The Five Focusing Steps: A Worked Design Agency Example
Key takeaways
- The five focusing steps apply to service teams: in a design agency, the constraint was a senior designer's review capacity, not a machine.
- Exploit the constraint before spending anything: triage, batching, and protected review windows cut average review time from 25 to about 14 minutes at zero cost.
- Subordination means the rest of the team paces itself to the constraint, accepting deliberate upstream idle time to protect system throughput.
- Elevate only when demand still exceeds constraint capacity after exploiting and subordinating, and test where the constraint will move before committing budget.
- A one-day paid scoping pass through the five focusing steps turns a constraint hunch into a quotable, evidence-backed engagement.
Most explanations of the five focusing steps stop at manufacturing theory. Here they run end-to-end on a small design agency's artwork-approval workflow, with the actual decisions, illustrative numbers, and trade-offs at each step: identify, exploit, subordinate, elevate, and repeat. If you scope client work before quoting a full engagement, this shows what the method looks like on real knowledge-work throughput, not a factory floor.
What are the five focusing steps, in one paragraph each?
Goldratt's five focusing steps, set out in his novel The Goal, are a repeatable loop for raising throughput of any system, including a service team.
Identify the constraint: the single step that sets the pace of the whole process. In knowledge work this is usually a person or approval stage, not a machine.
Exploit it: squeeze more through the constraint with zero spend, by removing rework or idle gaps.
Subordinate everything else: the non-constraint steps pace themselves to the constraint, even if that means idle time elsewhere.
Elevate it: only now buy capacity, adding a person or tool.
Repeat: once the constraint moves, start again. The same loop underpins a worked invoice-processing capacity example.
Which example process will we run the five focusing steps on?
A five-person design agency approving client artwork before release. One account manager briefs, two designers produce, one senior designer reviews and approves, one studio manager schedules. Everything the client pays for waits on that approval loop.
The process as measured
Illustrative numbers from a two-week observation, the kind of baseline you can gather in a single client visit:
| Role | Volume per week | Cycle time per item | Utilisation |
|---|---|---|---|
| Account manager | 40 briefs | 20 min | ~35% |
| Designers (2) | 40 artworks | 3 hrs | ~70% |
| Senior designer | 40 approvals | 45 min | ~95% |
| Studio manager | 40 scheduled | 10 min | ~30% |
The demand gap is visible: 40 artworks arrive weekly, the senior designer can clear roughly 38, so backlog grows by two items a week and lead time drifts from three days toward two.
The constraint hypothesis
Before any analysis, the working hypothesis is that senior-designer review capacity sets system throughput. The same structure appears in other service workflows, for example customer onboarding, which we break down in a customer-onboarding capacity example. The five focusing steps will test whether that hypothesis holds, and where the constraint moves once review capacity changes.
Step 1 — Identify: how do you find the constraint when work is invisible?
On a shop floor you watch parts pile up. In a design agency's artwork-approval workflow, the constraint hides inside inboxes and Slack threads, so you read queues instead of org charts.
Reading the queue, not the org chart
Pull three weeks of timestamps from the project tool: when each artwork entered review and when it left. The step with the longest queue and the oldest work-in-progress is your candidate constraint, regardless of who is loudest about being busy. Little's Law ties the evidence together: average WIP equals throughput times average cycle time, so a growing queue with flat throughput is a constraint, not a staffing problem (Little's Law). Our constraint calculator turns those same timestamps into a scoping answer before you quote the engagement.
Why utilisation alone misleads here
The studio manager reports every designer at 90% utilisation, yet artwork still waits five days. Utilisation measures how busy people are; a constraint is the step whose capacity sets system throughput. When reviewers juggle three accounts, utilisation looks healthy while the queue grows.
Step 2 — Exploit: what can you get from the constraint before spending anything?
Batching and triage at the review desk
Exploit means squeezing more throughput from the reviewer without buying anything or changing anyone else's behaviour. At the artwork-approval desk, three changes did most of the work, and each one is measurable in cycle time rather than opinion. First, triage: a two-minute checklist at intake splits approvals into three buckets, standard edits the reviewer waves through, borderline items needing full attention, and anything ambiguous sent back for clarification before it touches the queue; in our illustrative run, roughly 40% of items moved to the fast bucket, which cut average review cycle time from 25 minutes to about 14 without lowering rigour on the borderline work. Second, batching: the reviewer previously handled approvals as they arrived, paying a context-switch cost on every interruption, so consolidating into two fixed review windows per day protected focus time and raised the number of items cleared per session. Third, protecting the schedule: no meetings booked inside those windows, and WIP capped so the reviewer never carries more than one day's backlog.
The point of exploit is sequencing, not cleverness. These changes cost nothing but discipline, and they establish a baseline throughput you can compare against after Step 4. If you skip this step, you risk spending money elevating a constraint that was underused to begin with, and the capacity model behind that decision lives in the Constraint Impact Calculator, where you log the before and after numbers.
What we deliberately did not do
We did not hire, buy tooling, or ask the reviewer to work faster. We also did not touch the designers' process yet, even though their behaviour clearly fed the queue, because that belongs to Step 3. Exploit is deliberately narrow: one desk, one working pattern, zero spend. Keeping it that narrow is what makes the throughput effect attributable, and attributable numbers are what make the later steps defensible in front of a client.
Step 3 — Subordinate: why does the rest of the team have to slow down?
Subordination means the rest of the agency now works to the constraint's pace, not their own. Account managers stop releasing briefs in Monday-morning batches, and designers stop queuing six revision rounds at once. The rule is simple: never let the senior designer starve, never flood the studio with WIP waiting on her.
