From Outputs to Outcomes: Building an OKR Framework a Five-Person Non-Profit Can Actually Run
[Measuring What Matters series, Part 2 of 4]
In Part 1 we asked whether the numbers non-profits report are connected to the change they want to create. This post offers one practical way to start answering that question: a worked example of an Objective, three Key Results, and a follow-up method sized for a small team. We offer it as a working draft and we would value your corrections.
Three weeks after the quiet January debrief, Maya and her program lead, Lourdes, sit down with a printout of 847 names.
They have decided to try something small. They will call ten graduates of the financial literacy program and ask what has changed. To avoid calling only the women they remember best, they take roughly every eighty-fifth name on the list.
Each call follows the same five questions, all about the previous month. Did you put any money aside? Where did you keep it? Did you borrow, and from whom? Who decided how the household spent its money? Is there anything from the sessions that you tried and then stopped?
Seven of the ten women answer, a few after a second attempt. Three numbers are out of service.
Of the seven, two describe putting money aside in most months, one of them through the mobile wallet the program introduced. One saved for two months before school fees absorbed it. One attended every session, enjoyed them, and has never opened the wallet app because her husband keeps the phone. Two say the content made sense and that by the end of each month there was nothing left to save.
Seven conversations prove nothing statistically, and Maya says so at the next team meeting. They did show her three things the December report could never have shown. A tool the program teaches may be out of reach for women who share one handset. A predictable expense such as school fees can undo a new savings habit within weeks. And a contact list loses a meaningful share of its usefulness within a year, which matters for anyone planning a follow-up.
Maya is an illustrative character introduced in Post 1. The conversations above are imagined, to show what a small follow-up can surface.
What Maya and Lourdes did that afternoon was informal and unrepeatable. They had no agreed definition of success to measure against, no baseline to compare with, and no plan for what they would do with the answers. This post is about turning an afternoon like theirs into a habit that a five-person team can sustain, using the structure of Objectives and Key Results (OKRs) as a starting point.
The question we want to work through is a practical one: what is the smallest, honest measurement system that connects a program to the change it exists to create?
Where OKRs Come From, and What We Borrow
OKRs trace back to Andy Grove’s management practice at Intel in the 1970s. John Doerr, who worked there as an engineer, later carried the method to Google and many other companies, and described it in his 2018 book Measure What Matters. The book includes accounts from outside the corporate world, among them the Gates Foundation and the ONE Campaign, which is one reason the method has attracted interest in the social sector.
We should be candid about the fit. OKRs were developed for settings where results arrive within a quarter and success can usually be expressed as a number. Social change often arrives slowly and is contested. A framework built for one setting needs translation before it serves another, and it can produce either useful discipline or demoralising paperwork depending on how it is applied.
For that reason we borrow only the core structure, and leave the rest of the corporate machinery aside:
An Objective states, in plain language, the change we want to see.
Key Results (two to four per Objective) state the specific signals that would tell us we are getting there.
The value of the pairing is that it forces the second question, how will we know?, to be answered in writing, before the program runs.
Three Tests for a Good Key Result
Most first drafts of Key Results turn out to be activities in disguise. “Train 800 women” reads like a result and describes what the team will do. These three tests help catch that.
1. The “nothing changed” test. Imagine you hit this Key Result exactly. Could the women’s lives still be unchanged? If yes, the Key Result describes an output. “847 women trained” passes the test for delivery and fails it for change.
2. The “could we observe it?” test. Can someone with a phone, a short list of questions, and an afternoon find out whether this happened? A Key Result such as “women are more empowered” cannot be observed as written. “Women report making a major household money decision themselves or jointly” can be.
3. The “would it change our plans?” test. If the result came back much lower or higher than expected, would the team do something different? Measurement that cannot change a decision is expensive reporting. This test echoes one of the CART principles, discussed below.
A Worked Example to map output to outcomes: Maya’s Program
Let us build a framework for Maya’s program step by step. Every number below is invented for illustration. Your own targets should be set after you have a baseline, and we say more on that shortly.
