Trust
How we verify pricing
Pricing is the most volatile and most contested fact about software. Here is exactly how ours is obtained, and the four rules that constrain it.
The four rules
- Read it on the vendor’s own page. No price on this site came from a roundup, a review site or a summary. If we did not load the vendor’s page, there is no price.
- Record the date. Every figure renders with the date it was read. A price without a date is a claim about the past presented as the present.
- Never convert currencies. CHF 700, EUR 845, GBP 35, INR 6,999 and AUD 6 stay exactly as written. A converted figure is a number we invented, and rates move.
- Preserve the unit and the qualifier. “As low as $34,000”, “80c per project”, “from $95 per new system” and “+ 18% GST” are kept whole. A decimal column would discard the part that matters.
One consequence: where a vendor shows a bare “$” and never states the currency code, we
record the ambiguity rather than resolving it for them. One product on a .com.au
domain does exactly this.
The control
A “no price published” result means nothing unless the method is known to return prices when they exist. So every verification pass opens with a control before any negative is recorded:
- Positive controls. Two products whose records say they publish. Both returned real figures — CHF 700/year and $159/user/month.
- Negative control. One product whose record says it does not publish. It returned no figure and a “Talk to sales” call to action.
Only then is a “not published” admissible as evidence of absence. This exists because the recurring failure mode on this project was trusting a negative from a method never confirmed to return positives — it produced four wrong findings before the control was introduced.
When no price exists
51 of the 87 products whose pricing state we established publish nothing. Those pages say “Pricing not publicly disclosed”, name the date we checked, and quote the call to action the vendor offers instead. That is the answer to the query “what does X cost”.
Where a vendor’s surface could not be read at all — a bot block, an expired certificate, a login wall — we record “could not read”, name the failure, and exclude the product from every percentage. 10 products are in that state.
Why we refuse some comparisons
12 distinct pricing models are in use. Within a model, figures compare. Across models they do not, and the comparison engine refuses rather than inventing an assumption about your project volume. A page that tells you a per-project fee is “cheaper” than a per-seat subscription is guessing on your behalf.