Lesson 2: Choose a context that fairly tests the model
Another common approach to site selection is to pick the place most likely to succeed. For a model intended to scale, this is misleading. A pilot run in an exceptionally favorable context may succeed for reasons specific to that context, telling you little about whether the model could work elsewhere. Similarly, a pilot in an exceptionally difficult context risks the opposite: failing for reasons unrelated to the model. We recommend aiming for a middle ground. Look for a site with sufficient right conditions for the pilot to test the model fairly, but not so favorable that the lessons fail to transfer.
For our MBS model, that means a context where markets can function: strong enough to give the model a fair chance, but not so exceptional that success would fail to generalize. Keeping that in mind, we assessed our four candidate countries (Ethiopia, Kenya, Mozambique, and Zambia) against the necessary conditions. Zambia consistently performed well, but never ranked as the top choice or lowest performer on any of the criteria we had reviewed, which made it the right choice for our project.
- Sound policies and clear government leadership. Sanitation markets are shaped by the regulatory environment in which they operate. We looked for governments that treat sanitation as a national priority, set clear targets, and create regulatory space for private-sector participation. Zambia performed well (though not as well as Ethiopia). Rural sanitation is anchored in the National Rural Water Supply and Sanitation Programme. Recent reforms have expanded the mandate of commercial utilities to deliver rural sanitation services, and the Ministry of Water Development and Sanitation has actively engaged in market-based approaches. This provided an institutional infrastructure for scale that the partnership could build on.
- Functioning market infrastructure. For a sanitation program to scale through markets, physical and economic infrastructure must be sufficient for small businesses to operate. This includes road networks that move materials, supply chains that source and deliver sanitation products affordably, and a business environment that allows small enterprises to register, operate, and access finance. Our desk-based review identified Zambia as a relatively conducive environment across these dimensions (but not as conducive as Kenya). Zambia ranks well on indicators of business formalization and access to credit, and rural transport corridors – particularly in Southern Province, where districts such as Mazabuka, Monze, Choma, and Kalomo lie along key transport routes – enable the movement of materials and services.
- Household demand and purchasing power. Markets cannot scale without genuine demand. Specifically for sanitation, we must ask: Are households able and willing to invest? Do sanitation coverage levels, open defecation rates, and purchasing power suggest real demand? Zambia showed a useful pattern: low rural open defecation alongside heavy reliance on unimproved facilities. After years of Community-Led Total Sanitation (CLTS) programming, households already have demonstrated demand by moving off open defecation and into basic self-built latrines. However, these basic latrines often deteriorate quickly. MBS builds on this existing demand and offers households an upgrade from the latrines they have already built to more durable and desirable improved latrines.
- Operational readiness of implementing partners. Even a well-designed model in a well-suited context will not scale if implementing organizations cannot carry it. We considered country-office capacity, Savings Groups network strength, prior market-based programming, and strategic alignment for each partner. IDinsight’s enabling environment review pointed to Kenya and Zambia. CARE’s country-office readiness review ranked Ethiopia first and Zambia second. iDE’s preferences leaned toward Mozambique and Zambia, partly reflecting where its existing sanitation market development work could complement a joint effort. While no single country topped all dimensions, Zambia ranked consistently well across the required dimensions, making it our top candidate.
Why this matters
For an organization whose goal is scale, the choice of test site determines what the pilot can teach. A site that is unusually favorable can make a weak model look strong; a site that is unusually difficult can sink a sound one. Choosing an “average context” and naming in advance the conditions the model depends on is essential for developing the evidence needed to further scale the model. In our case, Zambia did not rank first on any of our assessments, but consistently ranked second against the other three countries in question.
Lesson 3: Don’t end your research with a desk review; verify your assumptions on the ground
A desk-based review can tell you where to begin, but it cannot tell you what you will find when you get there. Country-level indicators capture the broad conditions under which a program meant for scale can be piloted, but rarely the mechanics of how it will behave on the ground. This is why we recommend treating the desk review as a starting point and verifying your assumptions on the ground before piloting.
In our case, once we selected Zambia, we went on the ground to study the existing conditions and to further narrow down our search. The team zoned into Southern Province as a suitable site. It had strong Savings Groups networks centered on Choma and Monze, key transport corridors, and active district WASH structures converging in five priority districts: Mazabuka, Monze, Pemba, Choma, and Kalomo. Within those districts, we narrowed further still, to rural growth centers (RGCs): small market hubs where road access, basic infrastructure, and local economic activity converge. RGCs were rural enough to test the model in conditions representative of where it would need to work for scale in Zambia, but with enough density and infrastructure for market-based solutions to stand a chance.
We conducted on-the-ground assessments to stress-test our priors. This included mixed methods research (e.g., store observations and interviews with entrepreneurs, focus group discussions with Savings Groups, and a quantitative survey with ~500 households), mapping of active Savings Groups and their financial maturity, along with WASH systems assessments conducted with district authorities. Our most important finding was that affordability, not awareness, was the barrier preventing households from investing in improved latrines. The desk review had inferred demand from sanitation coverage gaps, and the ground research confirmed it: 98% of surveyed households strongly agreed that owning an improved latrine matters. But 76% named affordability as the main barrier. This finding validated that our model’s central task is on affordability and access to finance, not demand generation.
Why this matters
A desk review can tell you where the broad conditions for a model exist. It cannot tell you whether the model will actually work there. Only ground research reveals the specifics that decide that: what is really driving or blocking demand, how local systems function, and where the binding constraints sit. Those specifics often reshape the model itself. For any organization working toward scale, ground research is what ensures the model is designed for the context you find, rather than the one you assumed.
What this tells us about scale
The specific ingredients of scale will always differ across models, geographies, and sectors. For us, the evidence pointed to Zambia, and to Southern Province, as the right place to begin – where conditions for success are present, and where lessons could be generalizable to a broader set of contexts. But the more important output of this process was not the decision itself. It was the framework for making that decision: a structured way of asking what conditions are necessary for scale and then systematically testing and interrogating them.
We have started our pilot. Our Baseline data collection is complete, and we are now testing the model in rural growth centers across the Monze and Pemba districts. The insights we generate through this pilot will tell us whether the model can work there. If it doesn’t, we think it’s unlikely to be scalable. We will return in a future post to share what we learned from implementing our pilot, so stay tuned.
