HelpDeck
As described:StopNarrowed:NarrowNo as described. Weak Go, only as a Forge help-center app for Confluence, built as a second Forge app.
software teams · ~$50 a month per customer · last reviewed (next review by 29 Dec 2026) · 25 cited sources
Start this week
- Decide whether this would be your first or second product, and tell me your hours per week and whether you've shipped a Forge app. If it's your first and you haven't, pick StatusLens or DocPress instead.
- Trial Scroll Sites and Refined Sites for an hour each. Read every review of 3 stars or fewer. Write down 10 gaps in customers' own words.
- List 30 software companies that use Confluence and run a separate public help center (look for Zendesk, Intercom or Help Scout in their help-center footer). Book 10 calls.
- Run a Forge spike: publish one Confluence page tree to a public URL. Find out whether it needs an external remote or can run on Atlassian alone.
Checkpoints and kill gates
Decide in advance what counts as on track. Missing a checkpoint means changing the buyer or the offer, not adding features.
Day 21
~15 hrs/week
20 discovery calls. ≥8 teams say they write docs twice (Confluence plus a separate help center) and name what it costs them.
Full-time
20 discovery calls. ≥8 teams say they write docs twice (Confluence plus a separate help center) and name what it costs them.
If you miss it
Stop. The pain isn't there. Don't fall back to Notion.
Day 45 (kill test)
~15 hrs/week
5 paying pilots at $150–300 each
Full-time
5 paying pilots at $150–300 each
If you miss it
Stop. One retry is allowed only if ≥3 said "yes, next quarter".
Day 90
~15 hrs/week
Forge app submitted or live. 10 paying teams. ≥80% of pilots publish a page change every week.
Full-time
Forge app submitted or live. 10 paying teams. ≥80% of pilots publish a page change every week.
If you miss it
Cut to the doc-owner loop plus a plain site. Drop AI answers.
Month 6
~15 hrs/week
$1k MRR (~15 hrs/week) or $1.5k (full-time). ≥5 reviews.
Full-time
$1k MRR (~15 hrs/week) or $1.5k (full-time). ≥5 reviews.
If you miss it
Stop, or sell the customers and code to a Confluence partner.
Month 12
~15 hrs/week
$2k / $3.5k MRR. Revenue churn ≤2% a month. 15 reviews.
Full-time
$2k / $3.5k MRR. Revenue churn ≤2% a month. 15 reviews.
If you miss it
Stop adding features. Fix onboarding and the doc-owner loop only.
Month 18
~15 hrs/week
$2.5k / $5k MRR
Full-time
$2.5k / $5k MRR
If you miss it
Put everything into the single best channel (migration content or Marketplace).
Month 24
~15 hrs/week
$3k / $5k+ MRR (stretch: $10k full-time)
Full-time
$3k / $5k+ MRR (stretch: $10k full-time)
If you miss it
Above target: run it profitably. Below $1.5k: stop.
Do
1.Sell one job: "write once in Confluence, publish a help center your customers can search, and see what they couldn't find."
Target: Every visitor gets it in one line
Why: Double-writing between Confluence and a separate help center is the only pain found that incumbents don't lead with. Scroll and Refined sell site building.
2.Target software companies with 10–250 Confluence users who pay for a separate help center (Zendesk Guide, Intercom articles, Help Scout Docs, Document360).
Target: ≥60% of paying customers replaced a separate tool
Why: Standalone help centers cost $50–695 a month (vendor comparison, May 2026). Your price is $20–110. That's a saving you can state in dollars.
3.Sell before building. Offer 5 pilots a done-for-you move of their public help center into a Confluence space, hosted on their domain. Charge $150–300 once, with 3 months included.
Target: 5 paying by day 45, 10 paying by day 90
Why: It proves teams will switch before you spend months on a Forge app.
4.Build Forge-only, with no Connect modules.
Target: Qualifies for 100% revenue share
Why: Connect apps pay 20–25%, and Connect support ends 31 Jan 2027 (verified in StatusLens research).
5.Ship the doc-owner loop as the differentiator. Every week, file failed searches, "not helpful" votes and pages with no views as a Jira issue or Confluence task, assigned to the page owner.
