How to Identify Bottlenecks in Professional Services Delivery

By Two Pillars & Co Advisory Group · Sources checked October 2026

What is a bottleneck in a services firm?

A bottleneck is the step in a process that has less capacity than the steps around it, so work piles up in front of it and later steps wait. The Theory of Constraints Institute puts the principle bluntly: strengthening any link of a chain apart from the weakest is a waste of time and energy, and most initiatives to improve something in an organization fail to produce results because they are not aimed at the constraint.^[1] In a professional services firm the constraint is rarely a machine. It is usually a person or an approval step: a partner who reviews every deliverable, a specialist whose skills are needed on most projects, a client sign-off that always arrives late, or an intake process that cannot absorb more engagements.

The symptoms are familiar. Work is waiting for review, deadlines slip even though staff are busy, some people are overloaded while others have slack, and urgent requests keep jumping the queue. Lean’s vocabulary describes the damage: muda is any activity that consumes resources without creating value for the customer, mura is unevenness in an operation, such as an uneven pace that makes people hurry and then wait, and muri is overburdening people by requiring them to work at a higher pace for longer than appropriate.^[2] A bottleneck tends to produce all three at once.

How do you map the workflow and find the constraint?

Start with a simple map of how a typical engagement moves from first enquiry to final payment. List each step, who does it, how many hours it takes and, importantly, how long the work waits before the step begins. Typical steps include proposal, scoping, contract, kickoff, information gathering, analysis or drafting, internal review, client review, revision, final delivery and invoicing. Use your time and project data to estimate the average hours of effort and the average waiting time at each step for the last 20 to 30 engagements.

Two measurements help most. First, the elapsed time at each step, because queue time is usually much longer than touch time. Second, the capacity of each step in engagements per month. The step with the least capacity relative to demand is your candidate constraint. Check it by asking whether finishing more work at that step would immediately raise the number of engagements the firm can complete.

What should you do once you have found it?

Use the five focusing steps. The Theory of Constraints Institute describes them as identify the constraint, exploit it, subordinate everything else to it, elevate it and then prevent inertia by returning to step one.^[1] Exploit means getting the most from the constraint with what you already have: protect its time from meetings and low-value tasks, make sure it is never starved of inputs, and have it work only on what only it can do. In a firm where a partner’s review is the constraint, that may mean fixing what arrives for review (complete, checked and in a standard format), delegating routine review to a qualified manager and scheduling review time as a protected resource.

Subordinate means aligning the rest of the system to the constraint: do not release more work than it can handle, and do not push urgent exceptions through ahead of planned work. The institute notes that left unchecked, non-constraints produce bloated work in progress, elongated lead times and frequent firefighting.^[1] Only after exploiting and subordinating should you elevate the constraint by adding capacity, for example by hiring or training. The institute warns that people tend to gloss over the first three steps and jump straight to elevation, and says that implementing them properly typically exposes a minimum of 30 percent hidden capacity within the first few months; that figure comes from the institute and from manufacturing settings, so treat it as a prompt to look, not as a forecast for your firm.^[1]

How do you keep the constraint visible and stop it returning?

Put the status of the process where everyone can see it. The Lean Enterprise Institute defines visual management as the placement in plain view of tools, activities and indicators of system performance, so that the status of the system can be understood at a glance by everyone involved.^[4] For a firm, that can be a board or dashboard showing every engagement, its current step, how long it has waited and who owns it, with a clear marker for items waiting on the constraint.

Track three measures weekly: the queue in front of the constraint, the average time engagements spend there and on-time delivery. Also track overruns; SPI Research finds that overruns above 10 percent damage client satisfaction, project margins and future bookings, and that Level 5 firms hold overruns to single-digit percentages.^[3] When the numbers improve, the constraint will often move to another step, which is expected. Return to step one and repeat, and avoid leaving the earlier fix to decay.

What mistakes should you avoid?

The first is treating everything as urgent. If every project is a priority, the constraint is allocated by who shouts loudest. Agree rules for priority. The second is measuring busyness: high utilization at non-constraint steps can simply produce more work waiting at the constraint. The third is hiring first. Adding people to a process whose real problem is queueing or unclear scope can raise cost without raising output. The fourth is ignoring rework: work that returns to an earlier step consumes the capacity of the constraint a second time.

The last is solving the wrong problem. A bottleneck may be in sales (proposals take too long), in client behavior (late information), or in a decision rule (a policy that forces every document through the same reviewer). The mapping exercise tells you which; the fix then follows from the data.

Frequently asked questions

What to do next

1. Map one typical engagement from enquiry to payment with times for each step.
2. Measure queue time and capacity at each step for the last 20 to 30 engagements.
3. Name the constraint and protect its time.
4. Set up a simple board that shows every engagement and where it is waiting.