Software reviews usually describe features. Features are on the vendor's own website, written better, by people with more time. They are not why you are here.
What nobody publishes is how long the thing takes to set up, what it actually costs at your headcount once the add-ons land, and how hard it is to leave. Those are measurable. So we measure them.
The ten steps
Every tool is run through the same protocol, in the same order, with a stopwatch running and a screenshot taken at each step.
- Signup to empty inbox. Clock starts at the signup button, stops when an empty ticket queue is visible. We record elapsed time, how many form fields were required, whether a credit card was demanded up front, and whether a sales call was forced before access.
- Time to first ticket. We email the support address from an outside account and stop the clock when it appears as a ticket. This is our headline number, and as far as we can tell nobody else publishes it.
- Email channel connection. We connect a real custom domain and record the elapsed time, the DNS records required, and whether it worked first time.
- One automation. The same rule in every tool: auto-assign tickets containing "refund" to a named agent and tag them. We record the time taken and whether it required a paid tier.
- Canned response and one knowledge base article. One of each. We record the time and whether the knowledge base is bundled or sold separately.
- Second agent invited. Elapsed time, and what it adds to the bill.
- Report retrieval. Produce first-response time for the last seven days. We record the time, whether a higher tier was needed, and whether the number was actually right.
- AI feature audit. Every AI feature, what it costs, and whether it is billed per agent, per resolution, or bundled. Screenshots of the pricing. This is where the money hides in 2026.
- Support channel test. We send every vendor the same question — what happens to my ticket history if I cancel? — and record the channel offered, time to a human reply, and whether the answer was correct. A support tool with bad support is a finding.
- Cancellation and export. We cancel and export. We record elapsed time, how many steps, whether cancellation was self-serve or required contacting sales, and what format the data arrived in.
The score, and its weights
Six dimensions, each 0 to 10. The weights are published because a reader who cares more about cost than about reporting should be able to recompute our answer and disagree with it.
| Dimension | Weight | What it measures |
|---|---|---|
| True cost at 10 agents | 25% | The real monthly bill including add-ons, not the headline tier |
| Setup speed | 20% | Steps 1 to 3, weighted toward time to first ticket |
| Everyday usability | 20% | Steps 4 to 6 — can a non-technical agent do the common jobs |
| Reporting quality | 15% | Step 7, including whether the numbers are correct |
| AI value for money | 10% | Step 8 — capability against what it actually adds to the bill |
| Exit friendliness | 10% | Step 10 — cancellation and data export friction |
What we will not do
We will not claim a test we did not run. If a tool has not been through the protocol, its page says not yet independently tested and the comparison is explicitly based on published vendor information instead.
We will not fill a gap with a plausible number. Where a vendor does not publish a price or a term, this site shows "not verified" rather than an estimate dressed as a fact.
We will not quietly refresh a measurement. Every number carries the date it was taken and the plan it was taken on. When pricing changes, the page is marked stale until it is re-tested.
We will not let commission move a ranking. The weights above are the ranking.
Labels you will see
Each factual claim on this site carries its provenance, so you can tell measurement from repetition:
- Tested — measured by us under this protocol, with a date and the plan tested.
- Verified — read from the vendor's own page or a frozen dataset, with a date.
- Vendor claim — the vendor says so; we have not confirmed it.
- Not verified — we do not know. We say so rather than guess.
Re-testing
Every tested tool is re-run at least every six months, and immediately when a vendor changes pricing or ships a major release. Support software ages fast, and a two-year-old setup time is fiction.