Risk indicator methodology

Common rules

Every indicator starts from the same data and follows the same rules. They are stated once here, so each indicator below can be read on its own.

• Source. The open public procurement data published by the Romanian Government on data.gov.ro, from April 2018 to date. At every data update, all the indicators are recomputed from scratch.
• Three streams. Public money reaches suppliers three ways: direct purchases through the electronic catalogue, offline purchases (also direct purchases, made outside the catalogue and reported through an award notice) and tenders. Each indicator states which of them it works on.
• Money spent only. For tenders, what counts is signed contracts and subsequent contracts. Framework agreements are left out: their value is the most that may be bought under them, not spending, and the money flows through the subsequent contracts. A lot covered only by a framework agreement has no money spent yet.
• Year. Money is booked in the year it was spent: the finalisation date for direct and offline purchases, the contract signing date for tenders. An offline purchase whose finalisation date is not known is dated by its publication. Where an indicator uses another date, its steps say so.
• The direct-purchase ceiling. The ceiling applied is the one in force on the day the purchase was published, for its contract type, because the law applies to procedures started after it took effect. The ceilings changed on 4 June 2018 and on 10 September 2022; their table is under indicators #01 and #02.
• Lots and associations. A contract covering several lots is split between them by each lot's estimated value. A contract won by an association is split between its members, so the same money is not counted twice.
• Amounts. All amounts are in lei, rounded to the leu.
• Thresholds and minimums. The direct-purchase ceilings come from the law. The other thresholds are chosen from each indicator's national distribution, so that an indicator flags a few percent of the possible cases, not half of them. The minimum values and counts remove cases too small for a percentage to mean anything: over three purchases, any percentage is close to chance.
• Partial years. The first year of the data begins in April 2018, and the current year is not over. The charts by year mark them as partial, and the indicators that compare years allow for it.
• Different units. The indicators count different things: a purchase, a lot, an authority–supplier pair, a market, a year of an authority. That is why the cases and the money of two indicators are never added together. What can be combined is who was flagged.
• A case is not an accusation. A flagged case is a statistical anomaly, not proven wrongdoing. Each one needs checking in context, in the procedure documents on e-licitatie.ro. The signal is stronger when several indicators flag the same authority or the same supplier; the indicator dashboard shows those overlaps.

High risk

#01 Artificial slicing of direct purchases

A large need is covered by several direct purchases from the same supplier, in the same CPV group and the same year, each below the legal ceiling (today 270,120 lei for goods and services and 900,400 lei for works; half that until September 2022). Together they pass the ceiling, and the tender the law requires is avoided. Computed on direct purchases only.

A flagged case is: authority–supplier–CPV–year group

Why it matters

Procurement law forbids splitting a contract into several smaller ones to avoid a competitive procedure: the estimated value is reckoned over all purchases of the same kind in a year. A direct purchase is quick and has no competition, so a large need bought in pieces from the same supplier sidesteps exactly the rule that would have brought other bidders in. Same supplier, same product family, same year is the classic pattern of slicing.

How it is computed

Computed on every direct purchase from 2018 to today.

1. Direct purchases with a supplier (CUI) and a CPV code are kept.
2. They are grouped on five keys: the authority, the supplier, the first three digits of the CPV code, the year and the contract type.
3. Groups with a single purchase are dropped: one purchase alone is not slicing.
4. For each group the purchase values are added up, and the average and the largest are computed.
5. The ceiling is the one in force on the day each purchase was published, for its contract type. In the years the ceiling changed (2018, 2022) the group splits in two, each part with its own ceiling.
6. The group is flagged when its total passes the ceiling. The purchases are not judged one by one: their sum is what counts.

Thresholds

Ceilings exclude VAT (Legea 98/2016, art. 7(5)). Each purchase is compared with the ceiling in force on the day it was published. The ceilings and procedures, explained.

• At least two direct purchases to the same supplier, in the same CPV group and year

What it does not cover

Direct purchases only. Tenders and framework agreements are not counted.

Limits of the method

Grouping uses the first three CPV digits, so different products of the same family land in one group. That is how the legal ceiling applies too, but it can join purchases that do not replace one another.

How to check a case

Open the list of the group's purchases (the icon at the end of the row). Close dates, near-identical descriptions and values that each stop under the ceiling point to a split need. A foreseeable need (supplies for the whole year, works on the same building) weighs more than one that arose along the way. Many small purchases against a large total is the clearest sign.

