Need, setting and program boundaries · Lesson 3

Turn a need into an objective and commissioning gates

10 min read

Lesson objectives

  • Write a verifiable organizational objective, compare proposals after scope, safety and budget gates, and state evidence that would pause or alter commissioning.
Reading tools

Once the need is framed and affected workers have been heard, the board needs a decision it can fund and revisit. ‘Improve wellbeing’ is a respectable direction but it identifies neither a working condition within the board's control nor an outcome that can reasonably be examined. Write a two-part objective: a nearer change in the organizational system and a more distant result expected to respond, without promising a causal effect in advance. For example: ‘Reduce imposed schedule changes notified less than 24 hours before a shift from 40 to 30 a month within eight weeks, without increasing the reserve team's burden,’ followed by ‘examine whether planning and recovery feel more manageable through a voluntary follow-up with disclosed response coverage.’ Thirty is a site-specific pilot target based on local capacity, not a universal mental-health threshold. If the baseline is unsound or changes were worker-requested, revise the objective rather than ignoring inconvenient data. The board should also know what is not an objective: diagnosing workers, reducing their use of professional services, or deterring reports of distress.

Separate an activity measure from an outcome and a protective measure. Activity could be the number of meetings or schedules published; it tells us what was delivered. A work outcome could be the proportion of all scheduled shifts changed without notice, or median notice time, with a defined denominator. A wellbeing-related outcome might be a voluntary response about capacity to recover. A mean shift across different respondent groups cannot by itself establish the program's effect. A protective measure looks for harm, such as more overtime in the reserve team or more service errors. An improved overall average with worse burden in a small group is not success. Before the pilot, specify each number's source, collection time and permitted viewers. A number that cannot be collected safely is a poor decision condition however attractive it looks in a proposal. A bounded qualitative indicator, such as accounts of schedule predictability, can help when collected in a way that protects contributors.

Use eligibility gates before scoring proposals. Ask in order: is the offer inside the preventive, non-clinical boundary and safe for participants? Does someone have authority to change the proposed work conditions, with resources to gather necessary data? Is the total cost supportable, including follow-up and evaluation rather than only the advertised activity price? A condition that fails cannot be offset by popularity points or a low sticker price. You may redesign an offer outside scope rather than discard every useful idea. For instance, an unqualified provider's therapy package could be replaced by optional general education plus a separate route to qualified services. But do not announce a referral partner until its local availability and conditions are verified. Once proposals pass the gates, compare their relation to the suspected mechanism, participant burden, access equity and testability. Evaluation is not an afterthought if it determines whether funding continues; include its cost from the beginning.

Naming the decision owner prevents an attractive plan that nobody can operate. A wellbeing unit might be able to buy a workshop but unable to alter rosters. Putting roster redesign into its brochure does not confer authority. It needs an operations mandate and an agreed trial schedule, or the objective remains conditional on that mandate. For every gate, specify entry evidence and exit decision. Who confirms the baseline? Who approves a safe participation channel? Who checks the real cost of paid worker time? Who can stop the trial if a privacy or workload problem appears? A rejected or deferred offer then becomes a reasoned decision, not a judgment about the people who proposed it. Do not treat intended participants as free data sources: time spent in meetings or surveys can displace work and needs cover. If the board asks for precise causal proof within two weeks of a small intervention, challenge the expectation instead of inventing an indicator that the design cannot answer.

How do you select a numerical target without false precision? Start with the distribution of the baseline, not a single month's average. Perhaps the 30 short-notice changes occurred during a seasonal event while the three previous months recorded 20, 28 and 25. A target of 24 might then be met without intervention as demand returns to normal. Record prior months, whether changes were requested or imposed, sites and service constraints. Select a pilot threshold useful for a decision, not a public promise. The board might instead say that it seeks fewer imposed changes under 12 hours and will review after eight weeks against a sensible comparison period. If an engine requires a number, present it as a revisable internal estimate before implementation. If ‘imposed’ is redefined during the pilot, recalculate baseline and result or declare them incomparable. An honest comparison matters more than an appealing figure. Examine shifts separately as well: an apparent site-wide improvement could occur only in the day shift while nights remain unchanged.

Make the funding decision reversible. The first tranche buys a small test and better records; the second depends on a pre-stated result. Write down what happens if the near-term target is achieved but the protective measure worsens: pause expansion and revise the mechanism, not announce success. If changes do not fall but workers report clearer notices, do not turn a better message into proof that the risk has gone; wording may improve while the count remains. Keep the option of buying nothing or rescoping an offer in the minutes. The board need not claim prevention has no value because precise causation was not proven in a short pilot. It needs a decision proportionate to evidence and risk, with a review date, owner and evidence that could reverse it. A professional program does not sell a workshop or platform. It explains what should change, why, what can be learned after delivery, and what will remain unknown. Clear limits let managers, workers and funders examine the same reasoning instead of exchanging slogans about wellbeing.

Watch the denominator: changing it can reverse the comparison. In the board case, a count of 40 imposed short-notice changes does not show the risk per shift if schedule volume changes. Suppose 800 shifts were scheduled at baseline: 40/800=5%. In the pilot month there are 30 such changes among only 500 shifts: the count falls but the rate rises to 30/500=6%. A count may answer a question about the scheduling unit's total workload, while the rate may better answer how likely a shift is to be affected. Write the question first, then choose the measure. Preserve count and denominator so readers can interpret the percentage. The definition of ‘short notice’ might also move from 24 to 12 hours, making results look better without a real change. Freeze the definitions of period, scheduled shift, imposed change and notice time in the measurement protocol. If redesign makes voluntary worker-requested swaps easier, do not mix those with imposed changes in one figure that punishes flexibility. All these numbers are fictional teaching inputs, not a rate for any real site.

Read the commissioning language as a worker would, not only as a funder would. ‘Reduce use of support services’ could imply that seeking help is unwelcome, even if the board intended to reduce pressures. Replace it with a work-condition objective while preserving independent, voluntary access to support. Do not ask a supplier to promise that everyone will improve. Ask about the offer's assumptions and limits, how needs beyond its competence are routed, and what data it retains or transfers. Request a fallback if too few people choose to participate or the measurement proves unreliable, not just a price discount. In the decision record, connect every cost to an output: paid worker time to represent shifts, an operational change, baseline and follow-up data, and contingency for a foreseeable problem. If ‘evaluation’ consists only of an exit satisfaction rating, call it user feedback rather than impact measurement. Do not turn a seller's testimonial into proof of effectiveness. These details distinguish buying an activity from buying a chance to learn whether a work redesign merits continuation.

Lesson summary

Commissioning links a testable organizational objective to a more distant wellbeing outcome and applies scope, safety and budget gates before ranking. A limited pilot with protective measures is more defensible than a promise of success.