Target: ≥50% of customers close ≥1 of these tasks a month
Why: It turns traffic into better docs. It's also the weekly reason to keep paying.
6.Sell AI answers with citations in the Advanced edition only, with a cap per site.
Target: AI cost ≤8% of revenue
Why: HelpKit charges +$49 a month for AI. Scroll and Refined include AI search. Answers must cite a page, or they're a liability.
7.Price inside Atlassian's user tiers, with two editions. No free plan beyond the Marketplace trial. Push annual billing. No lifetime deals.
Target: Blended ≥$50 a month. ≥40% of revenue on annual by month 12.
Why: Low prices buy high churn, and Marketplace buyers expect per-user tiers.
8.Publish migration guides as your main content: "Zendesk Guide to Confluence", "Intercom articles to Confluence", "Notion help center to Confluence", plus a free help-center checker (paste a URL: search, broken links, llms.txt, speed).
Target: ≥30% of trials from content by month 9
Why: People search for the move they're about to make. It keeps working while you're offline.
9.Ask every pilot and customer for a Marketplace review.
Target: 15 reviews at ≥4.5 stars by month 12
Why: Scroll Sites has 111 reviews. Reviews and installs decide what the Marketplace shows first (the ranking method is unverified).
10.Automate operations from day 1: self-serve install, your own docs hosted on your own app, a status page, email-only support with a 1-business-day promise.
Target: Support ≤2 hrs/week per 50 customers
Why: It keeps the product workable at ~15 hours a week.
11.Make every site fast, indexable and AI-readable by default: static pages, sitemaps, clean URLs, llms.txt.
Target: Every site scores ≥90 on Lighthouse performance
Why: Help centers are found through search and, more and more, through AI assistants. HelpKit and Scroll both market this. It's table stakes.
12.Publish a data-flow page, a DPA and a security self-assessment at launch. Apply for Cloud Fortified after $3k MRR.
Target: Live at listing
Why: Atlassian buyers ask before they install. Both incumbents are Cloud Fortified.
Avoid
1.Any form builder on top of Notion
Why: Notion Forms is free on all plans. Tally offers unlimited free forms with a free Notion integration (500,000+ teams). NoteForms claims 130,000+ users. The original founder built OpnForm to get off Notion.
2.A generic Notion help-center builder
Why: HelpKit claims 1,000+ customers and publishes an "alternative" page against every rival. NotionDesk, Notaku, Super, Bullet and Potion also compete, and Notion Sites is free.
3.Competing on themes, page builders or site design
Why: That's what Scroll Sites and Refined Sites do best, with Platinum-partner teams behind them.
4.Supporting several content sources (Notion + Confluence + Google Docs)
Why: Three integrations to maintain, three sets of API changes, and no clear buyer.
5.Internal wikis or intranets
Why: Refined's home ground, and Confluence already does internal knowledge natively.
6.A full help desk (tickets, inbox, live chat)
Why: Zendesk, Intercom and Jira Service Management own it. Integrate with them instead.
7.A generic AI chatbot as the core product
Why: Intercom Fin, Chatbase, Atlassian Rovo and Notion Custom Agents all answer questions from docs. Citations inside your help center are a feature, not a company.
8.Promising "deflect X% of tickets"
Why: You can't measure deflection reliably without the help-desk data, and a missed promise becomes a refund. Report searches and clicks, not claims.
9.Connect modules or any non-Forge architecture
Why: Loses the 100% revenue share and hits the Connect end of support on 31 Jan 2027.
10.Data Center or Server versions
Why: A second codebase for a shrinking base. Cloud only.
11.Lifetime deals (AppSumo) or a free plan
Why: They cap revenue and fill support with teams that never pay.
12.Paid ads before month 6
Why: Until churn is proven, ads buy installs that leave.
13.Starting this while your first portfolio idea is still before its month-6 checkpoint
Why: At ~15 hrs/week, two new products means two half-built ones.
14.Adding features when a checkpoint fails
Why: A miss means the wrong buyer, or that teams don't want to leave their help desk's built-in help center. Change the buyer once, or stop.