Hey, I'm Damon Millar. I'm a UK accountant, fractional CFO and business advisor. I run a practice built on quarterly reviews and proactive planning rather than filing returns after the fact.

Three software companies have published research on the state of accounting practice recently. All three get quoted at conferences and in LinkedIn posts. Almost nobody reads past the headline number.

I did. Including the methodology sections, which is where the actual information is.

Every one of these is a marketing document. That isn't an accusation, it's just what commissioned research is. But a marketing document with a real sample still contains real information, and the skill is telling the two apart.

Every source is linked at the bottom.

Want to implement AI systems in your pratice?

<aside>

Book a 30-minute call with me: https://calendly.com/damon-millar-switchaccountants/30min

</aside>

How to read a vendor report in 60 seconds

Do this before you read a single finding. It takes a minute and it changes what you take away.

1. Find the sample. Who did they ask?

This is the whole game. If a software company surveyed its own customers, every question about that category of software is unusable. Not exaggerated. Unusable. The sample cannot produce a negative answer.

2. Find the n and the fieldwork dates.

An undated survey is a survey you can't place in time. A survey with no stated n is a survey with something to hide. Both happen more than you'd think.

3. Ask which direction the bias runs.

This is the useful move, and it's why "it's marketing, ignore it" is lazy. Bias has a direction. A survey of software customers is biased towards liking software. It is not especially biased towards reporting higher revenue, or towards a particular view on staff retention. Some findings in a biased sample still stand up. Work out which ones.

4. Separate the commissioned study from the ROI calculator.

Vendors often publish a proper third-party study and a set of product performance claims in the same press release. The first has a methodology. The second usually has "adjustable assumptions". They are not the same kind of object and they should not carry the same weight.

5. Then read the findings.