Common workflow bottlenecks (and how to fix them)

The usual suspects

• Weak handovers — work passes between people with missing information, causing rework and delay
• Waiting on the client — jobs stall for records with no system to chase and track
• Partner review queues — everything funnels through one or two reviewers
• Inbox-managed work — jobs tracked in email and memory rather than a workflow system
• Unclear priorities — the team unsure what matters most this week

How to clear them

Standardise handovers with a simple checklist of what must be complete before work moves on. Systematise client chasing so it doesn’t rely on someone remembering. Spread or streamline review so it isn’t a single choke point. And move work out of inboxes into a system that shows status at a glance.

Finding and clearing bottlenecks is one of the first things an Optivo engagement tackles — it’s usually the fastest capacity win available.

How to find yours, rather than guess

Every practice recognises the list above, which is precisely the problem: recognition is not diagnosis, and the bottleneck that is costing you most is rarely the one that annoys you most.

The method takes an afternoon. Pick your highest-volume job type and follow three real files end to end — not from memory, from the record. For each one, write down every point at which the job stopped moving, and how long it sat there.

What you are looking for is not the slow step. It is the waiting. In most practices the total working time on a set of accounts is a few hours and the elapsed time is several weeks, and almost the whole difference is queueing: waiting on records, waiting on a query, waiting on review, waiting on a signature. Speeding up the working time attacks the small number. Removing a queue attacks the large one.

The four queues, and what actually clears them

Notice that three of the four are cleared by writing something down rather than by working faster. That is the general shape of workflow improvement in professional services, and it is why it so rarely happens: there is no satisfying moment of effort attached to it.

The query queue is the cheapest win

Worth singling out because it is almost universal and almost free to fix.

A preparer working through a file finds a question and sends it. Twenty minutes later, another. The client answers the first at lunchtime and the second the next morning, by which point the preparer is on another job and has lost the context. Each query costs the sending, the waiting, the context switch, and the reload.

Batching changes this to one message, one answer, one return to the file. The rule is simple — no query leaves until the file is worked to the end, then all of them go together, numbered — and the objection to it is always the same: but then the job sits longer. It does, by a few hours. It also arrives back complete, which is worth several days of the alternative.

The inbox is not a system, and the symptom is the question

Jobs managed by email have a distinctive tell: somebody has to ask where something is. Not once, routinely. “Did that go out?” “Is Smith Ltd waiting on us or on them?”

Every one of those questions is two people's time spent reconstructing a fact the system should hold. In a practice of eight that happens dozens of times a week, and it never appears on any report because nobody codes time to “finding out”.

The fix is not necessarily new software — most practices already own something that does this and use a fraction of it. It is that job status has to be recorded in one place, by whoever changes it, at the moment it changes. That is a habit, not a purchase, and it is why practice management is a job rather than a system.

The order to do them in

Bottlenecks move. Clear the review queue and the constraint shifts to records; clear records and it shifts to approval. That is not failure, it is how constraints behave, and it is the reason to work on one at a time rather than launching an improvement programme.

1. Start where the elapsed time is longest, not where the frustration is loudest. Usually waiting on the client.
2. Fix it for one job type only. One genuinely under control beats eight half-documented, and it produces a template for the rest.
3. Measure the elapsed time again after six weeks. If it has not moved, you fixed the wrong queue — which is useful information, cheaply bought.
4. Then take the next one. Expect a quarter per queue, and expect the fourth to be much faster than the first because the practice has learned how to do this.

One caution about timing: the records queue is about to get considerably harder for most practices, because MTD for Income Tax increases the number of times you must extract records from an affected client from once a year to five. A practice that fixes that queue before April 2026 absorbs the change. One that leaves it will meet it five times a year instead of once.