← MethodsSCREEN READING EDITIONDownload the PDF

Would Knowing Change the Decision?

A practical guide to giving an evidence request a useful purpose

For:nonprofit leaders, program teams, boards, funders and commissioners

Use when:someone proposes an evaluation, a survey or a request for evidence before anyone is clear how it will be used

What you leave with:a completed decision-use record with a question, the responses that are feasible, who has authority, the dates and a next check

1

What would the answer change?

Someone proposes an evaluation. The topic matters, the budget is there and colleagues want to know whether the program works. Yet nobody has said what different answers could change.

Before you agree on the scope, name the choice, the people who are able to act and the uncertainty that stands between the responses you could feasibly make. Then ask whether the proposed work can resolve enough of that uncertainty in time.

What you leave with is a decision-use record. It is a short account of the question, the possible findings, the options, who has authority, the calendar and the next check. It helps people agree on useful work without pretending that evidence makes their decision for them.

Choose your route

  1. 1
    A proposal needs an answer today: use the sheet What Would This Work Help Us Decide?, and then the Decision-use record.
  2. 2
    You want the reasoning: read the framework and the renewal case, in the sections from “Which uncertainty matters to this choice?” through “Can the answer reach this decision?” Then read the contrasting case and the limits, in the sections from “When is a focused inquiry worth doing?” through “What if the answer will not change this decision?”
  3. 3
    You are working together: try the practice in the section “Can you improve this request?” Then do the 15-minute exercise and look at the completed example, in the sections from “What can your team clarify in 15 minutes?” through “What does a completed record look like?”

A decision that will not change does not make inquiry useless. Accountability, improvement, public explanation, safeguarding and wider learning can each justify work in their own right. Name the purpose, so that it can shape the question.

2

Which uncertainty matters to this choice?

Start with a verb: renew, change, expand, select, reduce or stop. "Evaluate the program" names a piece of work, not the choice it should inform.

Write downWhat makes the answer usable?
Choice and optionsWhat commitment is still open? Which alternatives are feasible?
The uncertainty that mattersWhich unknown could decide between those alternatives?
AuthorityWho can approve each response? Who should be heard?
Finding and responseWhat specific answer would give you reason to consider a different option?
Capability and calendarCan this work tell the possibilities apart before the decision closes?

The uncertainty that matters is an unknown whose answer could change the choice. If extra places would still be unaffordable even with excellent outcomes, another survey of outcomes will not remove that limit. The real question may be about cost, funding or the size of the expansion.

Keep disagreements about facts apart from disagreements about priorities. Colleagues can agree on how many people were reached and still disagree about whether reaching them should come before reducing cost. More counting cannot settle that judgment.

The framework guides a conversation. It is not a score. If the one person with authority is not available, that can block a proposed response, and four other reassuring answers do not cancel the problem. Record what needs resolving and who owns it. The record itself does not give anyone permission to act.

3

What is still open in the renewal?

Worked case. On 1 October 2026, a foundation considers renewing a workforce service. Its grant ends on 31 January 2027. Staff propose surveying a cohort that is just starting about their employment six months later, to "prove the program works."

The people, places, organizations and figures in this document's examples are fictional.

The foundation board can authorize renewal on 15 January. Papers are due on 4 January. The provider controls changes to delivery within the existing grant, but it needs agreement before it changes the commitments that are funded. A six-month bridge is possible only if the foundation confirms the funding and the provider can keep its staff.

Feasible option to checkWhat the commitment assumes
Two-year renewal at current capacityDelivery and finances can sustain the current service, and the unresolved questions about outcomes have a credible plan for learning.
One-year renewal with a focused changeA serious problem with delivery can be fixed within that year, with an agreed review.
Six-month transition or bridgeContinuity and staffing can be protected while a change, replacement, or exit is arranged.

An additional site is taken out of this renewal choice, because there is no confirmed budget for premises or staff. It stays a separate proposal for expansion.

The board deliberately narrows the immediate question: is the current referral service being delivered reliably, and can any important gap be repaired? Employment outcomes still matter. If an estimate of the effect on employment is essential to the renewal, an audit of delivery cannot stand in for it. In that case, go back to the scope, the timing and the feasible options. Reports of outcomes alone cannot show that the service caused anyone to get a job.

The provider also wants to understand the barriers people face. That question about improvement belongs in the scope, with its own owner and use.

4

What findings could support different responses?

These are planning notes, not observed findings or automatic rules. The board and the provider agree on examples of answers worth telling apart. They also agree to examine information that contradicts them.

