Theory of change for business development programs: how to explain why your program should work

Key Takeaways

• A theory of change explains how a program's activities are expected to produce results in companies, link by link, with its assumptions written down. It builds on the definition in UNICEF's methodological brief (Rogers, 2014).
• It is the starting point for designing an impact evaluation, and it is built at the start of program design. That is how the Impact Evaluation in Practice handbook by the Inter-American Development Bank (IDB) and the World Bank puts it (Gertler et al., 2017).
• The intermediate links are the most fragile and the least measured. Almost every business training program gets companies to adopt some practices, but the change is modest, and David McKenzie's 2021 meta-analysis estimates real but small effects: around 10% on profits and 5% on sales.
• An adopted practice can be abandoned if it does not pay off for the business. In an experiment with 160 tailors in Accra, Ghana, those who received consulting started keeping records and most stopped later, with no evidence of higher profits (Karlan, Knight and Udry, 2015). The design lesson is to test first whether the practice is profitable for the segment.
• Measuring at the start and at the end is not enough to know which link broke. Without data on the intermediate links, you cannot tell an implementation problem from a design problem.

Every business support program rests on an explanation of why it should work, even if nobody has written it down. "If we give them advice, companies will improve their practices and sell more" is a theory of change. It is almost always incomplete, because it skips what has to happen in between and what is being taken for granted.

If you design, fund, or evaluate programs with companies in Latin America, whether from a multilateral, an economic development department, a chamber of commerce, or a foundation, here is the method in seven steps, an example of a financial management program, and what the evidence says about which links break.

What is a theory of change?

A theory of change is the explanation of how a program's activities lead to the expected results, with the sequence of intermediate changes and the conditions that must hold at each step. In Spanish it circulates as both "teoría de cambio" and "teoría del cambio."

Patricia Rogers's methodological brief for UNICEF (2014) defines it as an explanation of how activities are understood to produce a series of results that contribute to the intended final impacts, and notes that it can be developed for a project, a program, a policy, or an organization. Many representations also show the assumptions: conditions needed for it to work that are outside the implementers' control.

The term spread in the 1990s through the Aspen Institute's work on evaluating complex community initiatives. Isabel Vogel's review for UK aid (DFID, 2012), based on 40 interviews with 25 organizations, found consensus on its minimum elements: the context, the long-term change, the sequence of change, the assumptions, and a diagram with a narrative.

In Colombia, the official language of the National Planning Department (DNP) for formulating and monitoring public investment is the value chain: inputs, activities, products, and results. The term theory of change appears mostly in policy evaluations under the Sinergia system, such as the evaluation of Vive Digital. A practical way to see the relationship: the theory of change is the value chain with the mechanism and the assumptions of each link written down.

What is the difference between a theory of change and a logical framework?

The logical framework matrix summarizes what the program delivers, how it is measured, and what is assumed, in a fixed four-by-four format. The theory of change explains why those deliverables should change how companies behave, and it has no single format.

Vogel notes that the matrix tends to hide the most uncertain stretch, the one from outputs to long-term results. German development cooperation (GIZ) emphasizes the use of outputs as an intermediate outcome to make that stretch visible. In a business program, that means asking whether the company uses what it was given.

It is best to build the theory of change first and then summarize it in the matrix. How to build the matrix is in logical framework for business support programs.

Which links in the theory of change fail most in business programs?

The links that fail most are the intermediate ones, between receiving advice, using it, and having that produce a result in the business, because each involves a decision by the company that the program does not control. McKenzie and Woodruff's review of business training evaluations in developing countries lets you follow each stretch:

• Attending. Even when training is free, on average only about 65% of invited firms take part.
• Adopting. Almost all programs get companies to apply some practices, but the change is modest: 5 to 10 percentage points in typical cases.
• Sustaining. Firms may start a practice and drop it, and the evidence barely captures this: eight of the 13 studies reviewed measured at one year or less.
• Earning. Few studies find significant effects on sales or profits, partly because samples are too small to detect them.

The contrast comes from an experiment with large textile firms in India, each with several plants (Bloom et al., 2013). With five months of intensive consulting, adoption of 38 management practices rose from 25.6% to 63.4% of practices, held after the consultants left, and productivity rose 17% in the first year. There the link held. The same study found that, as information barriers fell, owners' time became a growing constraint: it went from explaining 3.7% of non-adopted practices to 14% within nine months.

