Your sales team agrees delivery dates. Operations misses some of them. Sales finds out when a customer complains, three weeks later, in a meeting neither of them is in.
Nobody is failing. There is no path from the decision back to its consequence, and there never was. Someone is deciding and never seeing what their decision did.
Why these three models
The decision is where to spend your next intervention on a process that keeps producing the same problem. The features that fire are a recurring malfunction, a decision made without visibility of its result, and a proposed fix whose value nobody has established.
Three lenses. The leverage search produces a complex answer about where to intervene and in what order. The value of information produces an equilibrium answer about whether the missing feedback would actually change anyone’s behaviour. Behaviour modes produce a cycle answer about what you are looking at before you fix it. The first two disagree productively: one says information flows are high leverage, the other says only where somebody would act differently.
1. Where to look, in what order
Meadows offers a ranked list of places to intervene in a system, from weakest to strongest: numbers, buffers, stock-and-flow structure, delays, balancing loops, reinforcing loops, information flows, rules, self-organisation, goals, paradigms, and transcending paradigms.1
Two things about that ordering matter here, and the second is more important than the first.
The first is where information flows sit. They rank above rules and above every physical parameter, which is a strong statement given that most operational effort goes into the weakest end of the list. Meadows describes missing feedback as one of the most common causes of system malfunction, and adding it as often relatively cheap.1 That combination, common cause and cheap fix, is unusual and is why this article exists.
The second is what the ranking is not. Meadows states in the original article that the list arose from personal experience and from insights gleaned from systems analysis rather than from rigorous empirical testing, and calls it a work in progress and an invitation to think more broadly.1 The critical literature makes the same point about the absence of evidence for the ordering.
So use it as a search order and never as a measurement. It tells you to look past the parameter you were about to tune. It does not establish that goals are reliably more powerful than delays in your situation, and saying “this is a level 3 intervention” as though that were a quantity is a misuse.
The founder-relevant compression is that you can usually reach three of the twelve: information flows, rules and goals. Paradigms are not on the menu. So search from the deep end, and act at the deepest point you can actually reach, which in practice is very often the information flow.
Leverage search, the intervention-order lens
- Assumes: intervention points differ in power, and deeper ones are harder to move.
- Fits because: you are about to fix something and have not asked where in the system to fix it.
- Breaks when: treated as a measured ordering, which it is not and was never claimed to be.
- Evidence: grade C. The author disclosed that it came from experience rather than testing, and critics confirm the gap.
- Counteracts: tuning the first parameter that comes to hand.
- May reinforce: false authority, because a ranked list reads as a finding.
2. Whether the feedback would change anything
The second lens is the discipline that stops this becoming a dashboard-building exercise, which is how it usually fails.
Information has value only where it can change an action. Before connecting anything, ask what the recipient would do differently on receiving it. If the answer is nothing, you have not added feedback. You have added a report, and reports accumulate faster than anyone reads them.
Applied to the delivery example, the test is specific. If sales sees the miss rate and can adjust what they promise, the loop closes and the information has value. If sales sees the miss rate and has no authority over what they promise, because the quoted lead time is set centrally, then the feedback changes nothing and the real intervention is a rule change one level up.
That distinction is worth labouring because the two failures look identical from a distance. Both present as a team that keeps making the same mistake. One is a missing information flow and is cheap. The other is a missing authority and is not.
Feedback that reaches someone who cannot act on it is a report, not a loop.
Value of information, the decision-relevance lens
- Assumes: information is worth its cost only where some result would change what someone does.
- Fits because: you are about to connect a signal to a person.
- Breaks when: the value is cultural rather than decisional, where visibility changes behaviour without changing any specific choice.
- Evidence: grade A. A formal result rather than an empirical claim.
- Counteracts: building dashboards nobody acts on.
- May reinforce: withholding information from people on the grounds that they cannot use it, which is how authority calcifies.
3. Name the malfunction before you wire anything
The third lens runs first in practice and is skipped almost every time.
Plot the thing that is going wrong before diagnosing it. Each fundamental shape implies a different structure. A miss rate that has grown steadily points at a loop with nothing balancing it. One that oscillates points at a delay, and adding more information into a system that is already oscillating from delay can make it worse rather than better, because you have increased the gain on a lagged signal.
That last point is the reason this lens belongs in an article recommending feedback. More feedback is not uniformly good. In a system whose problem is delay, faster reporting of a stale number increases the amplitude of the correction, and the team ends up over-adjusting in both directions with more data than before.2
So the sequence is: plot it, name the shape, and only then decide whether the missing information is the constraint or whether the delay is.
Behaviour modes, the shape lens
- Assumes: a trajectory belongs to a small set of shapes, each implying a different structure.
- Fits because: you have the data on the recurring problem and have been arguing from impressions.
- Breaks when: the series is short or noisy, which at small volumes it usually is.
- Evidence: grade B. Descriptive and useful, with no strong predictive claim attached.
- Counteracts: adding information to a system whose problem is lag.
- May reinforce: analysis paralysis on a fix that costs an afternoon.
The levers, cheapest first
- Find the decision with no return path. Ask who commits to something they never see the result of. There is usually one obvious answer and everybody already knows it.
- Check they can act. Would they do anything differently? If not, the intervention is authority, not information.