See all the cases

High risk

#02 Direct purchases at the legal ceiling

Direct purchases whose value stops just below the legal ceiling. Free prices do not bunch under a limit; when many values stop exactly there, the amount was chosen for the ceiling rather than for the need. Computed on direct purchases only.

A flagged case is: direct purchase

Why it matters

The direct-purchase ceiling is an administrative limit, not a market price. If values were set by need and price alone, they would spread smoothly below the ceiling; a pile-up just under it shows the amount was fitted so the purchase could stay direct. It usually means either a larger need cut down to the ceiling, or a price raised up to it.

How it is computed

Computed on every direct purchase from 2018 to today.

1. Finalised direct purchases with a supplier (CUI) and a value are kept.
2. The ceiling is the one in force on the day of publication, for the contract type: until September 2022 it was half of today's.
3. The value is divided by the ceiling and expressed as a percentage.
4. The purchase is flagged when the percentage is between 95% and 100%.
5. Purchases above the ceiling are not counted: there a tender was mandatory anyway.
6. Each purchase is judged on its own; nothing is added together.

Thresholds

Ceilings exclude VAT (Legea 98/2016, art. 7(5)). Each purchase is compared with the ceiling in force on the day it was published. The ceilings and procedures, explained.

• Flagged band: 95% – 100% of the applicable ceiling

What it does not cover

Direct purchases only, each one on its own. Several small purchases to one supplier are indicator #01, slicing.

Limits of the method

A value close to the ceiling is not in itself a breach: it can come from an honest estimate. The signal is repetition: the more of one authority's purchases stop just below the ceiling, the less chance explains it.

How to check a case

Count how many purchases of the same authority, or to the same supplier, stop under the ceiling: one is chance, ten are a pattern. Compare the description with the market price and check whether the same need was bought again in the same year: if so, read the case together with indicator #01.

See all the cases

High risk

#03 New companies winning public money

Suppliers that won public money within 90 days of registration. A company with no accounts, staff or experience that wins straight away may have been set up for that very contract. Each row is one company, with what it won in those 90 days in each stream: tenders, direct purchases and offline purchases.

A flagged case is: new company

Why it matters

A company a few weeks old has no balance sheet, staff or past contracts to prove its experience. When it wins public money straight away, it may be a vehicle set up for that contract, tied to someone in the authority or to another supplier. The signal is strongest in tenders, where the qualification criteria should stop precisely the companies without a track record.

How it is computed

Computed on every award from 2018 to today.

1. The registration date comes from the trade register, by CUI. Where the register holds several dates, the earliest is kept.
2. The three streams are walked separately, each by its own date: the contract date for tenders, the finalisation date for direct purchases, the publication date for offline purchases.
3. Only awards within 90 days of registration are kept, the registration day included.
4. Per stream and per company: the number of contracts, the value, the date of the first win and the days until it. Of a tender won by an association only the company's share counts; a direct purchase counts as one contract.
5. The three streams go on one row, and their values are added into a total.
6. Only companies with a total above 10,000 lei are kept.

Thresholds

• Critical risk: first win at most 30 days after incorporation
• High risk: first win between 31 and 90 days
• Minimum value: more than 10,000 RON in the first 90 days, across all streams

What it does not cover

Only the company's first 90 days count. A company under 10,000 lei in that window does not appear. Subcontracting is not in the data.

Limits of the method

A new company is not necessarily a shell: entrepreneurs set up new companies for a project, and a restructuring brings a new CUI. The signal is in the combination: a very young company, a large sum and above all a tender won days after registration, when it could not prove any experience.

How to check a case

Look up the company's shareholders and directors in the trade register, and their ties to the authority or to its other suppliers. Look at what came next as well: a company that disappears after the contract or never works for anyone else is more suspect than one that went on to find other clients. The row shows, for each stream, how many days passed before the first win.

See all the cases

Medium risk

#04 Supplier captivity

A supplier takes almost all of its public revenue from one authority, and the authority gives it a large share of its budget. Such a relationship points to a closed local market that other suppliers have left.

A flagged case is: authority–supplier pair

Why it matters

The two percentages describe a dependence both ways: the supplier lives off the money of a single authority, and the authority gives it a large part of its budget. In an open market a capable supplier sells to several buyers, and a large buyer buys from several suppliers. So closed a relationship can mean a steady preference, or a local market where the other suppliers have stopped competing.

