Drawing the boundary: what's in your system

A system boundary is a decision about what you include in your analysis and what you treat as outside weather — and it is a decision, not a fact about the world. Meadows makes the point plainly in Thinking in Systems: there are no natural boundaries; we draw them to suit the question, and most bad diagnoses are boundary errors before they are anything else. Draw the boundary too narrow and every problem looks external and unfixable; draw it too wide and analysis dissolves into "everything affects everything". The founder's skill is drawing it deliberately, per question, and knowing they have done so.

Why do boundaries matter more than they appear to?

Because the boundary silently determines the verdict. Ask "why is sales underperforming?" with the boundary drawn around the sales function, and the answer will be found inside it — effort, skill, tooling — whether or not that is where the cause lives. Widen the boundary to include delivery capacity, pricing and the founder's calendar, and the same facts produce a different culprit. This is the core move in A Systems-Thinking Guide for Founders: before optimising elements, check whether the frame around them is the right frame. Blame is usually a boundary artefact. "The market is tough" and "the team lacks hunger" are both claims that the cause sits outside the speaker's chosen boundary — convenient, untestable, and frequently wrong.

What does a boundary error look like in a service firm?

Two failure directions, both common in firms of 5–50 people:

• Too narrow. The firm treats lead flow as external weather — some months are good, some are not — when lead flow is substantially a function of activities inside the firm's control: outreach volume, follow-up discipline, referral prompting. Calling a controllable variable "the environment" removes it from the improvement agenda. The entire discipline of list building rests on the opposite move: The B2B Database Building Guide is, at bottom, an argument for pulling "who hears about us" inside the boundary and engineering it.
• Too wide. The founder's diagnosis includes the economy, interest rates, AI, and buyer psychology — a frame so large that no intervention is visible inside it. Analysis at this width produces commentary, not decisions.

There is a third, subtler error: drawing the boundary at the org chart. Departmental boundaries are administrative conveniences, but causes do not respect them — a "sales problem" whose root is slow proposal turnaround in delivery will survive any amount of sales coaching.

How do you draw the boundary well?

The mechanism, step by step:

1. State the question precisely. When the question is "why did Q3 bookings fall?", then the useful boundary differs from "why do we depend on two clients?". One boundary per question; reusing yesterday's frame is how stale diagnoses persist.
2. Start narrow and widen on evidence. When the cause is not findable inside the current boundary — you have checked the elements and they behave — then widen by one ring (adjacent functions, then external actors) and look again.
3. Include whatever you influence, even indirectly. When a variable responds to your actions on any timescale — referral rates, inbound volume, staff turnover — then it belongs inside the boundary, however external it feels.
4. Park the genuine weather explicitly. When something truly is outside influence — regulation, macro rates — then write it down as an assumption to monitor, not a cause to litigate.
5. Re-draw on a schedule. When the firm grows or the model changes, then last year's boundary quietly stops fitting; boundary review belongs in the same periodic examination as mapping the business as a system.

What sits at the boundary itself?

The most interesting objects in any business system straddle the line: clients, suppliers, referral partners, contractors. A client is simultaneously outside (their budget, their politics) and inside (your onboarding shapes their behaviour; your reporting shapes their expectations). Treating clients as fully external produces the passive posture — "they went quiet" — while pulling their experience inside the boundary turns silence into a design input. The same widening move rescues growth analysis: a firm that draws its boundary around "getting leads" will keep buying lead generation while capacity, of the kind examined in balancing loops and capacity, throttles everything from inside the ignored region.

What does this change on Monday?

Take your current most irritating problem and write two sentences: "The boundary I have been using includes ___. It excludes ___." Then ask whether the excluded list contains anything you influence. In our experience the exercise takes ten minutes and relocates the problem about half the time — which is a high return for ten minutes, and a habit worth adding to your working set of mental models. The boundary is yours. Draw it like you mean it.

---

Keep the thinking coming: The Format — a fortnightly letter on business and personal systems. Subscribe by email.