Specific possible findingResponse to consider, with reasons
The audit of records confirms that the promised referral steps were delivered, the employer contacts are still usable and the finances support current capacity.Bring a proposal for a two-year renewal at current capacity. Check participants' experience and the unresolved evidence on outcomes before approval. Reliable delivery alone does not show that the service works.
A serious gap in referrals is confirmed, a scheduling problem that can be fixed explains it, and the provider can carry out an agreed fix.Bring a proposal for a one-year renewal with the fix, a named owner and an early review of delivery. Test the explanation before committing to it.
The promised referrals cannot be delivered and no feasible fix is found, or the service cannot keep its staffing.Examine a funded transition or a replacement. Confirm the duties to keep support going and the provider's response. Do not treat an uncertain record as proof of failure.

These notes bring disagreements into the open. One trustee may favor a longer commitment to protect staffing, while another may want a shorter one while outcomes are uncertain. The study can clarify delivery and outcomes. The board must still judge continuity, what else the money could do and how much risk is acceptable.

Suppose the audit cannot tell missing paperwork apart from missed referrals. Then the answer stays unresolved. Seek a limited check, or consider a shorter commitment if that is feasible. Do not turn "unknown" into the most worrying row.

Ask what would weaken each reading. A complete file may hide a referral that participants could not use, and an incomplete file may leave out a referral that did happen.

5

Can this work reach the decision in time?

Calendar. Six months is the period over which the outcome is measured. It is not the whole time needed to produce and use an answer.

RouteEarliest useful dates in this case
Existing cohort, started 1 May 2026The six-month point is 1 November. Follow-up and analysis are planned by 30 November, if access and adequate records are confirmed.
New cohort, starts 1 October 2026The six-month point is 1 April 2027. A report would come later still, after this renewal has closed.
Current renewalThe evidence is reviewed in December, papers are due 4 January, the board acts on 15 January and the grant ends 31 January 2027.

An existing cohort could inform the current renewal. Check who is eligible, who can be reached, what consent or other suitable permissions allow and whether the results can be read credibly. The dates make a route possible. They do not vouch for the data.

The revised scope for now: review the referral records and the staffing commitments, check discrepancies with staff and with a varied set of participants' accounts, and summarize any usable outcomes for the existing cohort, with their coverage and limits. An analyst confirms access and feasibility by 9 October. Review the information that supports the explanations about delivery and the information that challenges them.

Keep the follow-up of the new cohort for a later review. Design any question about effects separately, because reports of outcomes alone cannot show what would have happened without the service.

This scope can shed light on delivery and inform a judgment about renewal. It cannot promise proof of impact. Compare its staff time, its cost and its burden on participants with the alternatives: using existing information, staging the commitment or planning a stronger study for the next cycle.

The calendar in the section “Can the answer reach this decision?” puts these routes and dates together.

6

Can the answer reach this decision?

Renewal milestoneDate in the planImplication
The choice is framed1 October 2026Confirm the actual options, the inquiry and who has authority
The existing 1 May cohort reaches six months1 November 2026It can contribute if usable follow-up and analysis are ready by 30 November
Board papers are due4 January 2027Allow time for checking, interpretation and approval before submission
The board decides15 January 2027Evidence that arrives later cannot inform this meeting
The current grant ends31 January 2027A renewal or a transition that can actually be carried out must be arranged
The new 1 October cohort reaches six months1 April 2027Too late for this renewal decision. It may serve future learning with its own owner

The date the data become available is only one milestone. Confirm, within the calendar, safe access to the source, completion, analysis, review and the delivery of a usable answer.

7

When is a focused inquiry worth doing?

A contrasting case. A community advice service can run one weekly evening session or one Saturday session for the next eight weeks. The manager has authority to change the timetable. Premises and staffing work for either option, and the schedule must be agreed in three weeks.

The uncertainty that matters is when people seeking advice can realistically attend. A general satisfaction survey would not answer it well. A focused inquiry can ask about times that work, travel, jobs and caring responsibilities.

Invite people from across recent referrals, including those who did not attend. Offer more than one way to answer. Check who the contact list covers and compare patterns of response, because people who already attend may favor the current timetable. Use conversations to understand barriers, without presenting a small chosen group as an estimate for everyone.