How it is computed

Computed on every award from 2018 to today, in all streams.

1. The money from the three streams is added together: tenders, direct purchases and offline purchases.
2. For tenders each winner's share is used, so a contract with two winners is not counted twice.
3. The money is grouped by authority–supplier pair, in total and per stream.
4. Each supplier's total public revenue and each authority's total spending are computed.
5. The pair gets two percentages: how much of the supplier's revenue comes from this authority, and how much of the authority's budget reaches this supplier.
6. The pair is flagged when both percentages pass the thresholds below and the supplier's revenue is large enough for the percentage to matter.

Thresholds

• Captivity of at least 80% of the supplier's public revenue (the list can be widened down to 50%)
• Captured budget of at least 30% of the authority's spending
• Total public revenue of the supplier above 100,000 RON

What it does not cover

The indicator measures where the money concentrates, not whether it was lawful. A captive relationship can also mean the local market has only one capable supplier.

Limits of the method

The percentages cover the whole period in the data, not a year. One intense year looks the same as seven steady ones.

How to check a case

See which stream the money went through: a relationship built on direct purchases says something different from one built on tenders won. For tenders, check under indicator #05 how many offers came in. Look for ties between the supplier and the authority, and whether the supplier had other clients before or after.

See all the cases

High risk

#05 Tender lots with a single bidder

Lots in publicly advertised tenders that drew a single offer and on which contracts have already been signed. It is the risk indicator most used in European procurement research: no competition in an open procedure can mean a closed market or specifications written for one bidder. Computed on tenders only, counting only money spent.

A flagged case is: tender lot

Why it matters

In a publicly advertised procedure, a single offer means nobody else wanted to, or could, compete. The single-bidder rate is the risk indicator most used in European research and in the European Commission's procurement reporting, because it rises where specifications are written for one bidder, deadlines are too short, qualification criteria are restrictive, or companies have shared out the market. Without competition, an authority usually pays more.

How it is computed

Computed on every tender from 2018 to today.

1. Lots on which money was spent are kept: those with at least one contract or subsequent contract carrying a value and a winner. Framework agreements are left out, because the money is spent through the subsequent contracts.
2. Only competitive procedures with publication are kept. Negotiation without prior publication is left out: there the authority invites companies directly, so a single offer is normal.
3. The offers received on a lot are the unacceptable, non-compliant and admissible ones, added up. Withdrawn offers are not counted. For the authority's own procedures, with no count per lot, the contract's count is used.
4. The lot is flagged when it received a single offer.
5. A contract covering several lots is split between them by estimated value and booked in the year it was signed. With no year chosen each lot appears once, with its total across all years.
6. The winner shown is the company with the largest part of the lot's money that year; for an association the contract is shared evenly among members.

Thresholds

• Single bidder: one offer received on the lot, counting unacceptable, non-compliant and admissible offers together; withdrawn offers are not counted
• Procedures included: open and accelerated open tender, simplified procedure, restricted and accelerated restricted tender, competitive and accelerated competitive negotiation, competitive dialogue, the authority's own procedures and request for offers
• Only lots on which at least one contract or subsequent contract has been signed

What it does not cover

Tenders only, and money spent only: a framework agreement with no subsequent contract is not counted. Negotiation without prior publication and direct purchases are not included.

The data holds at most 200 contracts per notice. Where a notice has more, the money from the missing contracts is not in the figures.

Limits of the method

The offer count is the one the authority reported. On under 1% of lots the contracts report different counts; the largest is kept. A single offer does not prove a breach: in a market with one manufacturer it can be natural. The signal is repetition, at the same authority or with the same winner.

How to check a case

Open the notice on e-licitatie.ro (the number in the first column) and check the submission deadline, the qualification criteria, the clarifications requested and any complaints. A single lot says little; the same winner with a single offer, year after year, at the same authority, says a lot. Indicator #06 shows how often it happens at each authority.

See all the cases

High risk

#06 Authority ranking by single-bidder rate

The single-bidder rate is the share of single-offer lots among all of an authority's lots on which money was spent. It helps set priorities: an authority well above the national rate deserves a look first. The ranking is ordered by the money spent on single-bidder lots. Computed on tenders only.

A flagged case is: ranked authority

Why it matters

A single-bidder lot is a case; an authority where half the lots draw a single offer is a pattern. The ranking takes indicator #05 from the lot to the authority and sets it beside the national rate, to show where the lack of competition is a habit, not an accident. It is meant for choosing which authorities to check first.

