulty. Neither alone proves why the result occurred. Use them to choose a specific next investigation or correction, rather than assigning a motive to the person who struggled.

The same restraint matters when considering other practice goals. A treatment-completion, return-visit or referral rate is not direct proof of trust, safety or appropriate care. Patients may need time to consider, discuss affordability, postpone, reschedule or decline. Clinical decisions remain with the appropriate professionals and patients; a target does not supply a reason to press someone to accept treatment or repeated contact.

Required clinical and safety duties remain in force throughout an improvement project. They are not optional stretch targets, and an aggregate score cannot certify that a practice is safe. Starting with a bounded administrative problem gives the team a place to learn the measurement process without redefining those responsibilities.

6. Review the evidence and make a decision

Agree when the team will review progress and what information it needs at that meeting. A brief discussion can cover the latest observation, the next action, the main obstacle and the support required from management. Choose a rhythm suited to the work. If no new observation is available, say so and discuss the outstanding action or support needed instead of presenting an old number as new evidence.

If a definition or target needs to change, record what changed, why and when. Keep the original alongside the revision. Do not quietly lower a target or exclude difficult observations so that the final score looks better. Equally, do not preserve a mistaken target after learning that it is irrelevant or places unreasonable demands on the team.

At the end, ask three separate questions. Did the agreed measure change? What can the available evidence establish about the objective? What should the practice do next? A completed task, a better test result and a useful change in everyday work are related but distinct. State any uncertainty that prevents a wider conclusion.

A percentage of target achievement does not automatically define success or failure. Consider the scope, quality of the evidence, resource use and consequences for other work. Decide whether to keep the change, revise it, gather more information or stop an optional project. The decision should be understandable to the people who will carry it out.

For a first OKR, keep one working page: the objective, result definitions, baseline and target, period, owners and dependencies, practical tasks, quality checks and next review. The value is in the conversation it supports. A team should be able to explain what it is trying to change, how it will recognise progress and what help it needs to act on the evidence.