Possible findingFeasible response to consider
Several groups of referrals describe Saturdays that work for them and concrete barriers in the evening, and the records and responses support that account.Try Saturday sessions for eight weeks, while checking who loses access.
Evening access remains important and Saturday barriers are substantial.Retain evenings and address the identified barriers where feasible.
Information stays thin, or important groups disagree.Keep the current schedule for a short time, or test a rotating arrangement if staffing allows. Say what the trial will check.

The manager sets aside a week to gather information and a week to weigh it with colleagues. The cost and burden are modest compared with choosing a timetable people cannot use. The trial can be reversed, and it will track attendance and barriers to access.

Inquiry earns its place here because it can inform a choice that is open and authorized. The pattern must still be judged. A survey does not pick the timetable by vote.

8

What should we consult before commissioning new collection?

Look at the closest useful evidence you already have: the actual operating sources, relevant evaluations or reviews, and the people who know how the service works. Record which sources you looked at, their date, population, service and outcome, and the differences from this choice that matter. A list of references is not an appraisal of the evidence.

Suppose an earlier review of access already looked at barriers on weekdays. It may help narrow the current inquiry. If the new choice changes transport, opening hours or the people reached, the older finding may leave a question that matters open. On the other hand, if the existing record answers the current question about delivery credibly, another broad survey may add little. Neither a large body of research nor a gap in it decides on its own whether this next study is worthwhile.

Keep two tests apart: does existing evidence support this answer here? and would resolving the remaining uncertainty change a response you could feasibly make? Then compare capability, timing, cost, burden and delay. If the proposed work mainly serves future learning, give that learning a real owner, a product and a point to review it. Do not invent a decision that needs it now just to justify the work.

For choosing the next commitment on the evidence you have now, How Much Evidence Is Enough? works through the comparison of acting, inquiring, waiting and staging. This guide sets the scope of what proposed evidence work could usefully answer.

9

If several useful requests compete

Handling one request at a time stays the default. Use this optional step when worthwhile requests collect the same data twice, depend on an earlier answer or compete for the same paid people and sources. Three useful questions do not automatically make three pieces of work that can all be commissioned.

Put the requests side by side in your existing planning record. For each one, keep its question, its user, its duty or other purpose, the last date its answer is still useful and who actually has authority. Check whether requests that look like duplicates need the same population, definition and period. Share a permitted extract or collection only where those needs agree. Sharing data does not make the claims interchangeable.

Identify the answer that has to come first. Before comparing how well referrals succeed, for example, settle whether a row is a person, a referral or an attempt at contact, and what the recorded event means. Count the cost of that check once if the same version serves both requests. Then show the remaining effort for preparation, collection, interpretation and review, including the staff or participant work it pushes aside. Confirm the time people actually have and the permissions, rather than assuming that a colleague named in two scopes has twice the capacity.

A separate example of coordination, not the renewal case: one analyst has 24 protected hours, which can be used flexibly over four weeks. A shared check of what each referral row counts and how it is coded needs four hours. A required fair account for a grant needs eight more hours by week four. An inquiry into opening hours needs twelve more hours before a decision on the timetable in week six. Both use the same permitted referral extract. A design for future learning about outcomes needs sixteen hours, with a planning review in week eight. All of them are worthwhile, and none of them shows a causal effect just by using the extract.

A sound first order is the shared check, then the grant account and the inquiry into the timetable: 4 + 8 + 12 = 24 hours. Preparing the same extract twice would add four hours that are not needed. The design for future learning is put off, and its learning lead checks the funded time at the week-four review before promising to deliver in week eight. This keeps its purpose alive. Putting it off is not a verdict that future learning has no value. Before adopting a real order, the commissioner must confirm the stated reporting duty, how much freedom there is over deadlines, the resources and the people whose contributions are affected.

Record the reasons for choosing and for putting off each request, what each depends on and what continues. Then name a person and a review date or trigger. Reopen the order when a deadline, a source, a permission or the available time changes. The practice on setting a purpose asks you to make and revise this choice on your own.

For a longer review of a whole set of requests, use Manual 03, When Not to Measure. Its Tool A lists uses. Tool B weighs what information is worth for an open decision. Tool C looks at existing evidence and whether it transfers. Tool D covers how each request is handled, the commitments and when to revisit. Its workbook gathers those tools. Do not force probabilities or dollar values onto accountability, protection or future understanding just to complete Tool B.

The prompts on coordination adapt the undated learning-agenda guide from GSA OES, physical pages 7–8 and 27–29, on questions that must come first, shared work, ownership and a feasible order. Its federal offices, planning cycles and duties do not become requirements for nonprofits.

