Map the value stream
A 60–90 minute walk-through with the owner. End-to-end trace from "customer becomes aware" to "customer pays" to "customer returns." The output is a list of candidate resources to measure.
The methodology
Find the constraint. Fix the constraint. Find the next one.
Every engagement runs through the same operational protocol. It's grounded in Theory of Constraints, triangulated across three independent kinds of evidence, and sequenced cheapest-evidence-first so we don't spend money to learn what an hour of looking would tell us.
The diagnostic
A six-step protocol
The diagnostic is sequenced by cost-of-evidence: the cheapest tests run first, the expensive ones only run if needed. Most engagements resolve to a confident answer in two weeks.
Triage: supply-side or demand-side
Compute utilization curves for every candidate resource during revenue-weighted peak hours. If everything sits under 60% utilization at peak, demand is the constraint — different conversation. If anything hits 100%, the constraint is operational and we continue.
Rank candidates by impact
For each supply-side candidate, sum five impact categories:
- Stockout-driven lost throughput
- Mix optimization loss (constraint hours spent on lower-margin items)
- Pricing power loss (chronically-saturated items underpriced)
- Substitution loss (lower-margin alternatives the customer takes instead)
- Waste from chasing the constraint badly
Rank by total annual impact × persistence × saturation.
Throughput Dollar-Days
For the top candidate, compute TDD = annual impact × (days persisted / 365). The urgency metric. Anything over $50K (small business) or $500K (mid-market) is operationally significant and worth fixing now.
Trace to the policy
Effect-cause-effect reasoning from the observable physical constraint to the underlying rule. The fix targets the policy because changing the equipment without changing the policy reproduces the same problem at higher capacity. (See Constraints → Policy for why.)
Subordination test and final report
Simulate elevating the candidate constraint by 20%. Does system throughput rise? Does another resource immediately become binding? If yes to the first and no to the second, the diagnosis is confirmed. The deliverable is a six-to-ten page written report — yours to keep regardless of whether you continue.
Triangulation
Three independent kinds of evidence
No single source of evidence is sufficient. Each one is suggestive on its own; together they're conclusive.
Data
Quantitative, defensible, captures patterns the operator can't see directly. Blind to policy, customer abandonment, and declined orders unless those are logged.
Tests
Controlled experiments to confirm causation. Cheap, time-boxed hypotheses — re-sequence one Saturday's production, raise a price for two weeks, add eight hours of decorator time. Tests validate forward-looking impact.
Surveys
Owner, staff, and customer interviews. Surface policy constraints data can't directly show. Questions that work: "Walk me through your most stressful 30 minutes." / "What's a rule we run by that nobody questions?" / "When did we last say no to a customer?"
When all three agree, the diagnosis is high-confidence. When they disagree, the disagreement is itself informative — data shows the currently binding constraint; the owner often feels a constraint that's been binding for years; a failed test means the diagnosis was wrong and we revise.
The ongoing cycle
The Five Focusing Steps
Diagnosis is the start, not the end. The constraint moves once you resolve it — that's not a flaw, it's the system working as designed. The Five Focusing Steps are the ongoing improvement cycle every operation runs once it starts paying attention.
Identify
Find the binding constraint at this moment.
Exploit
Squeeze every drop of capacity from the constraint without spending money. Remove waste, eliminate interruptions, ensure it never sits idle.
Subordinate
Everything else in the system serves the constraint. This means deliberately under-utilizing non-constraints — the hardest step because it offends the "keep everyone busy" instinct.
Elevate
If exploiting and subordinating aren't enough, invest to expand the constraint's capacity. This is where AI and capital expense typically enter the picture — but only after the cheap moves are exhausted.
Repeat
The constraint has moved. Find the new one. This is a continuous cycle, not a one-time fix — which is why we offer a Throughput Cycle retainer for ongoing engagement.
When AI enters
Three prongs of implementation
Once the constraint is identified and the policy traced, the next question is what to build. AI implementations fall into one of three categories. Which one depends on where the constraint sits.
1
Find better customers
Lead generation, qualification, personalized outreach, market analysis. Targets revenue-side constraints — when the binding limit is demand rather than capacity.
2
Automate operations
Workflow automation, scheduling, inventory, fulfillment. Targets throughput constraints. The Goldratt test before building: is this aimed at THE bottleneck, or just at convenient busywork?
3
Amplify the people
Decision support, knowledge management, training, quality control. Targets human-capacity constraints — when a specific role or expertise is what the system is waiting on.
The acceleration precept
Build for a world where models get cheaper and smarter every quarter.
Every AI solution we build sits inside a foundational reality: models are getting dramatically cheaper and dramatically more capable on a cycle measured in months, not years. Building for that is operational — not philosophical.
Plan for cheaper
Don't hardwire a specific model. Build abstraction layers. Use the cheapest model that can handle each subtask. As prices drop, the unit economics improve automatically without re-engineering.
Plan for smarter
Architect to absorb capability as it arrives. Identify the "waiting room" — tasks almost automatable today, six months from full automation. That turns a one-time engagement into a recurring relationship that compounds.
Foundations
Built on a body of work
The methodology is operations-first, with intellectual roots in throughput accounting and constraint theory.
Eliyahu Goldratt
Theory of Constraints · The Goal · It's Not Luck
The system has one binding constraint. Find it. Subordinate everything to it. Elevate it. Repeat.
Throughput Accounting
Corbett · Smith · Bragg
Operational measurement built around throughput, inventory, and operating expense — not the cost-allocation fictions of traditional accounting.
20 years operating at scale
USAF · Morgan Stanley · healthcare · pharma · telehealth
Field-tested against real businesses, real constraints, and real pressure to produce results that show up on the P&L.
Ready to find your constraint?
The Throughput Diagnostic runs this protocol against your business. Two weeks, fixed fee, a six-to-ten page report naming the binding constraint, quantifying the annual leak, and recommending the specific change. Yours to keep with no obligation to continue.
Book a call