Click on the link below to view the webinar by Béla Teeken from IITA on 23 October 2025 entitled Breeding for social Impact: mapping cassava trait preferences through citizen science and farming typologies.
Presenter Bio: Béla Teeken is the Gender, Youth, and Social Inclusion Program Leader at the International Institute of Tropical Agriculture (IITA) in Cotonou, Benin. His research links social science with breeding and agronomy to understand how local culture, ecology, and institutions shape agricultural innovation and social inclusion. Through participatory, citizen-science approaches, he leads co-design initiatives that integrate gender dynamics, market insights, and user perspectives to guide breeding investments for social impact.
For more info, contact Béla at b.teeken@cgiar.org
Summary of presentation: Using tricot on-farm trials with 448 farmers over two years alongside the RHoMIS survey, the study mapped gendered and socio-economic differences in varietal and trait preferences from planting to processing. Three farmer segments emerged—market-oriented sellers, women-led processors focused on value addition, and resilient mixed-income farmers—each reflecting distinct value chain roles. The results demonstrate how combining participatory variety selection with social and market insights enables real-time customer segmentation and more targeted breeding investment cases for greater social impact.
Link to Béla’s slides (pdf) here
| Question | Answer |
|---|---|
| How are the data being used in advance and release decisions? | The team integrates multiple data streams—consumer testing, participatory processing, advanced clone evaluations, and multi-location trial results over several years—and reviews them in Product Advancement Meetings. Rather than relying on a single metric, decisions are made through interdisciplinary discussion to identify varieties that perform well across contexts. These meetings include IITA/CGIAR teams, and especially NARS (like the National Roots Crops Research Institute -NRCRI- in Nigeria), whose local expertise is essential and who are the only ones that can formally release a variety in the country. They’re formalizing inclusive protocols so all disciplines and partners contribute from the outset, with the goal of prioritizing breeding investments that maximize social impact. At the same time, industry needs are addressed through demand-creation trials and seed pipelines, which can expand markets and create downstream benefits for smallholders. Finally, attention follows the product map: pipelines and areas showing growth—particularly those linked to industry—signal future trends the breeding program should resource. |
| How should breeders handle diverse farmer preferences—where analysis shows multiple varieties favored for reasons beyond yield—when they typically aim for a single high-yield “blockbuster”? In other words, how can this work guide breeders to make decisions (and accept recommending 3–4 varieties) when farmers’ choices vary by context? | I agree yield drives genetic gain, but I believe breeding decisions must be more holistic to drive adoption. Our evaluations and tricot results show we should routinely integrate social preference data with multi-location trial performance—ideally through automated analyses. First, we clarify our impact objective (for example, focusing on medium and small processors who set quality standards), then we choose cross-cutting varieties—like Baba 70 and Game Changer—that meet broad needs and also serve industry. By synthesizing diverse evidence and using an adoption-oriented index that includes yield plus other valued traits, we can still identify blockbuster varieties and speed uptake. The objective of identifying trait prioritisation and variety preferences for different social segments is not to release different varieties for all these segments but to map the diversity among the crop users so we can focus on varieties that have a good composite of traits that can address as many needs as possible, so this informs which varieties can be blockbusters. In parallel, I see value in regionalizing efforts (e.g., separate tricot work for northern vs. southern Nigeria) to capture additional gains. |
| How does the tricot method compare with the mother baby approach for on farm trials? | We started with mother–baby trials because early surveys showed farmers struggled to articulate desired traits; offering options worked better. In that setup, a researcher- or farmer-managed “mother” trial sits alongside many farmer-managed “baby” trials. We structured it a bit like tricot (incomplete blocks), but the results weren’t representative enough. To get robust data from baby trials, we needed many more of them—at which point the mother trial became less essential, with multi-location trials effectively serving that role. In practice, tricot is a scaled-up mother–baby approach that boosts statistical power through more participants. There’s still high variance because farmers manage plots differently, so I think adding optional, predefined agronomic “packages” that farmers commit to could reduce noise and make tricot even more powerful. |
| How did you generate the list with traits being evaluated? Are they use generated? Did you manage for different social characteristics when generating them? | We chose which traits to evaluate at each stage—from planting to processing—based on survey work about what people value in a variety. Insights from women, especially through the RTB Food project, highlighted canopy coverage and plant architecture in the field, and quality traits during processing like moldability, color, and texture; in fact, texture/color issues flagged some varieties as poor candidates for advancement. Using this evidence, we defined stage-specific trait lists and made sure farmers clearly understood what they were scoring. We also captured overall “best/worst” preferences with the reasons why, which often surfaced traits we hadn’t asked about. Tricot is great for eliciting these emerging priorities, and with the AI tools we’re building in the 1000 Farms project, we can rapidly analyze that feedback to detect new, decision-shaping traits. |
| What were the main tasks of the lead farmers? Were the trials planted and harvested by the farmers themselves or by the lead farmers? | Because cassava stems are perishable, we can’t hand out standardized “seed packages” like with grains. Instead, we multiplied planting material at several stations across the country to keep delivery distances short, then our IITA/NRCRI teams went to fields and planted with farmers to keep trials comparable (e.g., consistent spacing), while still letting farmers include their own varieties. We appointed a lead farmer per group (about 10 farmers) to visit each plot at defined time points, record observations in the handbook, and submit data—sometimes by photographing pages for us to enter, or with help from family if phone skills were limited. Lead farmers were typically community members; while extension officers can play this role elsewhere, in Nigeria the community-based model worked best. |
| What is the estimated time frame for evaluating the variety with the farmers, and when can we expect it to be ready for release? | In Nigeria, tricot is now an accepted on-farm testing method for variety release, alongside required DUS and other trials, with “prior-to-release” sessions run by the release committee as needed. Once a tricot round and on-farm testing are finished, I aim to compile all necessary data within about a week; the farmer-engagement itself takes longer because we must build trust, prepare well, and run through a full crop cycle from planting to harvest and processing. With our AI integration, analyzing the qualitative “why” behind choices should also be possible within one to two weeks. We’ve been building the 1000 Farms “Climb Up” system with RTB crops to make this as efficient as possible. While some criticize tricot as costly, that depends on design: if you insist on villages 20 km apart, costs rise—far beyond what typical on-farm testing does. By clustering farms across regions, we reduce costs to roughly on par with standard on-farm testing. Early pilots were pricier due to heavy monitoring and learning, but increasing digitalization (e.g., field photos from start to finish) is driving substantial efficiency gains. |
| Do you make a selection of different environments to start the experiment? | We deliberately targeted states with intense cassava production and processing—especially small-scale processors—while ensuring coverage of the forest and derived savannah zones. In tricot design, agroecology and product profile matter: if you span multiple zones with one tricot, you need many more participants for statistical power; often it’s better to run separate trials within the same product profile. We never mix boil-and-eat types with bitter, processing types (cyanide risk and different end uses), because they serve distinct markets. Our segmentation anchors on the final product and market—home consumption (boiled), processed market products, or biofortified lines—with the understanding that carotenoids/vitamins behave differently in boiled vs. processed pathways. Bottom line: you don’t “just set up a tricot”—you design it around market segmentation linked to agroecology and end-use. |