Public roadmap¶
FitRadarHR's V1 is complete (epics E1–E8 delivered) and V2 is fully delivered (9/9 items). This page also presents candidate ideas for a V3, prioritized using the RICE method — in full transparency, like the rest of the project.
How to read the RICE score¶
RICE = (Reach × Impact × Confidence) / Effort
- Reach — how many users are affected per quarter (scale 1–10)
- Impact — effect on the North Star (reports viewed / week): 0.5 = low, 1 = medium, 2 = high, 3 = massive
- Confidence — confidence in the estimates: 50% / 80% / 100%
- Effort — workload in person-weeks
V2 prioritization¶
| # | Feature | Reach | Impact | Conf. | Effort | RICE score | Status |
|---|---|---|---|---|---|---|---|
| 1 | Automatic questionnaire reminders | 8 | 2 | 80% | 1 | 12.8 | ✅ Delivered |
| 2 | Inviting managers into the org | 7 | 2 | 80% | 2 | 5.6 | ✅ Delivered |
| 3 | Comparing candidates for a position | 6 | 2 | 80% | 2 | 4.8 | ✅ Delivered |
| 4 | CSV import of people | 6 | 1 | 100% | 1.5 | 4.0 | ✅ Delivered |
| 5 | Re-take & longitudinal tracking | 5 | 2 | 50% | 2 | 2.5 | ✅ Delivered |
| 6 | Retention cohorts in Analytics | 4 | 1 | 80% | 1.5 | 2.1 | ✅ Delivered |
| 7 | SSO Keycloak / OIDC | 3 | 1 | 80% | 3 | 0.8 | ✅ Delivered |
| 8 | ES / DE translations | 2 | 1 | 80% | 2 | 0.8 | ✅ Delivered |
| 9 | Public API (read-only) | 2 | 1 | 50% | 3 | 0.3 | ✅ Delivered |
The arguments, item by item¶
1. Automatic reminders — the Analytics funnel shows most of the drop-off happens between sending the link and starting it. A reminder at D+3 directly addresses the completion rate, for minimal effort. Best value/effort ratio on the list.
2. Inviting managers — today an org lives with a single HR account most of the time. Each invited manager is one more consumer of reports: this is the "referral" lever of the AARRR funnel.
3. Comparing candidates — the fit ranking already exists on the position page; the side-by-side view (overlaid radars, complementarities) is the natural next step demanded by the recruitment use case. High impact on report consultation.
4. CSV import — onboarding friction: entering 50 employees by hand is discouraging. Indirect impact on the North Star but maximum confidence (obvious need).
5. Re-take — Big Five profiles evolve slowly but they do evolve; useful for internal mobility. 50% confidence: the real need still needs to be validated in discovery.
6. Retention cohorts — deepens the Analytics page for mature orgs. Reserved for already-activated orgs, hence lower reach.
7. SSO / OIDC — necessary for larger organizations; lower reach on the current target (SMBs/scale-ups) but one IdP per organization, additive to the password (never a replacement), keeps maintenance cost acceptable. Delivered, not yet validated with a real IdP in production.
8. ES/DE translations — full parity, including the IPIP questionnaire,
not just the interface. The German questionnaire is fully sourced from the
official IPIP translation (100/100 items); Spanish is for the short version
(50 items) but the long version introduces an in-house translated area,
for lack of a published official source — see docs/user/about/big-five.md.
9. Public API (read-only) — real value but not differentiating today. Delivered with a deliberately restricted scope: positions/teams, people + questionnaire status, fit results — never raw Big Five profiles (GDPR minimization). API key authentication per organization.
V3 prioritization¶
Items #1, #2, #3, #4 and #5 are delivered; items 6, 7, 8, 9 and 10 remain undecided candidates.
| # | Feature | Reach | Impact | Conf. | Effort | RICE score | Status |
|---|---|---|---|---|---|---|---|
| 1 | Reverse fit — best positions for a person | 5 | 2 | 90% | 1.5 | 6.0 | ✅ Delivered |
| 2 | Monetization (free plan → subscription) | 8 | 3 | 80% | 5 | 3.84 | ✅ Delivered |
| 3 | Mapping a team's gaps | 5 | 2 | 60% | 2 | 3.0 | ✅ Delivered |
| 4 | Candidate/employee portal (access to their profile) | 6 | 2 | 70% | 3 | 2.8 | ✅ Delivered |
| 5 | Exportable AI Act compliance dossier | 3 | 2 | 70% | 2 | 2.1 | ✅ Delivered |
| 6 | Webhooks (complement to the public API) | 3 | 1 | 70% | 2 | 1.05 | Candidate |
| 7 | Claude MCP — check your activity outside the site | 2 | 1 | 60% | 2 | 0.6 | Candidate |
| 8 | Anonymized cross-organization benchmarks | 4 | 1 | 40% | 4 | 0.4 | Candidate |
| 9 | Native connectors (Workday, BambooHR, Personio…) | 3 | 2 | 30% | 5 | 0.36 | Candidate |
| 10 | Public sharing of the OCEAN profile (CV, social) | 3 | 0.5 | 40% | 4 | 0.15 | Candidate |
The arguments, item by item¶
1. Reverse fit — today you start from a position to rank people; the reverse (starting from a person and seeing which open positions fit them best) serves internal mobility, without collecting any additional data: it's a new view on the existing fit engine. Best value/effort ratio on the list. Delivered: no new calculation, reuses fit results already produced by the engine (E5).
