How Coira works — and what it can't tell you.

Every live figure comes from publicly published contract award or tender notices. TED (Tenders Electronic Daily), the EU's official procurement journal, is the only source on a live schedule — every award and tender in Coira's search, benchmarks and buyer pages came from it. Separately, Coira holds a fixed archive of Irish OGP/eTenders open data downloaded from data.gov.ie, which powers the contractor directory and nothing else. That archive is not refreshed on a schedule, and the directory shows the date of the most recent award it contains.

Each project in Coira links back to its original source notice, so you can check it yourself. Coverage spans 30 European markets, but not evenly: Ireland is by far the deepest, and the only market we refresh every night. National portals beyond TED are listed in the Source Atlas before they are treated as live coverage.

We only count a value as reliable for benchmarking when it sits in a sensible range for construction. We leave out:

• placeholder values, for example a disclosed price of €1
• suspiciously low figures, under €10,000
• mega-frameworks over €500m — multi-year, multi-supplier agreements that distort the per-project median

Excluded projects still show up in search. They are just left out of the statistics.

Every benchmark shows a confidence tier, based on how many comparable awards it rests on:

• Strong — 100 or more comparable awards
• Moderate — 30 to 99
• Early coverage — under 30

A median built from 400 awards is far more dependable than one built from 15. We show you the sample size every time, so you can judge for yourself.

Below five, we show nothing. Not a greyed-out figure or a low-confidence badge — no number at all, and a line telling you how many awards were found and how many are needed. This applies everywhere a figure is computed: benchmark reports, the Budget Builder and contractor bid behaviour. A number you could quote to a client should never rest on one or two jobs, and a caveat beside a euro figure is not a refusal — people take the figure and leave the caveat.

A median is not always a benchmark. Public construction awards are heavily right-skewed: framework ceilings and whole-campus contracts sit in the same sector as classroom extensions. Where the 75th percentile is more than three times the 25th, we show the median and the spread but stop calling it a benchmark, because at that width the middle figure tells you very little about your own job. The spread is shown beside every median so you can see this for yourself rather than take our word for it.

Not every notice discloses a contractor name, floor area, or full scope. Floor areas in particular are rarely published, so €/m² figures are only available for a subset of projects. Where a field is missing, we leave it blank rather than guessing.

Ireland refreshes every morning at 6am. We pull new construction and engineering award notices (CPV 45/71 — the EU's classification codes for building work) and live open tenders from the official TED API, then classify and de-duplicate them. Live tenders past their deadline are hidden automatically.

The other 29 markets refresh on a rotating sweep — 3 markets each night, so every market is revisited within 10 days. That means a brand-new award outside Ireland can be up to 10 days behind. Keep this in mind before you rely on a same-week figure from outside Ireland.

Two limitations we have not solved yet:

• Sub-threshold notices published only on national portals, and never forwarded to TED, are not included. In Ireland this is the specific gap worth naming: eTenders.gov.ie carries RFTs below the EU publication threshold that never reach TED, so Irish awards and live tenders in Coira are the above-threshold subset. The exception is the contractor directory, which is built from the separate OGP eTenders sub-threshold dataset.
• How far back we can search each market is capped by how far the TED search itself will page.

These tools are deterministic — no AI, no hidden model. They use a curated table of published Irish build-cost rates: the Mitchell McDermott Construction InfoCard (January 2025) and the Buildcost Construction Cost Guide (H2 2025), published by a practice affiliated with the SCSI and RICS — the Society of Chartered Surveyors Ireland and the Royal Institution of Chartered Surveyors. Every result shows its source.

These rates are build cost per m² GFA (gross floor area), excluding VAT (13.5%), professional fees, siteworks/abnormals and FF&E (fixtures, fittings and equipment). They are curated benchmarks, not figures derived from the award data. We always show the two separately, so you know which number you're looking at.

Of the awards Coira holds, roughly 37% publish an estimated value, and about 21% give us a usable award-versus-estimate figure. That gap is a limit of what buyers disclose — not a sign that we are not crawling deep enough. We sampled live TED pages beyond our normal refresh depth and found that only about a quarter of award notices publish an estimate at all. We already hold nearly every one of those, because the corpus was built by an ingest process that specifically requested that field.

We also discard cases that look like a real estimate-versus-award signal but are not:

• Awards that match the estimate to the cent — this usually means a buyer republishing its own figure, not a genuine result.
• Ratios beyond −80% or +200%, where a lot has clearly been compared against a whole tender rather than the matching item.
• Any award where the price and the estimate are quoted in different currencies.

So this figure will grow as new notices are published, but it will not jump suddenly. We would rather show you the honest number than a bigger one built from cases we cannot stand behind.

The headline number on a counterparty report is the median gap between the buyer's pre-tender estimate and the price the contractor was actually awarded, across every award of theirs we hold that publishes both figures. We compare it against directly comparable firms — same market and same trade, where the sample supports it — and the report always names which comparison it used and how many awards it rests on.

This figure tells you how tightly that firm prices, and so how much room is likely left in the job when they come to buy your package.

It is not a credit check, a solvency opinion, or a prediction of whether you will be paid. Coira holds no company accounts, no filings, no judgments and no payment history — there is no such source anywhere in the product. A low bid is not evidence of financial distress.

This section exists because the report once carried the heading "Can they pay you?", which wrongly implied otherwise. Treat the figure as a reason to check scope and contract terms carefully — and use a proper credit report for credit decisions.

Buyer and contractor signals are computed once a night from the same award and tender notices as everything else on Coira — nothing here is estimated, modelled, or generated by an AI at the moment you view the page. Every figure carries its own sample size, and — as everywhere else on Coira — we do not show a number at all when that sample is below five.

