Skip to content

Investigation guide

How to use PeBaniiNostri.ro for investigations and data journalism.

What is PeBaniiNostri.ro?

PeBaniiNostri.ro automatically checks every public purchase in Romania (tender, direct purchase or offline purchase) against 10 risk indicators, using the open data from data.gov.ro.

Three rules before anything else

  1. A flagged case is a reason to look, not proof. The indicators find statistical anomalies. The explanation may be perfectly legal, so every case has to be checked in context.
  2. Read the unit. Each indicator counts something different: a group of purchases, a purchase, a company, an authority–supplier pair, a lot, an authority, an authority-year or a market. Never add up the cases or the money of two indicators.
  3. Every figure can be checked. The last column of an indicator's table opens the purchases the case was computed from, and each purchase's SICAP number opens its record on e-licitatie.ro.

1Where to start

An investigation usually starts in one of three places: a name you already have, an indicator that found something, or an area you want to understand.

From a name

Type a name, a CUI, a purchase's SICAP number or a CPV code into the search. A CUI opens the authority's or the company's profile directly, a SICAP number opens the purchase, and a CPV code opens its group in the CPV statistics.

From an indicator

The indicator dashboard shows the authorities and suppliers flagged by the most indicators. On each indicator's page the table opens on the cases with the most money and filters by party, year and, where it applies, CPV code.

From a county or a market

The statistics filter by county, year and CPV domain. The authority list filters by county, by size of spending and by indicator, and the Authorities and suppliers page shows the largest relationships between an authority and a supplier.

2How to investigate a contracting authority

  1. Open the authority's profile, from the search or from the authority list. The first figures show how much it spent in each stream: direct purchases, offline purchases and tenders.
  2. Read the three scores (single-bidder rate, discretionary spend index and supplier concentration) beside the national reference and the authority's rank. A score only matters if it is high against the other authorities.
  3. Go through the strip of nine indicators. A lit indicator opens its cases, already filtered to the authority. Several indicators lit in the same year or at the same supplier say more than one alone.
  4. Look at the charts by year and by CPV code. A sudden jump in one year, or large spending in a field unrelated to what the authority does, are good questions for later.
  5. In the supplier table, read each supplier's share and the stream it was paid through. A supplier taking a large share of the money through direct purchases alone is worth comparing with indicators #01 and #02. Every amount in the table opens the purchases it was summed from.
  6. Finish with the latest purchases and the full lists under Acquisitions, filtered to the authority: that is where each purchase's description, value and SICAP number are.

3How to investigate a supplier

  1. Open the company's profile, from the search or from the supplier list. The header shows the registration date and, if indicator #03 flags the company, how many days passed before its first public money.
  2. Read the three scores (money won without competition, money won at exactly the estimated value, and dependence on the main client), each beside the national reference. Below a minimum number of lots or clients the score is missing, because it would say nothing.
  3. Look at the clients: each one's share of the company's revenue (how much the company depends on it) and of the authority's spending (how much the authority depends on the company). A pair where both shares are high is a relationship to follow over time.
  4. Check the association partners, the companies that won the same contracts together. The same partners returning to the same authorities may point to a shared-out market.
  5. Put the revenue-by-year chart beside the company's history in the trade registry: shareholders, directors, seat, headcount, balance sheets. A jump in revenue right after a change of shareholders is a question to ask.
  6. For a new company, read its first year's balance sheet. A company with no staff and no assets carrying out millions of lei of works in its first months has probably subcontracted the work, and that shows in the award documentation.

4Reading the indicators in context

Every indicator page is built the same way. Before the cases, read four things:

  • The unit of a case: what the indicator counts (a group of purchases, a lot, a pair, a market, an authority-year).
  • The threshold: the value from which a case is flagged. The thresholds are set from the national distribution and described in the methodology panel, under the headline figures.
  • The rate by year: the chart shows the cases and their share of everything that was checked. A year with many cases may just be a year with many purchases; the rate says whether the pattern grew.
  • What it does not cover: every methodology ends with the method's limits. Read them before writing about a case.
Reading the indicators in context
Indicator Key question Verification sources
#01 Slicing Is there a genuine need behind each tranche, or is this one disguised acquisition? Technical specifications, the procedure required at the real cumulative value
#02 At the ceiling Why did the value stop just below the ceiling? How was the need estimated? The needs assessment, the market study, the supplier's quote
#03 New companies Does the company have employees, comparable experience and real premises? Trade registry, balance sheet, share capital, headcount
#04 Captivity Are there personal or business links between the authority's management and the supplier? Asset and interest declarations, trade registry
#05 Single bidder Did the technical specifications deliberately exclude other operators? Technical specifications, appeals lodged with the review body
#06 Single-bidder rate Why is this authority's rate so far above the national average? Court of Accounts audit reports, annual procurement plans
#07 Exact price match Did the bidder know the estimated value before the notice? Market studies, prior communications with the authority
#08 Year-end Are the December purchases justified, or do they simply burn the remaining budget? Budget execution, needs assessments, the urgency invoked
#09 DSI index Why does the authority avoid competitive procedures at this budget size? The annual procurement strategy, the staffing of the procurement unit
#10 Concentration Why does this market's money go to one supplier when the country has hundreds? Market studies, tender specifications, the bidders in the CPV group's tenders

The indicators are best read together. A new company (#03) winning a single-offer lot (#05), at exactly the estimated value (#07), from an authority it depends on (#04), makes a pattern no indicator sees alone. The dashboard shows who is flagged by several indicators at once.

5From flagged case to finding

A flagged case is only the start. The steps below separate an explainable anomaly from a story.