Evidence closer to the region also shows that the link can hold when the support is well designed. In Colombia, an experiment with 159 auto parts firms compared individual consulting with consulting in groups of three to eight firms (Iacovone, Maloney and McKenzie, 2022). Both improved management practices by 8 to 10 percentage points, the effect lasted at least a year, and group consulting cost about a third as much as individual consulting. Three to four years later, the firms that received group consulting showed more employment and higher sales. In Mexico, small and medium firms in Puebla that received subsidized consulting improved their productivity and, five years later, had more employees; we cover that study in the indicators guide. And McKenzie's 2021 meta-analysis concludes that traditional training does move profits and sales, with effects too small for most individual studies to detect.

The design lesson is that the adoption link depends on the type of firm, the intensity of the support, and whether the practice really serves that business. A theory of change that takes it for granted is hiding the program's main risk.

How to build a program's theory of change step by step

1. Start with the situation of the companies

Before thinking about activities, describe the problem facing the companies you want to serve, its causes, and its consequences. Rogers recommends starting the theory of change from that situation analysis, and only then deciding which part of the problem the program will tackle.

It is the same analysis as the problem tree in the logical framework: do it once and use it in both documents. Work on it with data and with owners from the segment. If the diagnosis says companies do not grow because they do not keep accounts, also ask why they do not. The answer (lack of time, mistrust, records that serve no concrete purpose) changes the program.

2. Write the final change and the change the program can actually produce

Separate two things. The long-term change is what you hope to see in the business, such as more sales or more jobs, and it depends on many factors. The direct change is what the program expects to see in companies when it ends: a practice adopted, a decision made, a channel working.

The direct change is what will later become the purpose of the logframe matrix and the expected change in the program design. It has to be verifiable at the company.

3. Map the sequence backwards

From the direct change, ask what has to happen just before it, and repeat until you reach the activities. In a business program the sequence usually has links like these, and the exact number depends on the program: the company takes part, receives the deliverable, understands it, uses it, changes a practice, and the practice produces a result.

Working backwards avoids the most common mistake, which is to start from the activities already planned and connect them with arrows to the desired result. If a link has no activity that triggers it, the program is waiting for it to happen on its own.

4. Explain the mechanism behind each arrow

Each arrow is a causal claim and needs an explanation of why it would happen. Rogers distinguishes two parts: how change happens in general and how the intervention triggers it.

"Companies adopt the practice" is not a mechanism. "Companies adopt the practice because during the support they apply it to a real decision and see the result in their own business" is one, and it also tells you what the support has to include.

5. Write down the assumptions and look for evidence on each

Assumptions are the conditions outside the program's control that must hold for each link to lead to the next. Vogel describes them as the real theories inside the theory of change, and warns that making them explicit is the hardest part and requires discussing them with others.

For each assumption, look for existing evidence: studies of similar programs, data from previous editions, or the operator's experience. If the evidence says only two in three invited firms attend, the coverage target and the size of the open call have to reflect it. If nobody knows whether the practice is profitable for that segment, that is the first assumption worth testing.

6. Decide what to measure at each link

Choose one or two indicators per link, especially the intermediate ones: participation, use of the deliverable, adoption of the practice at three and six months. Rogers warns that if intermediate results are not measured and the program fails, you cannot tell where the causal chain broke.

The theory of change tells you what to measure. How to write each indicator with its baseline and technical sheet is in indicators for business support programs.

7. Use the theory of change to read results at closing

At closing, walk the chain with the data. Rogers separates two types of failure, and in practice a third is worth ruling out. If activities did not reach the intended firms, with the planned intensity or as designed, it is an implementation failure and the recommendation is to improve operations. If they were delivered as planned and the change did not come, it is a theory failure and the strategy needs rethinking. Before concluding either, check whether the measurement could detect the change: with few firms or a very early measurement, a real effect may not show.

