This section is organized around two main findings. First, supply chains proved to be both a useful anchor for wider system modeling and a recurring source of insight into the drivers of food security outcomes. Second, the process of building and using these maps enabled collective sense-making and more coordinated action among stakeholders.

Throughout our engagement in Karamoja, supply chains served not only as a useful entry point for mapping the wider system, but also frequently emerged as a central finding of the analysis, as they were often identified as key constraints on, or enablers of, food security and resilience outcomes. The mapping highlighted flows of agricultural inputs, crops, livestock, food, pharmaceuticals, and other goods and services into and out of the region as critical both to crisis response and to longer-term resilience.

Building outward from individual supply chains to the wider system also made it possible to uncover dynamics shaping household resilience. The Karamoja Beans Market System Map enabled the practitioners to identify interconnected causal loops within the markets (see Fig. 5). For instance, in the supply chain for iron-rich beans, successful aggregation by farmer groups is more likely to attract traders, which in turn encourages more farmers to participate in future seasons. However, without traders, farmers hesitate to engage in collective marketing, and without supply, traders are reluctant to enter the district. The success of this loop depends on the alignment of both elements, but also on the reliable production of iron-rich beans, which requires an adequate supply of inputs, forming another feedback loop between farmers and input dealers. By using these models, USAID partners learned that facilitating positive causal loops is crucial for developing robust supply chains that enhance household resilience.

Feedback loops in the supply chain for iron-rich beans

Modeling the wider system also made it possible to identify synergies across supply chains that would have remained invisible in a siloed value chain analysis. For example, the map showed how the supply chain for iron-rich beans could build on existing actors, infrastructure, and market linkages in more established chains such as sorghum, maize, or sunflower. This included opportunities to piggyback on trader networks, share transportation assets, exploit economies of scope in freight, and align with interventions led by other donors. Extending the analysis beyond an individual supply chain also showed that many of these supply chains depended on common enabling functions, especially transportation, which appeared repeatedly across the conceptual model (Fig. 6). This made clear that transport operated as a system-wide constraint affecting actors both upstream and downstream of the farmers. These effects collectively slowed market development by impeding not only the flow of inputs and outputs, but also the flow of money and market information. By making these interdependencies visible, the map helped identify opportunities for complementary interventions, ranging from transport infrastructure improvements to measures that improve freight market performance and truck utilization across commodities.

Transportation effects highlighted on the Karamoja Beans Market System Map

System models can support action in a direct and relatively conventional way. For instance, our simple system dynamics model of household resilience later informed a broader model for the International Committee of the Red Cross (ICRC) in Northeast Nigeria, integrating household needs and supply chain dynamics to explore the appropriate mix of humanitarian response options, including in-kind assistance, cash assistance, and credit for supply chain actors [36].

But for our Karamoja maps, the contribution lay not only in their prescriptive value, but also in the shared process of building and iterating them. That process created a shared understanding of resilience among a variety of actors and enabled them to reason collectively about how their interventions interacted across the wider system. In these respects, the work directly enabled what USAID referred to as “Collaboration, Learning and Adaptation,” an approach to improving development outcomes through adaptive management. Table 1 summarizes the main ways in which the mapping work proved useful. We elaborate below on several examples of how these contributions materialized through our engagement with the Karamoja Cluster.

In January 2020, we hosted the first workshop of the Karamoja Resilience Cluster with support from the Uganda Learning Activity. The workshop gave participants an opportunity to better understand one another’s work and to begin defining how the Cluster would operate going forward. The Karamoja Household Resilience System Map served as a key organizing tool, structuring communication and framing discussion. Additional detail on the workshop outputs is available in [37, 38].

The mapping process generated a number of positive outcomes at the workshop, which were not quantified directly but captured from post-event feedback. First, stakeholder participation in the creation of the map helped promote a collective understanding of the system and of the drivers of resilience. The map showed how the activities’ results chains were linked to the high-level outcomes that they cared about, which helped them to better understand the system, by locating their own work within it and framing their contribution through their existing understanding. This enabled the stakeholders to better understand each other’s mindset, perspective, and dynamic hypotheses about system change. As one participant noted, it was useful to “see how moving parts work together”.

The workshop also gave implementing partners their first structured opportunity to learn about one another’s work. Because the map captured all USAID-funded activities in the Cluster districts, participants could quickly identify where interventions were concentrated, who was working on related issues, and where opportunities existed for synergy, collaboration, or sharing of resources and expertise. We also encouraged participants to think in terms of “complementarity,” that is, ways in which interventions might support one another indirectly by acting on different parts of the system that jointly shape a desired outcome. The workshop produced a documented set of specific collaboration opportunities, and a participant reported that they had “learned of other activities’ interventions and potential areas of collaboration”.

Beyond identifying specific opportunities for collaboration, the workshop also surfaced broader priorities for collective action by highlighting key barriers to change and gaps in the existing approach. In particular, participants identified water access, conflict, and gender issues as cross-cutting factors affecting multiple dimensions of household resilience. These insights were made more visible by the map, which encouraged participants to view resilience from a systems perspective rather than through isolated interventions. In this way, the map helped clarify programming needs and opportunities in the region, particularly for newer activities.

The Karamoja maps were intended to serve as the foundation for a further stage of work: layering available indicators onto the conceptual models to show how the system was evolving over time. In Blair et al. [26], we argued that this combination of structure and partial data can substantially improve the usefulness of causal maps in development and humanitarian settings, where information is often incomplete but still sufficient to highlight important patterns of change. That next step was not completed in Karamoja because the onset of COVID-19 led USAID to redirect attention toward the immediate pandemic response. As a result, our engagement with the Karamoja partners was cut short, and we could not follow the longer-term use of the maps or document their influence on programming.

At the same time, this shift provided an important demonstration of the broader value of the approach. Rather than continuing in Karamoja, we pivoted to the larger Uganda market system mapping work, which USAID used to support rapid assessment of COVID-related disruptions. This underscored one of the main arguments of this paper: when interventions must be adapted in response to a shock, it is highly valuable to begin from a shared understanding of how the system works. A common mental model makes it easier to assess which connections are likely to be disrupted, which actors or functions are most vulnerable, and where adaptation may be most effective. Thus, while this paper focuses on the contribution of adapted systems modeling to shared understanding and intervention design under relatively stable conditions, the same modeling logic also informed later crisis analysis. We leave to a separate paper the explanation of how rapid model adaptation early during the COVID-19 shock helped assemble unstructured data, define sentinel indicators, and anticipate emerging risks across the agricultural market system.