Evidence first · snapshot 7 November 2026
Should we bid — and at what price?
312890 · Flash-rain hazard warning system
Äkksadudest ohustatud alade hoiatussüsteemi välja töötamine · KeMIT
Conditional pursue with a qualified team. No-go as an unqualified solo/software-only bid.
This is a methodology + hydrology + GIS + forecast integration project, not a generic website. Supplier CVs, availability and costs are unknown. No win probability or profitable target price is established.
Due 11 November 2026 · 12:00 Europe/Tallinn / 10:00 UTC. Recheck the live register; this report does not refresh automatically.
€250,000Published estimate · verified in XML + UI
60 / 40Quality / price points
8 rolesMax 2 roles/person · at least 4 people
9 monthsDelivery + 12-month software warranty
The findings that matter
- Do not price as a fixed discount off €250,000. The estimate is not an established hard ceiling. Historical examples cover different scopes and include prices above estimates.
- Prior engagement is visible. The market study names CGI, Nortal, Regio and TalTech; Cloud IS is named in a clarification topic. These are not confirmed bidders or a unique-competitor count.
- Three Q&A threads ≠ three competitors. There are separately three buyer announcements. Publicity can occur after answering; private questions are not counted.
- A close 2025 flash-rain/GIS procedure failed to produce a contract. Estimate €54,600; result notice says top-ranked bidder(s) refused to sign. Bid counts conflict (2 in contract statistics, 0 in result notice). No individual bidder motive or price cause is established.
- Quality can justify a higher price. The actual scoring formula—not historical average discount—belongs at the centre of the pricing decision.
Primary evidence is RHR. Multiple register views are consistency checks, not independent evidence of complete or error-free reporting.
Two-offer scoring sandbox
Published rule: price = 40 × lowest eligible price / evaluated price, rounded to 2 decimals. Quality is the sum of the published bands. Hypothetical inputs, not predictions.
Assumes these are the only eligible offers; a third lower offer changes price scores. Equal totals: higher quality, then lower price. Committee scoring is unknown; selecting 60 quality points does not make them likely.
Published rubric v2 ↗Cost and margin floor
Enter our actual costs. Blank means unknown, not zero. No supplier-specific assumptions are prefilled.
Floor = all-in cost / (1 − margin). Conditional on winning; no failed-bid amortisation, win probability or portfolio expected profit. Include cash-flow needs, acceptance rework, Estonian documentation and mandatory technical evidence.
What is in open data?
Competition XML ↗ · Result XML ↗
Presence counts are per notice version, not per procurement. Result notices can be no-award or modifications. Their PayableAmount fields require scope/unit/currency checks and must never be filled with the original estimate.
Submission-statistics trap: the downloaded result XML contains numeric statistics without type-code children. We did not interpret “first number” as total bids or add overlapping categories. Selected examples use typed contract statistics + rendered award notices.
Competition signals: observed vs unproven
Exploratory post-award diagnostic
Question timing / simultaneous-procurement workload predictive value is not established. No historical first-public snapshots or chronological holdout; no question-based win score is produced.
Public questions — timing is not competitor identity
Sent and answer-sent are source fields. First-public-visible time was not recorded and is unknown. One thread may contain multiple questions; several threads may be the same supplier. No numeric prediction of target bidders is justified.
Other opportunities with nearby deadlines
Four illustrative examples discovered in October XML and refreshed from live headers on 7 November; not a complete market inventory or proof of supplier overload. DigiVesi is now due 18 November (October XML said 13 November); PULSE is due 16 November (earlier entries said 4/9 November).
Compliance and commercial decision gates
Before spending heavily on the bid
- Named CVs for all eight roles: modelling, GIS, PM, analyst, developer, architect, UX/UI, tester. Max two roles/person; every role requires 12 months relevant work in the preceding 36 months plus its education/experience route.
- Current v2 analyst requirement allows a 3-year experience alternative to relevant higher education; do not rely on void originals.
- Estonian offer, work and documentation; RHR price form; ESPD/sanctions, CV signatures and authorisations. No conditional offer; meet both 90-day and 3-month validity wording.
- Cost methodology validation and data quality/licensing; national layers, GIS/GeoServer, forecast ingestion and warning interfaces; accessibility and technical handover evidence.
Before committing a price
- Acceptance evidence includes SBOM, security/load tests, recovery exercise, API/event/CAP documentation and traceability. Resolve critical/high security findings before acceptance.
- Payment after accepted stages, 21-day invoice term; no identified advance or fixed stage-payment percentages. Fund delivery and potential rework.
- 12-month software warranty; critical 8-working-hour repair, medium 16-working-hour repair. Delay penalty 0.5% of affected stage/work per day subject to caps.
- Broad non-exclusive IP licence + source handover; include relevant rights in price. Related extra work up to 10% is not guaranteed revenue. Future-at-signature MFN baseline remains a cost uncertainty.
Still needed from us: people/CVs, partner rates, capacity, effort/cost model, reusable-IP rights, contingency/warranty reserve, cash runway, target margin and bid-cost ceiling.
Scope, access and remaining unknowns
This is a local, self-contained prototype, not a live service or bid submission. It sends no background requests and stores no inputs. Public sources cannot establish our qualification, full losing-offer price lists, individual rejection reasons or final spend where it is not reported. A nonwinner is not automatically disqualified; a no-award reason is not an individual legal finding.
RTK terms distinguish open-data access from special-purpose automation of UI-only functions. Use documented monthly downloads/search exports for scale; obtain permission before scaling UI-only collection. No collector was scheduled. A current explicit feed Creative Commons licence was not verified; no claim of unrestricted republication is made. Schema changed to eForms in May 2023; older notices need legacy parsing.
Snapshot and dataset scope