This is the second post in IPA's series on cost-effectiveness analysis. The first blog covered which costing framework to use for different decision contexts. This post goes a level deeper, into how to structure the analysis once you've chosen your approach. 

What happens when two organizations conduct cost-effectiveness analyses (CEAs) of similar programs and reach different conclusions? One estimates a cost per outcome of USD 400, the other estimates USD 1,200. Are the programs actually that different, or did the organizations use different methodologies? 

In many cases, it is the latter. One team might have valued donated inputs based on what they would have cost to buy; the other might not. One might assume impacts continue over a lifetime; the other might cap them at five years. A CEA is a sequence of methodological decisions, each of which can meaningfully shift the final cost-per-outcome figure.

IPA has spent years working through these methodological decisions alongside partners. This blog post walks through one way to structure this process, drawing on IPA's experience with leading organizations in the cost-effectiveness field. We offer it as a starting point for teams navigating these decisions for the first time, for those looking to make their existing approach more rigorous, and to those relying on CEA to inform funding or program decisions.

Start with the question you are trying to answer

Before collecting any data, ask two questions: Who will use this cost-effectiveness estimate, and what decisions will they make with it? A funder comparing programs across a portfolio needs something different from an implementer planning a scale-up, and both need something different from a government deciding whether to adopt an intervention. 

In our work with  Rwanda's Ministry of Education  and gui2de on the Supporting Teacher Achievement in Rwandan Schools (STARS) program, we built the analysis from the outset around a single government planning question: what would it cost to sustain the program at national scale within existing Ministry structures? This framing determined which costs to include, how to present results, and how to separate research costs from implementation costs.

As the first blog post in this series outlines, several costing frameworks are suited  for different decision contexts, including cost-effectiveness analysis, social return on investment, and cost-effectiveness modeling. The choices made throughout an analysis—including which costs to count, which outcomes to measure, and how long to project benefits—should follow from the central question:  who will use the analysis output, and what decisions will it inform. It is also  important to assess how the results change when key assumptions or inputs are uncertain—a process, called cost-sensitivity analysis. 

Be honest about what your costs actually represent

The most useful question for the cost side of a CEA is: If someone tried to replicate this program tomorrow, what would it actually cost them? That question surfaces the decisions that most commonly drive variation across cost estimates:

Collect costs as you go: The most reliable cost data comes from tracking costs as the program is implemented. Reconstructing costs afterward from historical records introduces gaps and inconsistencies that compound through the rest of the analysis. In IPA's work with Ghana's Ministry of Education on the Strengthening Teacher Accountability to Reach All Students (STARS) differentiated learning program, we collected costs in real time using a structured monthly template. By the end of the project, staff had turned over and some expenditure records were no longer easily accessible, showing  that collecting this information retrospectively would have left significant gaps.  

Break costs into categories: IPA uses the ingredients method, which maps every line item cost to a specific cost category, such as administration, staffing, materials, transport, training, and targeting. Disaggregating costs  this way reveals what actually drives cost-effectiveness, makes sensitivity analysis possible, and allows readers to assess whether the cost structure would hold in a different context. 

Decide what counts as a cost: A few consistent, useful practices have emerged among organizations conducting CEAs:

• Count donated goods and volunteer time based on what they would normally cost, since these resources may not be free when the program expands. Separate money or goods given directly to recipients from the costs of delivering the program. This allows for a clearer view of what drives its overall cost. 

• Count what the program adds, not costs that would exist anyway. For example, in the Ghana STARS project, teachers and head teachers were paid their regular salaries whether or not they participated in the program. But the team counted the time they spent in training as a cost because they could have been doing their regular work during that time (the opportunity cost). 

  Each of these choices changes the cost figure. A structured decision framework can make each of these choices explicit and transparent. By documenting decision rationale, readers can interpret the number correctly and consider how to translate it to their context.

  Know how far to trust your impact estimates

  The effectiveness side of a CEA involves at least as many decisions as the cost side, and getting them wrong can fundamentally misrepresent a program's value. The key question is: what outcome should we measure, how much of that outcome can we expect in practice, and how long will it last? Three factors drive a significant portion of the variance across published analyses:

  Choose the appropriate outcome measure: For economic programs, many organizations use  household consumption because  it is the standard measure of material well-being and avoids the double-counting problems that arise when productive assets and savings are counted separately. For example, a household that receives livestock or builds up savings will typically use those assets to finance future consumption, so counting them independently inflates the estimate. Some organizations use income instead because it is easier to measure. Either way, documenting which approach is used, and why, gives readers a reference point for interpreting the result. 

  Assess how well the evidence from one context is likely to hold in another: The people reached, how well the program is delivered, and how much of the program people get in a pilot setting often differ from what happens when the program is implemented at scale. IPA has used different scenarios to estimate how results might change when programs are scaled or adapted, including changes in implementation quality, coaching intensity, the type of assets provided, and the number of people reached. Drawing on how other organizations have handled these adjustments can help analysts choose reasonable assumptions. In IPA and gui2de’s work with Rwanda’s Ministry of Education on the STARS program, for example, we tested two possible ways of expanding the program: extending it to secondary schools and shifting school visits to sector education inspectors. Each scenario produced a different cost-effectiveness estimate, helping the Ministry understand the trade-offs involved in each approach.

  Consider how long the benefits are likely to last: Most graduation programs, following Banerjee et al. (2015), assume lifetime impacts based on the Permanent Income Hypothesis; improvements in long-run economic capacity translate into permanent consumption gains. The right assumption depends on the program, the population, and the strength of the evidence. Showing what the estimate looks like under a range of time horizons allows decision-makers to see directly how much that single choice drives the result. Analysts also typically apply a discount rate to future benefits, reflecting the assumption that value received in the future is worth less than value received today. The choice of rate can shift the final estimate considerably. 

  A practical decision framework to document your work

  The value of a CEA depends not only on the choices analysts make but on how clearly those choices are documented. Decision-makers who do not understand the assumptions behind an estimate cannot evaluate whether it applies to their context or what would happen if key parameters changed. 

  Working with Village Enterprise, IPA developed a decision framework  that clearly shows all the parameters in a cost-effectiveness analysis, together with examples of how other organizations thought about parameter decisions. This allowed Village Enterprise to develop an internal documented analysis that they could then use to talk to peers and funders about where their assumptions differed.

  The most credible CEAs are the ones that explicitly test how sensitive the result is to the assumptions that are hardest to defend. In practice, this means varying key parameters across pessimistic, neutral, and optimistic scenarios, most often for impact duration, external validity adjustments, and discount rate.

  A decision framework is most useful as both a prompt for critical analysis and a documentation tool. As a prompt for critical analysis, it requires teams to make explicit decisions rather than defaulting to inherited assumptions. As a documentation tool, it gives teams a shared record of every decision made, the rationale behind it, and how comparable organizations have approached the same question. This record is what allows others to interrogate the analysis, adapt it to a different context, or update it as new evidence emerges.

  ---

  On June 23, 2026, IPA, Village Enterprise, and HereWeGrow hosted a webinar for implementers and funders on increasing transparency and comparability in cost-effectiveness measurement, featuring this framework and an applied example of how it has supported conversations between implementers and donors about cost-effectiveness. The recording is available here.