Setting the release cadence
The constraint needs roughly 45 minutes per artwork. If she starts at 9:30, the first brief must land by 9:15 and the next every 50 minutes after that. Anything released earlier just becomes WIP piling up in her backlog, which lengthens lead time for every client without adding a single unit of throughput. A visible queue with a cap, say three jobs maximum, makes the cadence self-enforcing. You can pressure-test whether the constraint actually moves after a cadence change with the Improve First calculator before you sell the client on it.
Handling the utilisation objection
The account managers will read the new cadence as their utilisation dropping, and they are right. Their capacity to release briefs now exceeds what the constraint can absorb, and that surplus is deliberately idle. Sell it honestly: the system's throughput is capped by the constraint, so upstream idle time costs the agency nothing, while upstream overproduction costs everyone lead time. Frame it as protecting the constraint, not policing the account team.
Step 4 — Elevate: when is exploiting and subordinating not enough?
The elevate-or-not arithmetic
After exploiting and subordinating, the studio's artwork-approval constraint still runs at 32 hours of demand against 30 hours of true capacity each week. Exploitation bought roughly two hours of throughput; subordination stopped the queue growing, but the constraint has no slack left for a bad week. That is the trigger for elevation: when the constraint is loaded past capacity even when everything upstream behaves.
Elevation here means adding capacity at the constraint, not hiring generally. The two realistic options are a second senior reviewer at roughly £2,400/month, or contracted overflow at £45/hour for peak weeks. Run the numbers against throughput: if the constraint adds one approved artwork per week at £180 contribution, a part-time reviewer pays for itself; paying for full-week overflow does not.
Testing the change before committing budget
Before signing anything, test the decision. Model the second reviewer in the scoping tool and check where the constraint moves — likely to account management's brief-writing. Then cap work-in-progress upstream so the new capacity is protected; teams using WIP limits on Kanban boards make the subordination visible rather than negotiable.
Step 5 — Repeat: where does the constraint move next?
Elevation changes the balance of the system. Once the studio's artwork-approval constraint gains capacity, throughput rises until the next-slowest step caps it — often upstream at briefing, where the backlog that once hid the approval queue now drains and exposes it.
Predicting the next constraint
Before you finish elevating, ask what the demand gap looks like at each step once the current constraint grows. If approval capacity rises from 40 to 55 jobs a week against 50 of incoming demand, briefing at 52 becomes the likely next cap. Write the prediction down; it is the cheapest test of whether you understood the system.
What to re-measure, and when
Re-measure cycle time, WIP, and utilization at every step two to four weeks after the change, not immediately — old WIP takes time to flush through. Then restart the loop at identify with fresh numbers, because the constraint you just relieved is rarely the one you will fight next.
How do you run these steps with a client before quoting a full engagement?
A one-day scoping sequence
Run one pass through the five focusing steps as a paid mini-engagement, not a free discovery call. Morning: map the artwork-approval workflow, time each step, and identify the constraint. Midday: quantify its impact — cycle time, WIP waiting, the demand gap against system capacity. Afternoon: test one exploit and one subordinate move in the Constraint Impact Calculator to see whether throughput shifts or the constraint simply moves. If the numbers show real headroom, the client has evidence to justify a fuller engagement.
Turning the findings into a client-ready report
The calculator's scoping report gives the client the constraint, its measured impact, and the predicted next bottleneck in one document. That converts a hunch into a quotable scope. The pattern addresses a familiar frustration: consultants asked for free advice on resource utilisation often get told the brief is too thin to even guess at. A paid, instrumented pass fixes the brief.
How do the five focusing steps compare with just modelling capacity in a spreadsheet?
A spreadsheet answers a different question. It tells you utilisation per resource against a plan; it will not tell you which step governs throughput, what happens to WIP when you subordinate the rest of the team, or where the constraint moves after you elevate it.
A simulation tool like ProcessModel or DO.Impact can model those dynamics, but at the cost of building a full model before you know whether the engagement is worth quoting.
The Improve First Constraint Calculator sits in the middle: it runs the identify-test-predict loop of the five focusing steps on one process, without simulation depth. Its honest shortcoming is scope — it scopes the constraint, it does not run the ongoing improvement programme a dedicated continuous-improvement platform covers.
FAQ
What is the difference between exploiting and elevating a constraint?
Exploiting squeezes more throughput from the constraint at zero cost: triage, batching, protecting focus time. Elevating means buying new capacity, such as a second reviewer, and only happens once exploitation and subordination still leave demand exceeding capacity. Skipping exploit risks spending money on a constraint that was simply underused.
How long does one full pass through the five focusing steps take in a small business?
A baseline takes a single client visit, and a one-day paid scoping pass can cover mapping, quantifying, and testing one exploit and one subordinate move. After a real change, wait two to four weeks before re-measuring, because old WIP needs time to flush through the system.
Do I need special software to apply the theory of constraints?
No. The core loop needs timestamps from your existing project tool and arithmetic like Little's Law. Simulation tools can model constraint movement, but they demand a full model before you know the engagement is worth quoting. A scoping calculator helps, though a spreadsheet and discipline can carry the first pass.
What happens if two bottlenecks appear at the same time?
One of them governs throughput; the other is just heavily loaded. Identify the step whose capacity actually caps system output, work it through the loop, and predict where the constraint will move next. Once you elevate and the true constraint gains capacity, the second bottleneck usually emerges as the next one to attack.
Are the five focusing steps only for manufacturing, or do they work in services?
They work in any system with a throughput-limiting step, and services are no exception. In a design agency, the constraint was a senior designer's review capacity, not a machine. Knowledge work hides queues inside inboxes and project tools, so you identify the constraint from timestamps and WIP rather than watching physical parts pile up.