4 Ideas for agriculture sourcing systems

Note: this is still rough, but with limited internet access I decided to send it anyways. Apologies in advance for shoddy grammar, spelling, and the like

As an EWB overseas staff member I’ve been partnered with a private company, TreeCrops. My work is this: set up a system of depots along a strip of lakeshore from Mangochi to Monkey Bay that we be used by rural Malawians to process, store and bulk baobab fruit. TreeCrops buys this fruit and exports it to the EU.

This is my second week in the field. My first job was to clean up and outfit a depot in Maldeco, a small town north of Mangochi. Yesterday, the depot was full of Malawians industriously cracking, separating, and bagging thousands of kilograms of fruit, each in her or his own way. Charming, but troublesome when trying to have hygienic food-quality baobab pulp as our end product. 

So I chatted with each of the dozen folks at work, sat down and wore my hands a bit raw extracting baobab pulp and fibre, and observed how each person went about their work. I stayed until quitting time, then helped clean and close up shop.

There’s a lot of need for improvement to the workers’ processing methods. It’s clear, however, that to change workers’ from what they’re doing to what we’d like them to do is going to take more than lecturing and demonstrating repeatedly. Everyone processes his or her own way, and for a reason. I want to get behind their eyeballs and see their work as they see it so I can understand how they might want to change what they do.

Thinking of all of this reminded me of a cantankerous English development architect, Eric Dudley, and his principle of designing for “maximum tolerance to failure”. This Dudley principle urges me to think not of how fantastically efficient everything could be, but of how I could set things up so if everything that could go wrong goes wrong, I’ll still achieve my most basic goals for my work.

All of this thinking meshed well with my engineering background, particularly control systems courses. Here are four ideas of ideas that should be incorporated into designing workable systems for rural agriculture purchasing.

A useful metaphor to help explain things comes from a course project in which we had to design a control system for a ball on top a “balance beam” of two parallel stainless steel bars. A disturbance knocked the ball off centre and our system needed to tilt the parallel bars until the ball was again safely at rest. Here are some general principles.

1. Definition of failure

What do I actually consider a failure when it comes to processing baobab? Germs found in a sample? Foreign matter above a certain threshold? Different grades mixed together? Perfection may not be required (or much less be achievable). With the balance beam failure is easy to spot—the ball falls to the floor. With baobab processing, failure is a fuzzier notion. A first step for me will be to adequately describe what actually is processing failure before I can design to avoid such an outcome.

2. Response

When controlling a balance beam, if the corrective action is too slow the ball will fall before the system is able to respond. If the corrective action is too fast, the system produces all kinds of jerky over-reactions (called over- or under-shoot) that will never let the ball actually settle down. With baobab powder, how long can we have people using substandard processing methods before we “fail” and also, what are the dangers of trying to quickly swoop in and correct every minor infraction? There will also be lag-time for any corrective response, but finding lag-time that is neither too fast or too slow is tricky.

3. Convergence

I expect that workers will eventually find methods that work well for them, and get themselves into a groove; they won’t endlessly change how they processing the baobab fruit. I don’t expect everything to go right the first time, but I want their eventual methods to be sufficiently hygienic by TreeCrops’ standards. How do I get people to converge on these targets instead of diverging towards bad practices? With the balance beam, how does the system produce the right type of feedback (positive or negative) that will cajole the ball back to a standstill? For us, what type of feedback helps people track towards good practice, not away from it?

4. Variation

No dynamic system will be in one state 100% of the time. There’s always variation. Even with the best balance beam control the ball will still oscillate back and forth a bit without the system having to respond. So should it be with us. If we overreact at every minor infraction—someone forgetting to remove their shoes, a newcomer mis-sorting grades, an unswept floor—we’ll burn ourselves out while frustrating our workers and gaining very little. Every control system must accept some level of variation about its target.

Each of these four points is just an idea inside of a balance beam metaphor. I’ll have to get out there and see if they actually apply to what we’re trying to achieve. Luckily, my next objective is to set up three other depots. This reminds me of another principle: the lack of truth in small numbers. I shouldn’t be okay with drawing many conclusions from one depot, or one dozen people. I’ve got to get out and see much, much more.