Step 1: Write the change in one sentence. This is Question 1 from Post 1: if this program works, what would be different for the women it serves? Maya and Lourdes settle on: graduates take more control of their household finances within six months of finishing the program.
Step 2: Turn it into an Objective. An Objective should be short enough to remember and specific enough to argue with. For example: Women who complete our program take more control of their household money within six months.
Step 3: Choose two to four Key Results that would show it. Each one needs a baseline (where women start), a follow-up point, and a method.
Key Result | Baseline source | Follow-up method | Illustrative target |
KR1 (saving behaviour): Share of sampled graduates who put money aside in at least four of the previous six months | Five questions added to the existing enrolment form | Phone follow-up at six months, same questions | 35% at enrolment to 55% at six months |
KR2 (high-cost borrowing): Among graduates who had a high-interest loan at enrolment, the share who have repaid it or reduced their borrowing | Same enrolment form | Same phone follow-up | Half of those with such a loan at enrolment |
KR3 (decision-making): Share of sampled graduates who made a significant household money decision, alone or jointly, in the last three months and can describe it | Same enrolment form | Same phone follow-up | 40% at enrolment to 60% at six months |
Step 4: Add a watch measure for who is missing. Post 1 asked, who is not showing up, and why? A framework that averages across everyone can look healthy while the women furthest from the training centre, or without their own phone, quietly drop away. We suggest one additional line that is tracked but carries no target: completion and follow-up reachability, broken down by village and by whether the woman has her own handset. In Maya’s case, the three disconnected numbers in her first ten calls would have shown up here.
Step 5: Decide in advance what each result would lead you to do. For example: if KR1 rises but KR3 does not, the team might ask whether savings are rising because women are earning more, or because they are skipping other spending, and examine the sessions on household negotiation. Writing these “if this, then we look at that” notes in advance is what makes the numbers usable for decisions.
Notice that the three Key Results measure behaviour reported by women, six months after the program. They sit alongside the attendance figures in the donor report, which remain useful for planning. They add the tier of information that Post 1 argued should drive program decisions.
Right-Sized Evidence: What a Small Team Can Run
Measuring outcomes does not require a research department, but it does require honest decisions about how much evidence is enough. Mary Kay Gugerty and Dean Karlan, in The Goldilocks Challenge (Oxford University Press, 2018), offer four principles for building what they call “right-fit” data systems. They call them CART:
Credible: the data is collected and analysed carefully enough to be believed.
Actionable: the organisation commits to using the data to make decisions.
Responsible: the cost of collecting the data is in proportion to the value of what it will tell you.
Transportable: the findings generate lessons that others can use.
Their guidance includes an important caution for small organisations: sometimes a full impact evaluation is out of reach or premature, because the sample is too small or there is no sensible comparison group. In those cases, strong monitoring that supports learning and improvement is the right target. We find the principles a helpful discipline when deciding what goes into a Key Results table and what stays out.
A realistic measurement routine
One routine a team of five could run for a program like Maya’s is as follows. This is a starting point, and the assumptions are ours.
At enrolment, add five short questions to the existing form (the same questions as KR1 to KR3) and ask for consent to a follow-up call. Record a second contact number, such as a relative or a community group leader, if the participant is comfortable giving one.
At six months, draw a random sample of graduates from each cohort. Random selection (every nth name on the list, for example) avoids calling only the most reachable or most enthusiastic women.
Make the calls using staff who did not deliver the training. Women may find it harder to describe struggles to the person who taught them. Ask about specific recent behaviour, such as “In the last month, did you put any money aside?” Questions like “Do you feel more confident?” invite the answer the caller seems to want, while specific questions are easier to answer accurately.
Follow-up assists to ensure output translates to desired outcome Record who could not be reached, and treat that as a finding. Non-response can hide the results that matter most.
Review the results together in a short team meeting, using the “what would we do differently?” notes written in Step 5.
What this costs, and what it can tell you
A sample of 40 graduates per cohort, at around fifteen minutes per completed call, comes to about ten hours of calling. If reaching each woman takes two attempts on average, plan on about twenty hours, or two to three working days, per cohort. These are our planning assumptions, and your costs will differ with geography, language, and call charges.