- Plot the problem first. Ten minutes. If it oscillates, more reporting may make it worse.
- Connect people before you build systems. The cheapest feedback loop is one person joining one recurring meeting. Dashboards come later, if at all.
- Shorten the path, not just the interval. Weekly reporting of a three-week-old number is still three weeks stale. The lag matters more than the frequency.
- Search deeper before settling. Information flows rank above rules on the list, and rules above parameters, so at least look at the deeper points before acting at the shallow one.
What to do this week
Do now, sized at half an hour, effect visible within a cycle. Take your most persistent operational problem and write down who makes the committing decision and whether they ever see its consequence. Reversible, free, and dominant across every scenario about why the problem recurs.
Hedge, where the premium is the whole loss, live before the next cycle. Put that person in the meeting where the consequence is discussed. One recurring invitation. If the loop was never the constraint you have spent one person’s half hour a week, and that is the entire downside.
Defer and trigger, size fixed now. Do not instrument the company. Pre-commit the trigger: if the problem persists a full cycle after the person can see the consequence and has authority to respond, the intervention moves up the list to rules or goals. Decide now who makes that call, because an escalation with no owner does not happen.
Note the arrivals. The invitation lands this week. Behaviour changes after the person has seen the consequence enough times to believe it, which is two or three cycles, and the temptation to declare failure will peak before then.
What usually happens next
One further caution belongs here because it is where the cheap fix stops being cheap. A loop that connects two people also connects two schedules, and the meeting that closes it competes with everything else in the week. Loops added without removing anything decay within a quarter, not because anyone decided to drop them but because the calendar reasserts itself.
Run the break test first. Has a rule changed, has an actor entered or left, has a measurement become a target? The third is the specific risk when adding feedback: the number you start showing people becomes the number they manage, and the part of the outcome it does not capture quietly degrades.
If nothing broke, the pattern is reliable. The loop gets closed, the problem improves for a quarter, and then a reorganisation moves one of the two people and nobody notices the loop has been cut. Feedback structures decay silently because their absence produces no error message.
There is a second failure worth naming because it looks like success. The loop gets built as a report rather than a conversation, the recipient reads it for two weeks, and then it joins the other unread reports. Nothing has been connected, but the item is marked done and the underlying problem now has a fix attached to it, which makes it harder to raise again.
Subtract the counterfactual before crediting the fix. A miss rate that fell in the quarter you added the loop may have fallen because volumes fell. The test is whether it stays down through a busy period.
What this ensemble cannot see
All three lenses treat feedback as information moving between roles. Much of what is missing is not informational at all.
People frequently know exactly what their decisions cause and continue anyway, because the incentive points the other way, because raising it is socially expensive, or because the person who would have to act outranks them. This framework will diagnose those as missing information and prescribe a meeting, and the meeting will happen and change nothing. The value-of-information test catches some of that and not the version where someone can act and will not.
There is also a real weakness in the central model, stated in its own card. The leverage ranking is a heuristic its author explicitly declined to claim as tested, and this article leans on its most useful single claim, that information flows rank high and are cheap. That claim is plausible, widely repeated and not established.
And one property none of these models contains: closing a loop changes the relationship between the two people in it. Sales seeing operations’ miss rate is also sales acquiring a stick, and how that lands depends on things no diagram captures.
The one action that survives the ignorance: this week, name one person who commits to something and never sees the result. Ask them what they would do differently if they saw it. If they have an answer, connect them. If they say nothing would change, you have found something more important than a missing loop, and it is not going to be fixed by a meeting.
Who has to move
The person who needs this usually owns neither side of the broken loop, which is why it has survived. The cheapest first test costs one recurring invitation and half an hour a week. If the problem improves over two cycles, the loop was the constraint. If it does not, you have eliminated the cheapest explanation and can stop paying for reporting that was never going to help.
Sources and notes
- Donella H. Meadows, Thinking in Systems: A Primer, Chelsea Green. The twelve leverage points, ordered from weakest to strongest as numbers, buffers, stock-and-flow structures, delays, balancing feedback loops, reinforcing feedback loops, information flows, rules, self-organisation, goals, paradigms and transcending paradigms, are chapter 6, with information flows at rank six and the observation that missing feedback is among the most common causes of system malfunction and that adding it is often relatively cheap. Meadows states in the original leverage-points article that the list arose from personal experience and insights gleaned from systems analysis rather than rigorous empirical testing, and describes it as a work in progress and an invitation to think more broadly about system change. It is used here as a search order accordingly, and the subsequent critical literature confirms the absence of empirical validation for the ordering.
- John D. Sterman, Business Dynamics: Systems Thinking and Modeling for a Complex World, McGraw-Hill. The fundamental modes of dynamic behaviour used in section 3 are chapter 4, and the treatment of delays as the standard source of oscillation, with the consequence that increasing the gain on a lagged signal amplifies rather than damps the swing, is chapter 17. Together these support the caution that adding information to a delay-driven system can make it worse.
A note on recommending a framework this article also undercuts. The leverage ranking is graded C here and used anyway, because its function is to widen a search rather than to settle a question, and a search order does not need to be true to find things. The specific claim carried into the recommendation, that information flows are a common failure and a cheap fix, is the part that should be treated as a hypothesis you test on your own process rather than a result you inherit.
Joshua Agonya Pi’Rwot, Founder.