Comparing your firm to its industry is a data-mapping exercise, not an analytical one. Pick the classification, pick a published dataset, then spend your effort making your own accounts line up with theirs — because that is the step that decides whether the comparison means anything.
Key takeaways
Three steps. Choose the NAICS industry that actually describes what you sell, not the one that sounds most prestigious. Pull the published benchmark for that industry and the revenue band you fall in. Then map your own general ledger accounts onto the benchmark’s line items, one by one, and record the mapping decisions. A model does that third step well and quickly, and it is the step firms skip, which is why so many benchmarking exercises produce a comparison of two different things.
What comes out is not a verdict. It is a short list of lines where you differ enough from the industry to be worth explaining — and the explanation is usually a deliberate choice you had forgotten you made.
NAICS Canada 2022 Version 1.0 is the classification behind almost all Canadian industry data. It is hierarchical and six digits deep: the first two designate the sector, the third the subsector, the fourth the industry group, the fifth the industry, and the sixth designates national detail, with a zero indicating no further Canadian breakdown. The current version comprises 20 sectors, 99 subsectors, 323 industry groups, 694 industries and 923 Canadian industries, replacing NAICS Canada 2017 Version 3.0.
For a professional firm the choice matters more than it looks. A practice that describes itself as “advisory” may sit in a quite different industry from where its revenue actually comes from, and the cost structures of neighbouring industries are not interchangeable. If your revenue splits meaningfully across two industries, benchmark the two service lines separately rather than averaging yourself into a code that describes neither.
Benchmark datasets use their own line items. Your general ledger uses yours, and yours has been shaped by a decade of ad hoc additions. The comparison is only valid once the two are reconciled, and the reconciliation involves real decisions: whether contractor payments sit in cost of sales or in salaries, whether partner drawings are compensation, whether software is an operating cost or is capitalised, whether recoverable disbursements are revenue at all.
This is where a model earns its place. Give it your trial balance and the benchmark’s line definitions and ask for a proposed mapping with a one-line justification per account and a flag on anything ambiguous. You get a document a partner can argue with in twenty minutes rather than a spreadsheet nobody audits. Keep the mapping and reuse it, so next year’s comparison is a comparison and not a fresh set of judgment calls.
Financial Performance Data gives access to more than 1,000 industries across Canada, with more than 30 performance benchmarks intended to help small businesses see how they measure up. Reports show the number of businesses in the industry, separate statistics for profitable and non-profitable small and medium-sized businesses, detailed revenue and expense data, industry averages for income-statement and balance-sheet items, and financial ratios. They cover revenue ranges of $30 thousand to $5 million and of $5 million to $20 million.
Two caveats matter. Reports carry quality indicators that measure how much imputation contributed to a given value — a heavily imputed line is a weaker comparator, and the indicator is there to tell you so. And the revenue bands are wide, so a firm at the bottom of a band is being compared with firms several times its size. Neither is a reason to avoid the data; both are reasons to write down what you relied on.
Statistics Canada’s own business performance and ownership statistics subject area is the wider starting point if you need something the industry-average tool does not cover.
Realisation, utilisation, effective hourly rate, write-off percentage and average days to bill are not in any national statistical product, because they depend on practice-management data nobody collects centrally. Those are the numbers that actually explain a professional firm’s performance, and they come out of your own systems. Benchmark the financial structure externally; benchmark the operating metrics against your own prior years.
A model asked for “industry average realisation for Canadian law firms” will produce a figure. There is no public Canadian source for it. Do not put it in a partner pack.
A twelve-person management consultancy in Ontario wants to know whether its cost base is out of line before it takes on two more staff.
It selects the NAICS industry matching its actual service mix, notes that a fifth of revenue comes from a different industry and sets that segment aside rather than blending it. It pulls the Financial Performance Data report for its industry and revenue band, and checks the quality indicators first: two of the lines it cares about are lightly imputed and one is not, and the third is flagged in the write-up as weaker evidence.
The mapping exercise takes an afternoon with a model and a partner. Three decisions are contentious — how subcontracted associates are classified, whether principal compensation is normalised, and where software subscriptions sit. All three are recorded with a reason. Once mapped, the firm sits close to the industry on most lines and clearly above it on one. The explanation turns out to be a deliberate choice about seniority mix made four years earlier, which the partners had stopped noticing. The conclusion is not “we are inefficient”; it is “we are choosing to be different here, and here is what it costs”.
Repeat annually with the same mapping. The trend against the industry is worth far more than any single year’s gap.
This page is written for a firm that delivers work to a book of clients. If the question is really about the front desk — intake, scheduling, recall and reminders — that lives on the professional practice owners page. If it is about your own month-end, reconciliation and payables rather than what you deliver to clients, that is accounting automation, or bookkeeping automation for coding and receipt capture. The tooling overlaps; the risk almost never does.
Can AI pull the StatCan numbers for me automatically?
Be careful. Many statistical tables are rendered interactively, so a model summarising “the table” may be summarising nothing. Pull the report yourself, save it with its date and vintage, and give the model the file. Then the citation is to a document you hold.
Which NAICS code should a mixed practice use?
Split it. Benchmark the service lines separately against the industries they belong to, and only combine at the end if you must. A blended comparison against a single code hides the thing you were trying to find.
How often is this worth doing?
Annually, with the mapping reused. The first year is the expensive one because the mapping has to be built and argued about; after that it is an afternoon and the trend is the output.
A 30-minute call is enough to tell you whether AI pays for itself here.