It is equally important to be clear about what 40 responses can tell you. Under simple random sampling and a high response rate, a result of 50% from 40 people could plausibly sit anywhere between about 35% and 65%. That width is a reason to treat a single cohort’s figures as a signal worth discussing, and to watch whether the pattern holds across cohorts. It also explains why we suggest setting targets only after you have a baseline: a target of 55% means little until you know how much the baseline itself wobbles.
Two further practicalities deserve attention. First, collecting follow-up information means holding personal data, so check what your local data protection law requires. In Maya’s case that is the Philippines’ Data Privacy Act of 2012. Second, plan how you will store and protect the responses before you collect them. Part 3 of our Technology Without Hype series covers small-scale data habits in more detail.
Why Key Results Should Reach Beyond Test Scores
It might seem simpler to measure whether women learned the material, using a short quiz at the end of the course. Knowledge is easier to measure than behaviour, and it is worth measuring. The research on financial education suggests that it should not be the only thing measured.
The evidence here is contested, which makes it a useful illustration. A 2014 meta-analysis by Daniel Fernandes, John Lynch and Richard Netemeyer in Management Science reviewed 168 papers covering 201 studies. It found that interventions to improve financial literacy explained only 0.1% of the variance in the financial behaviours studied, with weaker effects in low-income samples, and that effects faded over time. The authors suggested a narrower role for education delivered “just in time”, close to the moment people act on it.
A later meta-analysis by Tim Kaiser, Annamaria Lusardi, Lukas Menkhoff and Carly Urban, published in the Journal of Financial Economics in 2022, focused on 76 randomised experiments with more than 160,000 participants. It found that financial education programmes have, on average, positive causal effects on both financial knowledge and downstream behaviours, and that these effects were at least three times as large as the average documented in earlier work.
Reasonable researchers read the same literature and draw different emphases, partly because they looked at different bodies of studies and asked different questions. For a program like Maya’s, the practical lesson is that neither review can tell her whether her sessions, in her three provinces, are changing what women do with their money. Only her own follow-up can. It also suggests that Key Results should reach the behaviours the program hopes to influence, and that the timing of follow-up deserves thought.
What You Can Honestly Say at the End
A natural worry about small-scale measurement is that it cannot prove the program caused anything. That worry is fair. Without a comparison group, a rise in savings could reflect a growing local economy, a new wallet product, or a hundred other influences.
A defensible claim acknowledges this. Here is the kind of sentence we would encourage in a donor report:
Among a random sample of 40 graduates from the 2026 cohort, followed up at six months, 52% reported putting money aside in at least four of the previous six months, compared with 36% at enrolment. We did not use a comparison group, so we cannot say the program caused this change. We can say the pattern is consistent with our theory of change, that 11 of the 40 women we tried to reach could not be contacted, and that the most common reason given for not saving was household expenses at the start of the school year.
(The figures in this example are invented.)
This is the spirit of contribution analysis, an approach developed by evaluator John Mayne that asks whether the evidence supports a credible story of how a program contributed to a change, and what other explanations have been considered. A claim written this way gives funders more to work with, because it shows the team understands what its own data can and cannot support.
Running It as a Rhythm
A set of Key Results is useful when the team looks at it on a regular schedule. We suggest three habits:
Review quarterly. Fit a one-hour conversation into an existing meeting. For each Key Result, ask where we are, what surprised us, and what we will adjust.
Score for learning. Practitioners of OKRs commonly advise keeping the scores separate from individual performance reviews and pay, because people who fear being judged on a number tend to set cautious targets. In a small team where everyone knows everyone, this separation matters even more.
Revise the Key Results themselves. If a measure turns out to be unanswerable, misleading, or no longer relevant, change it and note why. A framework that never changes is probably not being used.
In Post 4 we will look at how boards can build this review into their governance calendar.
A Note for Funders and CSR Teams
If you receive outcome reports from grantees, a few questions help you read them well and signal that candour is welcome:
How many people were invited to be followed up, and how many responded? A strong result from a small and self-selected group means less than a modest result from a random one.
