How we rate help desk software
One method for every vendor. Scores run from 0 to 100, and every page states its method version.
The test inbox
Our test inbox is a fixed set of 25 tickets: billing questions, bug reports, refund requests, how-to questions, an angry customer and tickets in other languages. Each help desk receives the same set on a trial account. We record how long setup takes from signup to the first routed ticket, and compare each AI suggested reply with a reference answer.
Method versions
v1 scores rest on documented evidence only: pricing pages, plan comparisons and product documentation. The two measured dimensions, setup time and AI reply accuracy, show as "not yet tested" and are left out of the total, with the other weights scaled up to 100.
v2 adds the test-inbox results. A page moves to v2 once every tool on it has been through the inbox, and the page says what changed.
What we score
- Channels covered. Email, live chat, social, WhatsApp, SMS and phone, and whether each one lands in the same inbox.
- AI reply accuracy. How often the AI suggested reply matches a reference answer on our test inbox.
- True cost for 5 agents. The monthly price for five agents with the features the page's team needs, add-ons and AI fees included.
- Automation and SLA rules. Routing, assignment, SLA timers and escalation.
- Knowledge base. Help center quality and how well it feeds the AI.
- Integrations. Shopify, Slack, CRM and the other tools the page's team uses.
- Data hosting and compliance. EU, UK and Australian hosting options, GDPR and security documentation.
- Reporting. First response time, resolution time, CSAT and agent reports out of the box.
How weights are set
Each ranking weights these dimensions by what people searching for that page ask about most. A ranking for ecommerce teams weighs Shopify integration more than one for B2B software companies. When we change a ranking's weights, its method version changes and the page says what moved.
Where the evidence comes from
Official pricing pages, product documentation, marketplace listings and our own tests, each with the date we checked it. A figure we could only find second-hand is written as "about $X". A marketplace listing with fewer than ten reviews counts as unproven.
How ties break
When two tools score the same, the one that costs less for a small team ranks higher, and the page says so.
How we keep pages current
We recheck prices and free-plan limits every quarter and confirm every listed vendor is still trading every month. A vendor that stops answering moves to a "no longer recommended" note instead of disappearing.