The contractor gap-vs-baseline figure takes the existing bid-behaviour gap (a contractor's award price against the buyer's own pre-tender estimate, as described above) and compares it against that same buyer's own median gap across all their awards, not a market-wide average. A contractor who prices 5% under estimate looks tight against a buyer who typically sees 15% swings, and unremarkable against one who typically sees 2% — the normalised figure tells them apart; the raw gap alone cannot. We show the raw figure alongside it, labelled, for anyone who wants the unadjusted number.

Buyer signals — award history, the estimate gap, how often a deadline gets extended, how often a procedure ends with no winner chosen, and how many tenders a buyer typically receives — are all counted directly from the notices we hold for that authority, with the same n-based confidence rule as everywhere else. Deadline-extension and no-winner tracking read from our own capture history of each notice over time, which only began recently — these two figures will read thin for most buyers at first and fill in as more notices are captured across their full lifecycle. That is expected, not a fault in the count.

Comparable-award medians shown outside Bid Check use the same classification and the same framework/unreliable-value exclusions as Bid Check's own comparison, but are a coarser, nightly-refreshed reference bucketed by country, work basis and building type — not narrowed to a specific job's value the way Bid Check's live comparison is. Use Bid Check itself when pricing an actual bid; use this figure to get a general sense of a market before you commit to checking a specific one.

Nine of the 30 markets we cover do not use the euro — Sweden, Denmark, Poland, Czechia, Hungary, Romania, Norway, Iceland and Switzerland all publish awards in their own currency. TED reports each award exactly as the buyer published it, so a Hungarian award of 400,000,000 HUF is roughly €1.0m, not €400m.

Coira stores the award exactly as published, and separately stores a converted euro figure. Every cross-market comparison — medians, percentiles, spend rankings, value filters — runs on the converted figure. We show the original figure beside it, so you still recognise your own number.

The conversion uses a single fixed rate per currency, dated on the page — not the rate on the award date. An award from 2019 converts at the same rate as one from 2026. That makes a converted figure a useful order-of-magnitude comparison, not a settlement amount.

We chose this approach deliberately. Ignoring rate drift misleads you by 10–20%. Mislabelling a currency misleads you by 5 to 400 times. Where we cannot identify a currency, we convert nothing and leave the row out of euro statistics, rather than assume it is euro.

We assign sector and trade from each notice's CPV code (the EU procurement classification code, explained above), refined with a title signal pass that catches M&E work: mechanical, electrical, HVAC, fire, lifts and BMS (Building Management Systems) that generic CPV codes miss. Classification is deterministic, and we apply it again on every refresh.

One detail matters: a notice usually carries several CPV codes, and the construction one is often not listed first. A flood defence contract can lead with a "floodgates" product code. An electrical installation can lead with "electric accumulators". So we select the works code (CPV 45) or engineering services code (CPV 71) from the full list, rather than taking whichever code appears first, and we keep every code on the record.

We measured this against 400 live notices: without this step, 11% would be filed under an unrelated product or service code, and effectively unfindable by the trade they actually belong to.

Sector and trade answer what kind of work, on what kind of building. They do not say how much of that building's life a given award represents. A school building sector on its own blends a new national school with a small window and door replacement job.

To fix this, Report, Estimator and Budget Builder each classify a works basis award as new build, extension, refurbishment, fit out or maintenance. This uses the CPV code plus a title signal pass: a "construction of a new" phrase takes priority, even when demolition of an existing structure is mentioned in the same title. Wherever there is enough data to do so safely, we narrow to this classification before computing a median.

The gap this closes is real and measured, not theoretical. For Irish schools, the unfiltered work-type median was €1.72m. Narrowed to new build only, it is €6.65m.

Two honest limits:

• CPV and title text carry this signal reliably for English-language notices. A title in another market's language, without a matching CPV code, falls back to a general "building" classification rather than a guess.
• A genuinely ambiguous title (bare "Construction work", nothing else) defaults to new build. Most substantial works-basis awards in this dataset are new build, so this is safer than leaving the award unclassified and dropped from every benchmark.

We record award values as published — generally excluding VAT, though this is not guaranteed across every notice. Headline figures are as-awarded, and not adjusted for inflation.

Bid Check and Tender Intelligence separately index historical Irish awards to present-day prices, using a stored year-by-year factor table. This currently implies roughly 2% a year over 2023–2025, below the roughly 3–4% a year that the SCSI tender price index has published for the same period. We are reviewing our factor table against the SCSI series. Until it is updated, treat any inflation-adjusted figure as directionally right, but possibly a little conservative.

Ireland is a small market. Some sectors — healthcare, specialist civic buildings — simply have fewer public awards. Where coverage is thin, we tell you. A thin benchmark is a starting point for judgement, not a precise answer.

A contract award value is the tendered sum a contractor agreed to. It is not a full development cost plan: it typically excludes client-side costs, design fees, site acquisition, abnormal works, and risk allowances carried elsewhere. Treat Coira benchmarks as the construction-works component, not the all-in project cost.

In a bid room you can record the submission deadline, each clarification deadline, and internal team tasks with their own due dates. Coira reminds you as each one approaches: in the app for everyone on the room, and by email if you opt in on your account settings. Reminders escalate as the date nears: a week out, three days out, the day before, the morning of, and again if the date passes while the deadline is still open. You choose which of those steps you want.

One rule matters more than the schedule: the authoritative deadline is the one on the eTenders notice, not the one in the tender document. An ITT frequently says the return time is “as per eTenders”: the document defers to the portal. So Coira records where each date came from, and a date read out of a document rather than confirmed on the portal is flagged for you to check. We remind you of the date you entered; the published notice remains the record.

Reminders are per person and per bid room. You only ever see dates for rooms you own or have been added to: a reminder never carries another company's tender.