How were they chosen, and who made the calls?
What did the team learn that surprised them? A report without a single surprise may be telling you what the team expected to find.
Which participants are hardest to reach, and what is being done about it?
When a grantee reports a result below its target together with a clear explanation of what it will change, that is evidence of an organisation that is learning. We encourage funders to say so explicitly, because grantees take their cues from how earlier honest reports were received.
Where We Are Still Working This Out
We want to name the limits of this approach, since several are real.
Goodhart’s Law applies. The economist Charles Goodhart is associated with the observation that a measure tends to lose its value once it becomes a target. If staff feel pressure to hit KR1, the follow-up calls may start to nudge answers. Random sampling, independent callers, and a team culture that welcomes bad news all help, though none eliminates the risk.
Not everything that matters can be counted. Trust, dignity, and women’s sense of voice in their households are central to many programs, and a short phone survey captures only part of them. Qualitative stories and the judgement of experienced staff remain important sources of evidence.
The vocabulary may travel poorly. In some organisational cultures across Asia, “Objectives” and “Key Results” sound imported and hierarchical. The underlying logic of naming the change you want and the signals you will look for can be expressed in many other ways, and we would like to hear which phrasing works in your setting.
Funder-mandated indicators sometimes conflict with your own. Many organisations report to several funders at once, each with its own indicator list. We have not yet worked out the best way to reconcile these with a small set of your own Key Results, and it may merit a post of its own.
We have tested this example on paper. The framework above draws on published guidance and on our experience with organisations in the region. We have not yet seen how it holds up across many different programs, and we would welcome your experience.
Three Things to Try This Month
For non-profit teams: pick your largest program and write one Objective and no more than three Key Results, then run the three tests above on each one. If a Key Result fails the “nothing changed” test, rewrite it. Aim to finish in a single one-hour team conversation.
For CSR and foundation partners: at your next grantee check-in, ask one question from the list above, such as what surprised you in this year’s data? Pay attention to how the answer is received on your side as well.
For board members: ask your executive director to share one Key Result for the largest program at your next meeting, including how it will be measured and what the team would do if it came back low.
Whichever applies to you, we would be glad to hear how it went. You are welcome to share your draft Objective and Key Results with us, anonymised if you prefer, and we will respond with questions and suggestions. You can also tell us in the comments where the framework felt useful and where it did not fit your context.
Next in this series (Part 3): The 12-Month Problem, on what can honestly be measured inside a one-year grant, and what funders and non-profits can do about it.
About This Series: Measuring What Matters
This is the second post in a four-part series exploring how non-profits and their corporate partners in Asia can build better impact measurement practice: practical, proportionate, and honest about the real constraints organisations face. Each post includes frameworks you can use, questions worth asking in your own context, and examples from the region.
Part 2: From Outputs to Outcomes: Building an OKR Framework a Five-Person Non-Profit Can Actually Run
Part 3: The 12-Month Problem (working title)
Part 4: Making the Board Meeting Count (working title)
About Huse Infinity
Huse Infinity works with non-profit organisations, social enterprises, and corporate partners across Asia on strategy, governance, and capability building. If you would like to talk through what outcome measurement could look like for your organisation, we would welcome the conversation.
Sources referenced
Doerr, J. (2018). Measure What Matters: How Google, Bono, and the Gates Foundation Rock the World with OKRs. Portfolio/Penguin.
Gugerty, M. K., & Karlan, D. (2018). The Goldilocks Challenge: Right-Fit Evidence for the Social Sector. Oxford University Press.
Fernandes, D., Lynch, J. G., & Netemeyer, R. G. (2014). Financial literacy, financial education, and downstream financial behaviors. Management Science, 60(8), 1861-1883.
Kaiser, T., Lusardi, A., Menkhoff, L., & Urban, C. (2022). Financial education affects financial knowledge and downstream behaviors. Journal of Financial Economics, 145(2), 255-272.
Mayne, J. (2001). Addressing attribution through contribution analysis: using performance measures sensibly. Canadian Journal of Program Evaluation, 16(1), 1-24.