2. Monetization — this is the very reason for the move to the Fair Source license (FSL-1.1-MIT): Damien explicitly considered a free trial then a subscription. High reach (potentially touches all orgs eventually) and strong impact (it's the business model), but significant effort: Stripe integration, per-plan quotas, billing screens, overage handling. Delivered: a single paid plan (€39/month, Stripe Checkout + Customer Portal) and a permanent free plan (single threshold of 25 people — unlimited positions and questionnaires). The initial 14-day trial model was abandoned in favor of freemium: simpler to understand, without artificial pressure to convert.
3. Mapping a team's gaps — extends the complementarity already
calculated: from a team's aggregate profile, surfacing under-represented
OCEAN dimensions. Stays within the product's scope — presented as avenues
to explore for a recruitment plan, never as a firm recommendation
(consistent with the "never an automated decision score" rule). Delivered:
accessible directly from the team page, without selecting a person — see
docs/user/reports/team-gaps.md.
4. Candidate/employee portal — today only HR/managers view reports. Giving the person themselves access to their profile reinforces the "human in the loop" principle and the GDPR access right already implemented (E8). Delivered with a deliberately restricted scope: the person only sees their OCEAN profile (radar, scores, evolution), never Position/Team Fit results — showing those directly, outside any human mediation, would look like an automated recruitment verdict, contrary to the "never an automated decision score" rule. Access via a permanent, password-free link, on the same principle as the questionnaire link, sent automatically by email upon first completion and revocable (automatically via the GDPR right to erasure, or manually by an HR user who resends it). Authentication by link fingerprint only, never storing the plain value — same mechanism as API keys. Extended on August 9, 2026 with self-service GDPR erasure (the person can request erasure of their own data from their portal, with explicit confirmation and an automatic email to HR) and a public example on the demo. See candidate/employee portal.
5. Exportable AI Act compliance dossier — an export of the "traceability file" — measures already in place (human oversight, traceability, minimization), key figures and audit log — for organizations that need to document their compliance under the "high-risk" classification. Directly extends E8, mainly targets more regulated organizations (lower reach). Delivered ahead of the December 2027 deadline — see EU AI Act and recruitment for the timeline. Separate PDF (summary + measures + figures, readable) and CSV (full detail of the audit log for the chosen period, potentially large) rather than a single file — see compliance file. Explicitly framed as supporting the client organization's own compliance analysis, never as proof of compliance in its place — the "high-risk" classification applies to how the tool is used, not to the tool alone.
6. Webhooks — natural complement to the public API (V2 item #9): notifying a third-party system when a questionnaire is completed or a fit is recalculated, rather than polling. Useful only for orgs already using the API, hence restricted reach.
7. Claude MCP — expose an MCP server as a complement to the public API (V2 item #9), to check your FitRadarHR activity directly from Claude (questionnaire status, fit results, reminders to send) without opening the site. Technically a thin wrapper around the existing v1 API: same API key authentication scoped to an organization, same minimization rules (never raw Big Five scores, only derived results). The point of vigilance isn't data access but how the model uses it: deliberately narrow and structured tools (list pending questionnaires, get a position's or team's fit summary) rather than a single open-ended tool like "ask any question about this organization," so as not to let the agent drift toward a firm recommendation — contrary to the "never an automated decision score" rule. Modest reach (early-adopter public/orgs already comfortable with AI tools), limited effort by reusing the existing API.
8. Anonymized benchmarks — comparing an organization's OCEAN distribution to an aggregated cross-organization average, opt-in and anonymized. Real analytical value but low confidence: raises data governance questions (cross-tenant aggregation, consent) that still need to be scoped before any development.
9. Native connectors — beyond the generic API, pre-built integrations with specific HRIS/ATS systems would reduce adoption friction. Lowest confidence of the lot: demand not yet validated by a real customer use case, high effort (one connector = a project in its own right). To be reserved if an organization expresses a concrete need.
10. Public sharing of the OCEAN profile — an idea proposed by Damien: let a person share their profile beyond the candidate/employee portal (item #4), for example on a CV or a professional network. Lowest RICE score on the list, not just for lack of perceived value but because the feature, as proposed, directly conflicts with two product principles. First, security: the portal link is designed as a personal secret ("don't share it", see the access email) — its security relies on it never being broadcast, which using it on a CV would completely reverse, with a real risk of uncontrolled circulation (job boards, third-party CV parsers) and no way for the person to "unshare" it without erasing everything. Second, GDPR: the consent given at completion covers evaluation by the organization that sent the questionnaire, not public disclosure to any third party who happens across the link — a reuse that raises a real purpose-limitation question (GDPR Article 5(1)(b)). A version that would hold up would need an entirely separate sharing mechanism (a distinct public token, revocable independently of the portal, dedicated explicit consent) rather than a simple extension of the existing link — hence the high effort despite a limited reach.
What will never be built¶
See the anti-metrics and the project's non-negotiable rules: no CV analysis, no automated decision score, no scientifically unvalidated typology (MBTI…).
This roadmap is indicative and re-evaluated based on user feedback — the feedback widget built into the app directly feeds this prioritization.