10

What if the answer will not change this decision?

First check why. The choice may already be settled, the answer may arrive too late, or the same duty may limit every feasible option. A report can still serve another legitimate purpose.

PurposeA question and an owner to name
Accountability or public explanationWhat fair account is owed for the resources and results? Who must approve or publish it?
ImprovementWhat can delivery staff change, and what would help them choose?
Wider learning or institutional memoryWhat will future teams need to know? Who will preserve and use the account?
SafeguardingWhat needs prompt attention, and through which responsible route?

Do not hold a safety concern until a scheduled study reports. Equally, do not invent a safeguarding duty from an ordinary result that is hard to read. Use the organization's actual responsibilities, and specialist advice where it is needed.

How an answer could be used and whether a method can produce it are separate questions. A clear table linking findings to responses cannot turn a satisfaction survey into an impact evaluation. A specialist in methods should check whether the design, measures, coverage, comparison and analysis can answer the stated question. Systematic qualitative inquiry may explain how processes work and how change happens, but chosen positive stories alone cannot do that job.

Waiting has a cost too. Delay may interrupt services or lose staff, while committing too early may lock in an arrangement that does not suit. Compare the likely consequences of inquiry and delay with the alternatives. A staged commitment can keep learning going, but only when funding, staffing, duties and authority make it possible. A simple high or low score does not settle that balance.

11

Can you improve this request?

Practice. Try it before you read the response. A youth service has a fixed grant for the next year. Its manager can change the opening hours by 20 November. Only the board can approve extra staffing. Someone proposes a twelve-month follow-up on wellbeing, asking: "Should we open another evening?"

The partly completed record says:

  • Choice: keep the current hours or open another evening.
  • Authority: the manager, within existing staffing and safety requirements.
  • Unknown: "whether young people benefit."
  • Possible responses and the dates by which they are useful: not yet written.

Revise the unknown, name a useful finding and a response, and give the proposed study an honest purpose.

A reasoned response

First confirm whether another evening is feasible within current staffing. If it is, ask which unmet needs for access that change could meet, who might lose access somewhere else and whether the service can run safely. Current attendance, information on referrals and accounts from a varied group of young people could help before 20 November.

If the barriers to access support another evening and staffing allows it, propose a change to the schedule for a limited time, with a review. If the change would simply move access around or cannot be staffed safely, consider changing an existing session, or keeping the hours while tackling another barrier.

The twelve-month follow-up might serve longer-term learning about outcomes. Its findings for a new cohort cannot inform this November's choice. Asking the board for more staffing is a separate route with its own authority, not something the manager can approve.

Keeping the hours is sound when the limits call for it. So is a limited trial, when it is feasible. "Cancel all evaluation" loses legitimate longer-term learning, and "wait for the follow-up" does not meet the deadline for the timetable.

12

What kind of disagreement are we having?

DisagreementAppropriate next response
What the source actually recordsFind the definitions and look at the relevant source
Which explanation the evidence supportsCompare the competing accounts, and choose a check that tells them apart
What the institution values or can acceptSet out the priorities and trade-offs plainly with the people responsible
Who can authorize the actionConfirm actual responsibility and permission
Whether this inquiry can answer in timeTest what the method can do, and the full calendar

Another study cannot automatically settle a disagreement about values or authority. Keep unresolved answers apart from a finding of failure.

13

What can your team clarify in 15 minutes?

Use one real proposal for a survey, a study, a dashboard or a reporting request. Work from a suitable summary, and do not pass around files that identify participants for this exercise.

Minutes 0-3: Name the choice. Write what could change, the feasible options, who can authorize each one and the last date an answer is still useful. Confirm what is truly open. If the person with authority is not there, record the question to take to them.

Minutes 3-7: Imagine specific answers. Each colleague writes two plausible findings that differ in ways that matter, and what each would give reason to do. Avoid labels such as "good" or "mixed." Include an unresolved answer: what if the work cannot tell the important possibilities apart?

Minutes 7-10: Find the disagreement. Compare the notes. Is it about facts, priorities, authority or feasibility? For example, a disputed count of attendance needs checking, and a preference for continuity needs discussion. A limit on funding needs a conversation with the people who have authority.

Minutes 10-13: Check the proposed work. Can it tell apart the findings that matter? Put the date the data are ready, the analysis, the papers and the meeting of the person with authority next to the deadline. Consider the cost, staff time, privacy, people's ability to refuse and the burden on the people asked to respond.