How it is computed

Computed on the same lots as indicator #05.

1. For each authority, the lots in competitive procedures with prior publication on which at least one contract or subsequent contract has been signed are kept.
2. Only lots whose offer count is known are counted; the others are reported separately and stay out of the rate.
3. The single-bidder rate is the number of single-offer lots divided by the number of lots counted.
4. The money spent is added up over all lots and over the single-bidder ones; their ratio is the share of the money.
5. Only authorities with at least 5 counted lots are ranked. The national rate, however, is computed over the lots of every authority.

Thresholds

• Rate = single-bidder lots divided by the lots counted, times 100
• Only authorities with at least 5 counted lots are ranked

What it does not cover

Only authorities with at least 5 counted lots are ranked; the others count only towards the national rate. Negotiation without prior publication, direct purchases and framework agreements without subsequent contracts are not included.

The data holds at most 200 contracts per notice. Where a notice has more, the money from the missing contracts is not in the figures.

Limits of the method

The rate covers the whole period. An authority that changed its practice recently looks the same as one that did not. A few large single-bidder lots give a large share of the money at a small rate.

How to check a case

Compare the authority's rate with the national one, then the share of money with the share of lots: a high money share at a low rate means a few large lots with no competition. Then open indicator #05 filtered to the authority to see the lots and who won them. On the dashboard, an authority counts as flagged from a rate of 50%.

See all the cases

High risk

#07 Price matching the estimate exactly

The money spent on the lot is exactly the authority's estimated value, to the leu. With competition the price falls below the estimate, by amounts that vary from case to case; here many cases land right on the estimate. The estimate is public in the contract notice, though, so bidders know it: the likely explanation is not a leak, but that nobody had a reason to go below it.

A flagged case is: tender lot

Why it matters

A tender price is the outcome of competition, so it should land somewhere below the estimate, by margins that differ from lot to lot. Freely formed prices do not pile up on one exact value. When the money spent equals the estimated value to the leu, the price was not pushed down by competition: the winner asked for exactly what it knew the authority was willing to pay.

How it is computed

Computed on every tender from 2018 to today.

1. Lots on which money was spent are kept: those with at least one contract or subsequent contract carrying a value and a winner.
2. Competitive procedures with publication are kept, as for #05, without framework agreements and without lots lacking an estimated value.
3. A contract over several lots is split by estimated value, and the parts of one lot are added up across all years.
4. The lot is flagged when the money spent and the estimated value, both rounded to the leu, are equal.
5. Near misses are not flagged: on a large lot, a tenth of a percent is hundreds of thousands of lei. The anomaly is the bunching right on the estimate.
6. The offers received on a lot are the unacceptable, non-compliant and admissible ones, added up; withdrawn ones are not counted. The second view keeps the lots with at least two offers.

Thresholds

• Exact match: the money spent on the lot and its estimated value, both rounded to the leu, are equal
• Two views of the same table: every match, and only those with at least two offers
• Only lots on which at least one ordinary or subsequent contract has been signed

Price matching the estimate although there was competition

The same lots, only those with at least two offers. With a single offer, a price equal to the estimate is easy to explain: there was no one to beat. With several offers the simple explanation goes. Matches fall from about 16% of single-offer lots to under 1% from three offers up, so this view shows what indicator #05 does not already show.

What it does not cover

Tenders with prior publication only, and money spent only. Framework agreements are left out: there the estimate is the agreement's ceiling, and a match means the agreement was used up, not a price. Lots without an estimated value cannot be checked.

The data holds at most 200 contracts per notice. Where a notice has more, the money from the missing contracts is absent, and the lot can look below the estimate.

Limits of the method

The estimate is public, so bidders know it. An exact match does not prove a leak or collusion: it shows nobody had a reason to go below it. The signal is repetition (the same authority, the same winner) and above all the second view, where there was competition and the price still did not fall.

How to check a case

Look at the number of offers first. With a single offer, the match is explained by the lack of competition (indicator #05); with two or more, the question is why the competing offers did not bring the price down: they were rejected, they were dearer, or they were submitted for show. The procedure documents on e-licitatie.ro show the offers and the reasons for rejection.

See all the cases

Medium risk

#08 Year-end spending

Every authority buys in December, and the share of those three weeks varies from year to year: 9.5% of the money in 2022, 4.1% in 2023. So the indicator compares each authority with the national average for the same year: it is flagged when it puts at least three times as much into the window. The table is ordered by the amount spent above normal.

