When growth stalls, the instinct is to attack everything: buy more traffic, rewrite the offer, hire salespeople, redesign onboarding, add features, and ask the team to work harder. That creates motion without leverage. A system can have dozens of problems and still have one problem that limits the next dollar of valuable output.
The Constraint Identification System is a weekly operating discipline for finding that limiting point. It compares acquisition, conversion, retention, and capacity on the same economic scoreboard, proves where flow is actually breaking, and directs the company toward one attack until the constraint moves.
Choose a checkpoint to see the operating question at that stage. The sequence keeps the company from confusing a visible symptom with the constraint that controls valuable throughput.
Map acquisition, conversion, retention, and capacity with the same period, scope, and definition of valuable output. Pick the lane with the strongest combination of economic impact, evidence, controllability, and learning speed.
flow map → comparable evidence → candidate constraintUse the lanes to locate where the system stops turning effort into valuable output. Do not treat them as four departments competing for attention.
Acquisition is the constraint when the system cannot create enough qualified opportunities to keep the downstream engine meaningfully loaded. Do not label it an acquisition problem just because lead volume is low. A small number of high-fit opportunities can outperform a large number of poor-fit inquiries, and a full pipeline can still be a demand problem if the wrong segment is entering.
Ask: Where does qualified demand first fail to arrive? Compare source, segment, promise, response time, and fit. Measure qualified opportunities per period, not raw leads. Then test whether additional demand would increase valuable output or merely create a larger queue for a later constraint.
Conversion is the constraint when qualified demand reaches the system and stalls at diagnosis, proof, offer, risk, price, follow-up, or decision speed. The most useful unit is the stage where a buyer should move but does not. A low close rate is a symptom; the constraint may be slow response, weak scoping, missing proof, unclear next steps, or a mismatch between promise and buyer.
Trace the queue: measure stage-to-stage conversion, time in stage, proposal age, no-decision rate, follow-up completion, and contribution per qualified opportunity. Read won and lost behavior together. If the change produces more closes but lower-quality customers or heavier delivery burden, the conversion improvement did not necessarily improve the system.
Retention is the constraint when customers enter, buy, and then fail to reach or repeat the promised outcome. This can look like churn, contraction, low usage, missed onboarding milestones, support escalation, or a cohort that never becomes economically durable. The question is not “Why did they cancel?” alone. It is “Which leading behavior changed before the cancellation, and which part of the value path failed to become normal?”
Segment the leakage: compare cohorts by acquisition source, use case, onboarding path, plan, customer size, and time to first value. Separate preventable fit, activation, delivery, support, and product issues from external events. Retention work is a constraint attack only when fixing the leakage increases durable contribution without overloading delivery.
Capacity is the constraint when work queues at a role, handoff, decision, or delivery step that cannot process demand at the needed rate. The bottleneck may be a senior reviewer, a founder approval, a scarce technical skill, a broken intake path, or rework caused by unclear standards. Hiring is not the first move. First prove where the queue forms and how much output the suspected constraint is actually producing.
Measure flow: track throughput, work in process, queue age, cycle time, utilization of the suspected step, rework, defects, on-time delivery, and contribution after service cost. Protect the constraint, remove avoidable interruptions, and subordinate adjacent work. Add capacity only after the evidence shows the old constraint remains.
A constraint score is useful when it forces comparable evidence. It is dangerous when it creates false precision.
| Test | Strong constraint signal | Weak signal or symptom |
|---|---|---|
| Economic impact | Improvement would increase valuable output, contribution, cash, or durable customer value. | The metric is interesting but not connected to an economic result. |
| Flow evidence | A queue, drop-off, delay, or repeated failure appears at the same step across comparable periods. | One anecdote, one bad week, or a complaint without a measurable location. |
| System effect | The step limits downstream output or creates a queue that other work cannot compensate for. | Improving the step would only make a non-constraint look better. |
| Controllability | The team can run a bounded intervention with a clear owner and review date. | The proposed fix depends on a market event or an undefined transformation. |
| Learning speed | The test can produce a leading signal before large spend or irreversible change. | The team wants to buy capacity before proving the mechanism. |
Use a consistent scoreboard. For each lane, record the same period, denominator, output definition, and economic guardrail. A qualified opportunity in one lane cannot be compared with a raw lead in another. A renewal rate calculated on surviving accounts cannot be used as if it measured the full starting cohort. A capacity estimate that excludes rework will overstate the output the system can actually deliver.
Score the candidate, then challenge it. A practical priority heuristic is economic impact multiplied by evidence strength and controllability, divided by time to learn. The number is not a law. It is a forcing function that makes the team explain why this constraint deserves attention now and what observable result would prove the diagnosis wrong.
Focused improvement is not passive analysis. It is a sequence that protects the constraint, removes avoidable loss, and earns the right to invest.
Map the flow from demand to cash or from customer need to delivered outcome. Locate the step where throughput is limited, work accumulates, or value repeatedly leaks. Use comparable data and observe the work directly when possible. The goal is to name one limiting mechanism, not to produce a list of every imperfection.
Use the existing constrained resource better before adding resources. Remove avoidable meetings, interruptions, rework, unclear inputs, unnecessary approvals, and low-value tasks. Improve the quality of work arriving at the constraint so its scarce minutes produce more valuable output.
Align adjacent activity to the pace and quality requirements of the constraint. That can mean limiting work in process, changing intake rules, sequencing requests, pausing low-value projects, or refusing to send incomplete work downstream. Local teams may move slower while the whole system moves faster.
Only after exploitation and subordination show that the constraint still limits output should you add people, spend, equipment, automation, or new distribution. Define the capacity purchased, the expected economic return, the ramp period, and the quality guardrails before approving the investment.
After an intervention, measure whether the constraint moved. If it did, return to the map. A new bottleneck may appear in acquisition, conversion, retention, or capacity. Do not keep optimizing the old constraint simply because the team became comfortable with its dashboard.
Write the counterfactual. If this lane improved by twenty percent, what downstream output would increase, and what evidence would show it? If you cannot answer, the lane is probably a symptom, a local optimization, or an investment thesis that has not been tested.
Map the four lanes, choose the constraint with the strongest evidence, write one bounded attack, and give the system enough time to show whether throughput actually moved.
Run The Constraint Path