Skip to content
← All posts
7 min readBy

Reviewing write-offs on fixed-fee tax jobs

A write-off on a fixed-fee tax job has four different causes. Code each one separately, or next year's quote repeats the same loss with a new excuse.

  • Practice management
  • Pricing
  • Review
A flat illustration on sage of a large navy coin split into four wedges, one coral wedge pulled slightly away from the others.

It's the second week of November and the work-in-progress report has finally been run. You scroll down the write-off column with a coffee going cold beside you. Forty-two jobs. Most of them a little under. A handful well under. One family trust where the time on the file is roughly double the fee, and the only note in the job is WO - see Sam.

Sam left in August.

Most small practices review write-offs once a year, if at all, and they review them as one number. The total is down on last year, or up, and someone says the fees need to go up by a bit. That is a pricing decision made from a figure that mixes four unrelated problems together.

My position is simple: a write-off on a fixed-fee job is only useful if it says why. Without a reason, the number can't tell you whether to raise the fee, change the engagement letter, fix something in your own workflow, or leave it alone because you meant to give it away.

Four causes that look identical in the WIP report

The quote was wrong. The job was what the client said it would be and the records were what you expected. It simply took longer than you priced it at. Two rental properties quoted like one, or a company return priced from last year's fee when last year was the easy year.

The records were not what was quoted on. The job changed underneath you. A second bank account turned up in April, the bookkeeper stopped in November, the client "forgot" the share sale. This is the scope problem, and if your practice has a stop point on fixed-fee work, most of these should have become a separate conversation before they became a write-off.

The practice did the work twice. A preparer built the schedule, the partner sent it back, and the second pass took longer than the first. Or a file sat for six weeks and the person who picked it up again had to read it from the beginning. The client did nothing differently, and the time came from inside your own process.

Someone decided to absorb it. A long-standing client, a hard year for them, a referral source you want to keep. This is a perfectly good commercial decision. It only causes trouble when nobody wrote it down and it arrives next year disguised as the normal cost of that client.

Four short stacks of cream paper side by side, each tied with a different coloured band, on a plain light surface.
One column in the WIP report, four separate problems. Each needs its own owner.

Code the reason when the time is written off

That doesn't need a new system. It needs a short, closed list of reason codes and a rule that a write-off isn't processed without one. Four codes, matching the four causes above, plus a free-text line for the sentence that explains it:

CodeMeansSentence that should go with it
UnderquotedSame job, more time than priced"Two properties, priced as one."
RecordsThe job changed after the quote"Second account found in April; client declined a separate quote."
ReworkTime spent redoing our own work"Schedule rebuilt after review; first pass used the wrong period."
ConcessionA deliberate decision to absorb"Partner agreed to hold fee this year. Review next year."

Keep the list closed. A practice that lets people type their own codes ends up with "other", "misc", "client" and "see notes" within one season, which tells you exactly as much as no code at all.

And the code belongs to whoever decided the write-off, not to the person who logged the hours. A preparer can tell you the file was a mess. Only the person who agreed the fee knows whether the mess was in what was quoted on.

Split a mixed write-off rather than picking one cause

The comfortable one is usually Records. It puts the blame on the client, and that might well be partly fair. But if a third of that overrun was the schedule being rebuilt after review, coding all of it as the client's fault hides the one part of the problem you can fix yourself.

Rough thirds are fine. What matters is that next November you can see rework as its own line, instead of finding it spread across every other code.

What each code tells you to change

Underquoted jobs go to whoever sets the fees. Look for the pattern: rental properties, a particular entity type, clients with a new sale or a new activity that year. The fix is in the quote, and often in what you ask for before quoting. If you are pricing a messy file from what the client says about it, a paid diagnostic is a cheaper way to find out.

Records write-offs go to the engagement process. Each one is a question: was there a point where the change could have been raised with the client, and why wasn't it? Sometimes the answer is that it was raised and the client said no. Fine. Sometimes it is that the preparer didn't feel able to stop, and that one is yours to fix.

Rework goes to whoever runs review. This is the code that a busy partner least wants to look at, and it is the one the practice has the most control over. Read the files with the biggest rework write-offs and ask what came back and why. The answers are usually boring and fixable: a figure with no source, an empty prior-year column, a judgement call buried in a total. The same things that send workpapers back from review in the first place.

Concessions go back to the partner who gave them, with one question: still meant? Some are, and the client relationship is worth it. Some were a one-off for a bad year that has now quietly lasted three years. The point of coding them is that you are making that decision again on purpose.

A coral sticky tab marking one sheet in a thick stack of cream pages, the rest of the stack plain.
A concession recorded is a decision. A concession forgotten becomes the price.

Look at the client, not just the job

Sort the coded write-offs by client and by year. The useful rows are the clients who show up every year under the same code. A client who is Records every year is a client whose bookkeeping you are paying for. A client who is Concession every year has a fee nobody has told them about.

Either can be worth it. Decide that with the number in front of you, before someone does the maths in a slow week and finds out.

Make the reason travel with next year's file

Put a short note on the client's file for next year: what ran over, which code, and what was agreed. "FY25 overran on records: second account. Quote FY26 with both accounts. Ask for statements before starting." The preparer opening the file in March then knows what last year cost, without anyone having to remember it.

The same goes for concessions. "Fee held at last year's level as a one-off" written on the file stops it becoming the price by default. It also gives the partner a natural time to raise it with the client, before the work starts rather than after the invoice.

Two cream folders on a plain surface, the front one with a small navy note card clipped to its cover.
Last year's reason, on the front of this year's file.

Where software fits, and where it doesn't

Software can shrink the Rework code. Our workpapers link each figure to the page it came from and raise unclear items as review points, so a file that comes back from review comes back for a reason someone can read. The workpaper templates show the schedules we work from.

The other three codes are decisions people in the practice make, and no tool makes them for you. The FAQ covers what the product does and doesn't do.

The write-off review is one afternoon a year. Run the report, code what isn't coded, and put a note on next year's file for every client that ran over. Then do something about Sam's family trust before it happens again.