Your product is an organism

Let's face it, the universe is messy. It is nonlinear, turbulent, and chaotic. It is dynamic. It spends its time in transient behavior on its way to somewhere else, not in mathematically neat equilibria. It self-organizes and evolves. It creates diversity, not uniformity. That's what makes the world interesting, that's what makes it beautiful, and that's what makes it work.

D. Meadows — "Thinking in Systems"

The dominant metaphor for talking about digital products is still the machine. It has parts, it has blueprints, it gets assembled, switched on, broken and repaired. We talk about architecture, components, modules, pipelines. Somebody, somewhere, knows where every screw is.

It is a comforting metaphor and it describes nothing of what actually happens. No digital product with a few years behind it has blueprints that correspond to what is there. It has parts that grew because a large customer asked for something on a Tuesday, scar tissue around an incident from 2019, vestigial organs nobody dares excise because nobody knows what they do and something stopped failing when they were left alone.

That is not a machine. It is an organism.

What the wrong metaphor costs you

The most important practical consequence is the one Meadows points at in the quotation: the world does not live in mathematically neat equilibria but in transient behaviour on its way somewhere else. Neither does a product.

From which follows the hardest thing to accept: disproportion. In a non-linear system, the size of an effect bears no relation to the size of its cause. You change two words in a sign-up form and a metric that had been frozen for two years moves. You rewrite an entire module over a quarter and nothing moves at all. That experience is so common almost everyone has had it, and yet we keep planning as though results were proportional to effort, because that is how budgets are built.

The second consequence is delays. The effects of a product decision almost never appear when it is made. They appear months later, mixed with the effects of ten other decisions, once nobody remembers the order. A system with long delays and murky feedback is a system that teaches very badly, and it explains why organisations attribute their results to whatever they did most recently rather than to whatever caused them.

The third is that the levers are where you are not looking. Meadows ranked the leverage points of a system and put parameters — the numbers we tune — in last place. Almost everything we do in product work is parameters: the price, the threshold, the trial length, the number of items per page. The real levers sit in information flows, in rules and in goals. Which is to say, in things that do not look like product tasks and have no ticket.

The organism includes the organisation

There is one part of the metaphor usually left out, and it is the most uncomfortable. The organism is not only the software. It includes the people who build it, how they communicate and the structures that keep them apart.

Conway's observation has been on record for sixty years and still surprises whoever hears it: the structure of a system reproduces the communication structure of the organisation that built it. Not as a vague tendency, but fairly literally. If two teams barely speak, there will be a brittle integration precisely at the boundary between what each of them does. If support has no channel to development, the product will have a hole exactly where learning from failure should be.

This means some organisational decisions are architectural decisions, even when they are taken in a committee of people who believe they are discussing reporting lines. And it means some persistent technical problems will not be fixed by refactoring, because their cause is in an org chart.

How to steer something that organises itself

If the product is an organism, the word "steer" has to be understood differently. You do not steer an organism by issuing orders to its organs. You create conditions for it, you observe its behaviour over time, and you correct.

In practice this comes down to three habits. Look at series, not snapshots: the current value of a metric says little, its shape over time says almost everything. Intervene in small and observable ways, because in a non-linear system large interventions do not produce large effects, they produce illegible ones. And be suspicious of explanations that work too well, because in a system with long delays there is always a coherent and false story available.

Meadows closes her paragraph by saying that this mess is what makes the world work. It is a useful reminder for an industry that spends a remarkable amount of energy pretending its material holds still.