A flagged case is: authority-year

Why it matters

The public budget is annual: appropriations not spent by 31 December are lost. Authorities that have put off spending buy in a hurry at the end of the year, sometimes with no real need or without comparing prices, just so as not to lose the money. Because the whole country spends more in December, the indicator does not look for authorities that spend in December, but for those that do so far more than the rest of the country in the same year.

How it is computed

Computed on every direct purchase from 2018 to today.

1. Direct purchases are taken by finalisation date.
2. They are grouped by authority and year. A year counts only if the data covers all of its December, so the current year is left out.
3. The first year in the data starts in April, so its total is estimated for a full year from the nearby complete years. The correction makes the total larger, so it can only remove a case, never create one.
4. Spending from 10–31 December is divided by the yearly total: that is the authority's share.
5. The same share is computed for the whole country in the same year. It varies a lot (9.5% in 2022, 4.1% in 2023), which is why no fixed threshold is used.
6. The authority's share is divided by the national one. The amount above normal is the window spending minus what the authority would have spent at the national share.
7. Authority-years at least 3 times the national share are flagged, with at least 5 purchases in the window and at least 100,000 RON above normal, among those that spent over 200,000 RON in the year.

Thresholds

• Window: 10–31 December, by the direct purchase's finalisation date
• Flagged: at least 3 times the national share, at least 5 purchases in the window, at least 100,000 RON above normal, on a yearly budget above 200,000 RON

What it does not cover

Direct purchases only. Tenders are left out: they are prepared months ahead, so a December signature is not a last-minute decision. Actual payments are not in the data.

Limits of the method

Year-end spending has a legitimate side: the budget closes on 31 December and unspent money is lost. The indicator does not say a purchase is unlawful, only that its timing departs from the national pattern. A year in which everyone spent late produces no flags.

How to check a case

Open the authority's purchases of 10–31 December of the flagged year. Stock goods bought in large quantities, equipment that could not have been delivered before the year closed, or the same suppliers again and again point to spending done to use up the budget. The year 2023 is missing data at the end of December; the indicator's page says how many days of the window are covered.

See all the cases

Medium risk

#09 Discretionary spend index

The discretionary spend index (DSI) is the share of an authority's money spent without a tender: direct purchases plus offline purchases, out of everything it spent. It is not in itself a sign of fraud, because a direct purchase is legal below the ceiling. But a large share of a large budget shows how many decisions are taken without competition. One row is one authority in one year.

A flagged case is: authority

Why it matters

A direct purchase is lawful below the ceiling, but the authority decides it alone, without competition and with little transparency. The index shows what share of an authority's money goes that way. At a large authority, a high share means many decisions taken without competition and, often, needs that could have been gathered into a tender.

How it is computed

Computed on every award from 2018 to today.

1. Direct purchases and offline purchases are taken. Both are direct purchases; an offline purchase does not even go through the catalogue.
2. Tender contracts are taken separately, by the value signed. Framework agreements are left out.
3. Money is dated by when it was spent: the finalisation date for direct purchases and offline purchases, the contract date for tenders. If an offline purchase has no finalisation date yet, its publication date is used.
4. It is grouped by authority and year. Each authority also has a row with its total across all years, shown when no year is chosen.
5. The index is the two direct streams divided by everything the authority spent, as a percentage.
6. Only authorities with at least 3 million lei in a year, or 20 million across all years, are included. Below that the index says nothing: a small authority need not run tenders, and almost every row would read 100%.

Thresholds

• Minimum spending: 3,000,000 RON in a year, 20,000,000 RON across all years
• Above 60%: a strongly discretionary profile

What it does not cover

The indicator measures how the money was spent, not whether it was lawful. It does not check whether a purchase passed the ceiling; indicators #01 and #02 do that. Actual payments are not in the data.

Limits of the method

A high index can have legitimate reasons: an authority with no investment in a year spends almost only on running costs. Read it with indicator #06, which shows how much competition the tenders drew.

How to check a case

Compare the index with that of similar authorities and with the same authority's earlier years: a sudden rise says more than a steady level. Read it together with indicators #01 and #02, which show whether the direct purchases sidestepped the ceiling, and with #06, which shows how much competition there was in the tenders. On the dashboard, an authority counts as flagged from an index above 60% in a year.

See all the cases