3. Feature-by-Feature Scrutiny (The Stated Idea)
Categories:
- Table stakes (TS): customers expect it. It earns nothing, but its absence costs you.
- Differentiator (D): a real reason to choose you over the alternatives.
- Distraction (X): built but unlikely to move revenue or retention.
- Liability (L): creates support, compliance, legal, platform or trust risk.
Some items sit in two buckets. The primary bucket comes first.
3.1 Classification table
| # | Stated element | Category | Reasoning |
|---|---|---|---|
| 1 | Time-in-status reports | TS | 12+ Marketplace apps, a free one with ~10.5k installs, and a native (limited) gadget. It earns nothing on its own. Keep it inside the audit job: "time in approval", "time between Approved and Done". |
| 2 | SLA reports | TS → D only for audited service desks | JSM has native SLAs, and Appfire's Time to SLA has ~3.4k installs. It becomes a differentiator only as SLA evidence: breaches with reasons, approvals on breached tickets, for an ISO 27001 or SOC 2 availability control. Phase 2. |
| 3 | Cross-project roll-ups of estimates | X | JXL, Structure, Jira Plans (Premium), eazyBI. It's a PMO job and a different buyer. Agile Docs built a business on this in 2019–20, but the category has filled up since. |
| 4 | Cross-project progress roll-ups | X | Same as #3. Portfolio dashboards are a sales-led, political purchase. |
| 5 | Audit and compliance exports ("who changed what and when") | TS as raw export / D as control evidence | Raw history export is already sold (SaaSJet Issue History, 3,060 installs, with HIPAA and FDA Part 11 claims). The differentiator is turning it into evidence: approval by a second person, skipped-step exceptions, sampling per audit period, a tamper-evident export. This is the product. |
| 6 | Confluence version later | X (for now) | A second product with its own competitors and its own history APIs. Consider it only after $10k MRR, and only if paying Jira customers ask. |
| 7 | Built on Forge | TS (mandatory) | It's what qualifies you for 100% revenue up to $1M lifetime. Connect loses support on 31 Jan 2027, so every competitor is moving here too. It's a requirement, not an edge. |
| 8 | Sold on the Atlassian Marketplace, per-user pricing | TS / D for distribution | The right channel: built-in discovery, billing and trust. Per-user pricing lets revenue grow with the customer. But you don't control ranking. Treat Marketplace search as rented land. |
| 9 | Buyers: "Jira admins and PMOs" | L (as positioning) | Two buyers, two jobs, two messages. The listing can't rank or convert for both. Pick the admin or compliance lead at an audited company. |
3.2 Scoreboard
| Category | Count | Elements |
|---|---|---|
| Table stakes | 4 | Time in status, raw history export, Forge build, Marketplace listing |
| Differentiator | 2 | Change-control evidence (from #5), SLA evidence for audited desks (from #2, phase 2) |
| Distraction | 3 | Roll-ups of estimates, progress roll-ups, Confluence version |
| Liability | 1 | "Admins and PMOs" dual positioning |
About a fifth of what's described is a reason to buy. The rest is table stakes or distraction.
3.3 Feature bloat and "building to avoid selling": explicit call-outs
- Four report families at launch. Each one is 3–6 weeks of Forge work, plus edge cases (workflows with loops, renamed statuses, deleted issues, business calendars). Four at once is a year of building before you know if anyone pays.
- Roll-ups. They look impressive in a demo and they serve a buyer you're not selling to.
- Confluence "later". Fine as a note. Dangerous as a plan, because it invites building a second app before the first one ranks.
- Dashboards and charts polish. Incumbents have 15+ report types and years of polish. You can't win on charts. Win on a report the auditor accepts.
3.4 Platform and engineering realities (Forge, 30 Sept 2026)
| Constraint | Verified figure | What it means for this app |
|---|---|---|
| Revenue share | 0% up to $1M lifetime Forge revenue, then 16% (from 1 Apr 2026) and 17% (from 1 Oct 2026) | Must qualify as a Forge app: no Connect modules, OAuth, Forge UI. |
| Jira API rate limit | 65,000 points/hour shared across all installs (Tier 1 default). 1 point per request plus 1 per issue returned, 2 per user or group. Tier 2 per-tenant pools after review. | Don't rescan changelogs. Capture events as they happen and backfill slowly. Plan to apply for Tier 2 once you have ~100 installs. |
| Function time | 25 s for user-led calls, 900 s for async events and scheduled triggers | Build evidence packs in async jobs, and show progress in the UI. |
| Memory and payload | 512 MB default, up to 1,024 MB. 5 MB payload. Front-end requests 500 KB. | Paginate and stream large exports. Generate files in chunks. |
| Storage | 1 GiB per production install (secondary source, medium). KVS values up to 240 KiB, 4,000 reads and writes a minute per install. Forge SQL available. | Store derived facts (transitions, approvers, timestamps), not full issue copies. |
| Platform billing | Since 1 Jan 2026, with free monthly allowances (for example 200,000 GB-seconds of function time) | Watch usage per install. The largest sites drive cost. |
| "Runs on Atlassian" badge | Atlassian-hosted compute and storage only, no egress except analytics, matching data residency | Rules out external email services, Slack, and remote backends, unless you accept losing the badge. |
3.5 What a disciplined v1 looks like
Keep: change-control evidence for one or more Jira projects:
- approvals, and who approved;
- self-approval and same-person exceptions;
- skipped required statuses;
- time in approval;
- a random sample of changes per audit period;
- XLSX and PDF export with a content hash;
- a monthly control-check summary inside Jira.
Defer: SLA evidence for service desks (phase 2), scheduled delivery to Confluence pages, and multi-framework control mappings.
Cut: roll-ups, portfolio dashboards, generic time-in-status reports as a product, the Confluence app, and a Data Center version.