How we researched this.
ClearLogic’s accounting and bookkeeping research draws on publicly posted customer reviews on Google — the reviews attached to a firm’s Google Maps and Google Business Profile listing. That is a different source from our trades, property maintenance and construction research, and the data has a different shape, so this page states exactly how we collected, scored and filtered it, and where it differs from our other work. (Our trades, property maintenance and construction methodology pages each cover their own separate corpus.)
Public Google reviews, read as any member of the public could.
We read only what any member of the public can see on a firm’s Google listing. We did not log in, create an account, contact any firm, or access anything that wasn’t already published.
We started from a list of 485 New Zealand accounting and bookkeeping firms with a public business address, compiled from public business listings (not from reviews, and not from any client or enquiry data). Each firm was matched to its Google listing by name and address; 484 of 485 matched, and every review on each matched listing was collected. The collection was done through a commercial data-collection service that reads public listings, on 1 October 2026.
| Count | |
|---|---|
| Firms sampled | 484 |
| Firms with at least one public review | 374 |
| Firms with no public review at all | 110 (22.7%) |
| Reviews collected | 7,571 |
Scope is New Zealand only. Five regions had enough firms to report on separately (Auckland 155, Canterbury 85, Wellington 51, Bay of Plenty 45, Waikato 36); firms elsewhere are included in the national figures only.
This is a cross-sectional snapshot of each firm’s full public review history as it stood on 1 October 2026. It is not a rolling monitor and does not update automatically — if we republish an updated version later, we’ll date it separately from this one.
One star rating, not four sub-scores.
Our trades, property maintenance and construction research used a review platform where each review carries four named sub-scores (communication, quality, reliability, value) on a 0 to 100 scale. Google reviews carry one rating, 1 to 5 stars, plus free text. So there is no “lowest-scoring dimension” finding available here; the comparable signal is the star-rating distribution (average, share of five-star, share of two-star-and-below, spread between firms) and, as in the other studies, what reviewers choose to praise or complain about in the text. We kept the same two-stage coding method so the praise and complaint findings are built the same way, but the categories themselves were written for professional services, not carried over from trades.
How reviews are scored, and how we coded them.
The rating
Each review’s 1 to 5 star rating, as published by the reviewer. We didn’t adjust it. A negative review is defined as 2 stars or fewer; a positive review as 4 stars or more. 6,190 reviews were positive by that definition, 328 negative.
Firm size
Google doesn’t publish staff counts, so “size” here is the firm’s public review count split into three equal groups (terciles): small, mid and large. It’s a proxy for public footprint, not headcount, and the pages that use it say so.
Complaint coding (6 categories)
Negative reviews were coded against a fixed list of six reasons written for accounting and bookkeeping: overcharging or fee disputes, unresponsive, errors or mistakes, missed deadlines, unclear advice, rude or unprofessional conduct. Coding was by deterministic keyword and phrase matching, fixed before the data was read.
Praise coding (11 categories)
Positive reviews were coded against a fixed list of eleven traits: responsiveness, clear explanation, proactive advice, accuracy, value for fees, accessibility (easy to get hold of), deadline reliability, politeness, problem solving, referral language (“would recommend”), and repeat-use language (“will use again”). Coding was by an LLM classifier (Claude Haiku 4.5, run in batches) because praise is phrased far more variably than complaints. The taxonomy was fixed in advance of coding, not built from the results. A review can carry more than one trait, so percentages don’t sum to 100.
A statistic only gets published if it clears two thresholds.
- 01At least 30 reviews behind any single figure (a region, a firm-size group, or similar breakdown), so a figure isn’t resting on a handful of opinions.
- 02At least 5 different firms contributing to that figure — our own addition, on top of the review-count threshold, because a stat built from one to four firms’ reviews can be traced back to roughly which firms it describes even with no name attached.
Nothing in this corpus failed either threshold. Every regional and size breakdown we report clears both, with the thinnest cell (Waikato) at 36 firms. This is the first of our four industry corpora with no suppressed cells. One figure is flagged rather than suppressed: the small-firm group’s praise percentages rest on 34 positive reviews, which clears the floor but only just, and every page that uses them labels them as directional.
Missed-deadline complaints did not reach the reporting threshold in any cell, so no figure for them is published anywhere, and pages that discuss deadline reliability say so.
No names, ever.
No firm name, owner name, staff name, or individual review appears in anything we publish. Every figure is an aggregate across many reviews and many firms. No verbatim review text is published or linkable from our accounting research. Reviews of accounting firms often name the individual accountant, and sole practitioners often trade under their own name, so the anonymisation pass for this corpus was done twice: an automated pass, then a hand check of every extracted quote for personal names and trading names the automated pass missed. Quotes were used only to check the coding; none are published.
Read the numbers with these in mind.
- —This is public reviews, not a random sample of every client relationship. People with an unremarkable experience are less likely to leave a review than people who were delighted or upset, so the corpus skews toward strong opinions in both directions.
- —It skews strongly positive: 92.6% of reviews are five-star. That leaves a negative sample of 328 reviews, and four of the six complaint categories have single-digit review counts. We report complaint figures as rank order, not precise measurement, and we don’t build a headline on them.
- —Professional services are reviewed far less often than trades. The median firm here has four reviews, and 22.7% have none. A firm with no reviews isn’t a bad firm; it’s an unmeasured one, and that gap is itself one of the findings.
- —“Size” is review count, not headcount or revenue.
- —Any regional difference is a description of this corpus, not a tested explanation of why.
All 172 figures computed from it were reviewed and approved for publication on 7 October 2026. This page will be updated if the underlying data is refreshed. Last updated: 8 October 2026.
