# Compare a client workflow fairly

Version 1, 29 October 2026. A blank test protocol, not published benchmark results. Use first-client-project.md and the same fictional records in each product. Flori authored this protocol; no competing product was hands-on tested for this kit.

## Set the conditions

Record product, plan, billing interval, region, date, tester and any enabled add-ons. Use the same number of internal people and client viewers. Separate initial account/configuration time from the time to repeat an already configured workflow. Record prior experience with each product. Do not call a free trial's access the free plan's feature set.

## Tasks and evidence

| Trial | What to record |
| --- | --- |
| Import 1 account, 2 people and 1 deal | Field mapping, created/skipped counts, retained relationships, warnings |
| Start delivery from the deal | Manual steps; which scope and contacts survive; project and task links |
| Share a deliverable with the client | Required plan; exactly which internal content is exposed; client account requirement |
| Record a rejected then accepted draft | Version, approver, timestamp and how unresolved work remains visible |
| Recover context in a new assistant session | Connection/plan prerequisites; whether a saved decision is retrieved accurately |
| Limit the assistant to one project | Scope offered; known out-of-scope record denied; read-only write refused |
| Finish a handover | Open issues, owner, access confirmation and support boundary retained |

For each trial mark Pass, Fail, Not supported or Not tested; attach redacted evidence. A documentation claim is not a passing test. Stop if you would need to use real customer information or change a paid subscription without the account owner's authorization.

## Cost worksheet

Calculate required plan price times billable seats, applying seat minimums. Add mandatory add-ons and the assistant subscription or API cost for each actual user. Show annual cash commitment separately from monthly equivalent. Include tax and regional terms when known; otherwise mark them excluded. Client viewers may be billed differently from internal collaborators.

Record results even when another product performs better. Do not publish a fastest, cheapest or easiest winner from an unrun trial, unequal plans, an unverified checkout or one unfamiliar tester's first attempt. A useful conclusion can be "keep the existing tool".