13. Systems thinking

§The second habit of mind is to see the product as a part of something, not as a thing.

§ 13.1The product is inside a system#

§A founder pictures the product as the object at the center: users come to it, use it, leave. Systems thinking redraws the picture. The product is one node. Around it are the users, who have habits and alternatives. The technology it runs on, which changes on its own schedule. The competitors, who respond. The incentives of everyone involved, including the team's. The regulations, the platforms, the channels. Every one of those is connected to the others, and the product's behavior is the behavior of the whole.

§Donella Meadows, whose primer is still the best short introduction, defined a system as elements, interconnections and a purpose, and observed that most people see the elements and miss the interconnections. In product this is the founder who sees users, features and revenue and misses that a change to the features changes which users arrive, which changes the revenue, which changes what the company can build next.

§ 13.2The four moves#

§Map the relationships. Before deciding, draw the system: who and what is connected to the thing you are about to change, and how. It does not need to be elaborate. A page of boxes and arrows exposes connections that a feature specification hides.

§Find the feedback loops. Some connections circle back. More users produce more content, which attracts more users: a reinforcing loop. More users slow the service, which drives users away: a balancing loop. Product Foundation's chapter on growth loops was one instance of this habit. The general skill is spotting where the arrows form a circle, because circles are where small changes become large ones.

§Identify the leverage points. Meadows's most-cited essay ranks the places to intervene in a system, and the ranking surprises people: adjusting numbers, like a price or a limit, is the weakest lever; changing the information flows, the rules, and the goals of the system are progressively stronger. In product, this means the setting you can change in a dashboard is usually the least powerful change available, and the one that alters what users can see about each other, or what the product is for, is the most.

§Anticipate second-order effects. The first-order effect of a change is what you intended. The second-order effect is what the system does in response. Add a referral bonus: first order, more referrals; second order, referred users who came for the bonus and leave when it is spent, and existing users who feel their friends are being bought. A team that only plans for first-order effects is surprised by the system every time.

§ 13.3Why this matters more for AI products#

§A product built on a model sits in a system with an unusual property: one of the elements, the model, improves on its own schedule and someone else's decision. A capability that is a differentiator in March is a commodity in September. Systems thinking is how a founder keeps that in view: which parts of the product are theirs, which are the model's, and what happens to the value of each when the model moves.

§ 13.4The discipline#

§Draw the map when the decision is load-bearing. Look for circles. Ask what pushes back. It takes an hour, and it is the hour that separates a decision from a surprise.