Minutes 13-15: Record the next check. Choose a limited action, an owner and a point to review it. As the situation calls for, use existing information, revise the question, commission a focused inquiry, stage a commitment or plan for the next cycle. Fifteen minutes clarifies the scope. It does not prove the study is sound.

The funder and the provider may need to learn different things. Record both purposes and who will see the findings. People with different powers do not need to agree on every response for the record to be useful.

14

What does a completed record look like?

This example sums up the sections from “What is still open in the renewal?” through “Can the answer reach this decision?” Use the Decision-use record for your own case. Its fields match this example.

FieldCompleted example
Purpose and choiceInform the renewal of the workforce service, and support a fair account and learning about delivery.
Authority and people affectedThe foundation board authorizes funding, and the provider manages delivery. Hear the views of participants and staff.
Feasible optionsA two-year renewal at current capacity, one year with a fix, or a funded six-month transition if that is workable. No new site is part of this choice.
The uncertainty that mattersAre the referral commitments delivered reliably, and can a confirmed serious gap be fixed?
Specific findings and responsesIf delivery is reliable and the finances are workable, consider two years. If there is a gap that can be fixed, consider one year with a review. If the gap cannot be fixed, examine a transition. These are still proposals that need judgment.
Unresolved answerIf missing files cannot be told apart from missed referrals, check the discrepancies and assess the feasible ways to keep support going.
Information and capabilityAudit the records, look into varied accounts and competing explanations, and summarize outcomes for the eligible cohort. This scope does not show an effect on employment.
CalendarConfirm feasibility by 9 October. The existing cohort reaches six months on 1 November, analysis is due 30 November, papers 4 January and the board meets 15 January. The new cohort's point of 1 April is too late.
Cost, burden and delayThe analyst scopes the staff time, the cost and the contact with participants. The board considers continuity, staffing and the consequences of delay. There is no agreed cost estimate yet.
Next check and reviewThe foundation officer and the provider lead confirm access, dates, funding options and scope by 9 October. Revise the record then. This exercise alone does not authorize any study.

The record should keep important disagreement on it, rather than manufacture agreement.

15

What should happen after this first pass?

Take the record to the right people. Confirm the open choice, the options, the dates and the resources before you commission work. Arrange help with methods where it is unclear whether a design can answer the question. An independent review needs real independence, and drafting the scope is a different role.

Record who approves the inquiry, who can see its findings and when the decision-use record will be reviewed. Go back to it when the timing, costs, options or question change. Keep a dated account of the reasons for the final decision, including what the information did and did not establish.

Recall it at the next meeting

Before you come back to this guide, ask: what is the difference between a useful question and work that can answer it? Then try a new request. If a future choice about staffing is already fixed, what other learning might still justify the work? Check your answer against the section “What if the answer will not change this decision?” This is optional practice, not a test of whether you agree with Prova.

Keep the record useful

Bring the completed record and a suitable summary of the information to the next conversation. Name the unresolved question, the person responsible for checking it and the date for review. Keep in view who in your organization is responsible for approval. Do not send records that identify participants through a public contact route.

16

What supports the approach?

These sources support the methodological distinctions.

  1. 1
    U.S. Government Accountability Office (2012). Designing Evaluations, GAO-12-208G. Printed page 6: intended users, constraints, burden, timeliness, and alternative information. Chapter 3, pages 18–23: connect researchable questions, source choices and permissible conclusions.
  2. 2
    HM Treasury (2026). The Magenta Book, updated 15 May. Sections 1.9-1.10, chapter 2 and section 3.2: purpose, users, proportionality, scoping and method choice. Government guidance is adapted here to nonprofit work.
  3. 3
    Grantmakers for Effective Organizations (2014). Where Should We Start in Using Evaluation as a Tool for Learning?. Sections on shared purpose and evaluation systems: shared learning interests, capacity, existing systems and reporting burden. Professional guidance, not proof of this guide's effects.
  4. 4
    Gertler, Martinez, Premand, Rawlings and Vermeersch (2016). Impact Evaluation in Practice, second edition. Chapter 3: observed outcomes and the counterfactual question. An outcome survey alone does not identify an effect.

Notes on method and use

Six calendar months from 1 May is 1 November, and from 1 October it is 1 April. Analysis and consideration take extra time. No response threshold, probability, expected-value calculation or automatic funding rule is implied.

Follow your organization's responsibilities and approval processes. The sheet and the record use this guide's definitions.