When there is no comparison group (similar firms that did not receive the program), John Mayne's contribution analysis allows reasonable conclusions with four pieces: a plausible theory of change, evidence that the activities were carried out, evidence that the expected links occurred, and an examination of other factors that may have played a role. The question shifts from how much the program caused to whether it is reasonable to conclude that it made a difference.

Theory of change example: a financial management program for microenterprises

Picture a program that supports microenterprises in keeping their numbers and using them to decide how much inventory to buy. For a committee, the theory of change fits on one page: a diagram like this one and a short narrative explaining each arrow.

Attends the sessions → Has a record set up → Keeps records for three months → Adjusts inventory orders using its numbers → Less idle inventory

The use link is the one to watch most closely. If most firms stopped recording after three months, the program learns in time that it has to review what the tool asks of the owner and what it gives back, before blaming the training. Evidence also helps set the participation target: if on average two in three invited firms attend, the open call has to be larger than the target.

To move this theory of change into the logframe matrix, the participation and deliverable links correspond to the outputs, the use and decision links make up the purpose, and the inventory result goes to the goal, because it depends on factors the program does not control. The assumptions of each link go into the assumptions column of the corresponding row. Without a comparison group, that last link is reported as a contribution, supported by evidence on the earlier ones.

Real case: the Ghanaian tailors who adopted and then abandoned

Between 2008 and 2011, Dean Karlan, Ryan Knight, and Christopher Udry ran an experiment with 160 tailors in Accra, Ghana (Karlan, Knight and Udry, 2015). One group received a year of standard consulting from an international firm, another a grant that doubled their working capital, another both, and another nothing. Almost all were workshops of five people or fewer, and at the start only 17% of the tailors kept written records.

Consulting changed the tailors' practices in the short term, but the change did not last. Within the consulting group, the share keeping records fell from 63% to 28%, and the drop was not because the control group caught up. One year after the consulting ended, the two groups' practices were no longer distinguishable, and the authors found no evidence that consulting increased profits. Their interpretation is that the tailors tried the practices, found they were not profitable in their business, and went back to what they did before. The figures come from the 2014 working paper published by Innovations for Poverty Action.

Read as a theory of change, the case leaves three lessons:

Compared with the India study, the case suggests that the same arrow ("if they adopt, they earn more") can hold in a large textile firm and not in a five-person workshop when the practice does not fit the business. The two studies differ in many ways, so this is less a rule than a reason to write assumptions by segment and test them. In microenterprises, the mechanism has to run through a concrete decision by the owner, as in the inventory example.

Frequently asked questions about theory of change in business programs

Is "teoría de cambio" the same as "teoría del cambio"?

Yes. Both are Spanish translations of theory of change used across Latin America: UNICEF's Spanish brief uses "teoría del cambio," while many funders and evaluators say "teoría de cambio." They refer to the same method.

How is a theory of change used to evaluate a program without a comparison group?

It guides a contribution analysis: you verify that the activities were carried out, that the intermediate links occurred, and that other factors that may have played a role were examined, acknowledging their weight where they had it. It does not tell you precisely how much the program caused, but it lets you argue, with evidence, that it is reasonable to say the program contributed.

Should the funding institution require a theory of change?

It can ask for one, taking care that it does not become just another form. Vogel's review found that many organizations fear that if it becomes mandatory, it will be filled in to comply and lose its analytical value. It works best as a design conversation between funder and operator, revisited with data during implementation.

Who should build a program's theory of change?

The funder and the operator, together, with owners from the target segment. The funder knows the change it needs to report, the operator knows what happens in the sessions, and the owners know which assumptions do not hold in their businesses. A theory of change written by only one of these parties usually leaves the other two's assumptions unexamined.

How often should the theory of change be reviewed?

At least at a midpoint in implementation and at closing, with data on the intermediate links in hand. If an assumption proved false, the program or the theory is adjusted, and the next edition starts from a corrected version. That cycle is what makes the theory of change useful.

A theory of change tells you where to look

A well-built theory of change does not guarantee that the program will work. What it does is say in advance what has to happen in each company, what is being assumed, and where to measure, so that when something fails you know which link it was and what to do differently.

If you are designing a program and want to review its theory of change with someone who has operated similar programs, we can work on it with you in a design session.

Request a design session | How we work with institutions

Read also: How to design a business support program | Logical framework for business support programs