  1. Redo the arithmetic. Open the purchases behind the case (the table's last column) and check their number, values and dates. If something does not match, settle the figures first.
  2. Read the source. The SICAP number opens each purchase on e-licitatie.ro: the full description, the notices and, for tenders, the award documentation, the clarifications and the outcome of the procedure.
  3. Compare. Put the case beside the national reference, beside the same authority in other years and beside other authorities buying the same CPV code (CPV statistics). A 60% single-bidder rate in a market where the average is 55% is not a story.
  4. Rule out the innocent explanations. Check the usual explanations listed below first. Every one you can rule out makes the case stronger.
  5. Ask for the documents. The procurement file (dosarul achiziției publice) is a public document. You can request it from the authority under Law 544/2001, together with the needs assessment and the document that justifies the estimated value.
  6. Connect the people. The trade registry shows the company's shareholders and directors, and the asset and interest declarations of the authority's management are public. A link between them turns a pattern into a question of public interest.
  7. Give a right of reply. Send your questions to the authority and the company before publishing, and quote the answer. Write "flagged by indicator #05", not "fraud": the figures here describe purchases, not intentions.

Common innocent explanations

  • #01 and #02. Consumables bought every month (fuel, food, medicines) naturally make many small purchases from the same supplier, and a price close to the ceiling may simply be the market price. What matters is whether the need was known from the start and could have been covered by one procedure.
  • #03. A new company may carry on an older one's business (a reorganisation, a transfer of activity) or be a company set up by the local authority itself. The trade registry shows where it comes from.
  • #04 and #10. Water, waste collection and district heating are often locally licensed services with only one possible operator. A small supplier serving a single commune naturally depends on it.
  • #05 and #06. Products under exclusive rights, spare parts for equipment already bought and niche markets have few bidders everywhere. Compare with the national rate of the same CPV group.
  • #07. An estimated value set from a single quote, perhaps the eventual winner's, produces exactly this match. It is not an innocent explanation in itself, but it shows where to look: the document that justifies the estimated value.
  • #08 and #09. Autumn budget revisions, purchases planned for the end of the year (subscriptions, licences, firewood) and the direct-purchase ceilings doubled in September 2022 move money towards December and towards direct purchases with no intent behind it.

What the data does not hold

  • The data starts in April 2018, so the first year is incomplete; late December 2023 is missing from the data.
  • The money is money spent: a framework agreement is not counted as such, but through the subsequent contracts signed under it.
  • The values are those the authorities reported. Impossible values (above the legal ceiling for direct purchases, or zero) are removed, but the reporting errors that remain reach the calculations too. That is why the first check is to redo the figure.
  • The data describes the award, not the performance: it does not show whether the work was done, whether it was paid in full or whether the contract was later amended.

Read the methodology

6With an AI assistant

PeBaniiNostri.ro has an MCP server. You connect it to Claude, ChatGPT, Copilot, Cursor or Gemini, and the assistant reads the same data and the same indicators as the site, through the same operations. Connecting takes a minute and is described on the AI assistant page.

What it helps with

  • Triage. Ask about a county, a field or a kind of authority and get, within minutes, the list of those worth a look, with each one's indicators.
  • Chaining. The assistant finds the CUI, the CPV code or the county code itself and moves from a profile to an indicator's cases, then to the purchases behind them, without you knowing what each filter is called.
  • Searching in your own words. "Laptops bought by schools in Iași in 2025" becomes a search by CPV code, county and year, and the results are compared on the spot: prices, suppliers, periods.
  • Summing up. At the end, ask for a table of what you found: each finding, its figure, the indicator and the link to the page where you can check it.

An investigation, step by step

An example conversation. Each question builds on the previous answer; under it, the tools the assistant usually calls.

  1. „Which authorities in Vaslui county spent more than 10 million lei and are flagged by at least three indicators?”

    Tools used: get_reference_valueslist_authorities

  2. „Take the first one. Read its profile and tell me which scores stand out against the national reference.”

    Tools used: get_authority_profile

  3. „Show me its lots awarded on a single offer, over the whole period, largest first, with the winners.”

    Tools used: find_single_bidder_lots

  4. „Does the winner of the largest lot also appear among its direct-purchase suppliers? What share of the company's revenue comes from this authority?”

    Tools used: get_supplier_profilesearch_purchases

  5. „Make me a table of everything we found: the finding, the figure, the indicator and the link to the page where I can check it.”

How to use it without being misled

  • The figures are ours, the conclusions are its own. The assistant can misread a table or draw a conclusion the data does not support. Open the link of every figure you use.
  • Ask it to look up, not to remember. A CUI, an amount or a company name given without a tool call is a guess. Ask it to check it in the data.
  • Name the period. "In 2024" and "over the whole period" are different questions. For #05, #09 and #10 the whole period is a separate test, not the sum of the years.
  • Do not let it add indicators up. A lot, a pair and a market do not add up. A "total flagged money" across several indicators means nothing.
  • It only knows what is in the data. It does not read specifications, contracts or the trade registry. The documents and the people remain your part of the investigation.
  • It has a daily budget. Every tool call is charged to your account's budget, the same as the API's. A whole profile costs a single call.

Connecting it, every tool and more examples

7The data and citing the platform

Everything on the site can be retrieved as JSON through the public API. You get the key from your account after signing in with Google, and at the foot of every page a box lists the API addresses it was built from.

The site's data and analyses are published under CC BY 4.0, which asks for one thing: credit PeBaniiNostri.ro. We do not ask you to credit data.gov.ro as well. Only if you redistribute the source records in bulk, rather than our figures and analyses, keep the statement their licence requires: "Contains public information under the Open Government Licence v1.0".

The data is periodically reloaded in full. Note the date you consulted a figure: after an update, the cases of a year still in progress can change.

Suggested citation: Source: PeBaniiNostri.ro (pebaniinostri.ro), CC BY 4.0, [month/year].