Analysis of historical data is a good basis for most decisions about changing rules or approaches. Calculations help to predict the effect of innovations. Therefore, we always support the use of analytics to develop changes in the field of public procurement.

Recently, our team has received a request to analyze construction auctions. And using this example, we will show how data can be used to find answers for future decisions.

Brief background: There is an opinion that for some construction works, the auction stage may not make as much sense because these works are estimated. According to the hypothesis, any reduction in price by the winner during the auction is highly likely to be dumping. As a result, the terms of the contract will have to be changed, or the contract will be terminated altogether due to the impossibility of performing it at the proposed cost.

Analysis question:

  1. How common is it for the winning bidder to underbid in construction auctions (compared to the procurement system in general)?
  2. How does the reduction of the bid amount by the winners of construction auctions affect the progress of the contract?

First you need to select the data itself. The sample was as follows:

  1. Year of announcement: 2019. During that year, the procurement system was still operating as usual (there had been no changes to procurement related to COVID-19 yet), and most of the contracts started in 2019 should have been completed by today. Thus, we can consider 2019 to be the most recent “full” year.
  2. The status of the contract: completed (that is, both fulfilled and terminated contracts). This filter is needed to have statistics only for contracts with a higher probability that all information about the progress of their execution has been entered into the system.
  3. For the selection of construction, we chose:
    1. CPV code — “45000000-7 Construction works and maintenance repairs”
    2. type of procurement item — “works”
  4. Types of procurement method — open tenders and open tenders with English language because in 2019, there was an obligation to report on the progress of the contract only under these categories.
  5. Expected cost: at least 1 million, as there was a request to consider the behavior of participants in large-scale procurement.
  6. Only the initial and final offers of the participant in procurement were taken into account. Bidders who offered lower prices but were disqualified were not taken into account. In those situations, it is often difficult to tell whether the participant was disqualified for formal reasons, whether they prepared incorrect calculations, or they did not confirm they were able to fulfill the contract.


In total, 4,561 construction contracts were signed in 2019 which have been completed to date. This is less than 5% of all contracts in 2019.

In 45% of cases in construction contracts, the winner reduced their price during the auction, which is 3% more often than the Prozorro system average. The higher the expected purchase price, the more likely the winner was to lower their price during the auction.

So, the auction does encourage lowering the price. However, it is difficult to predict based on the available data what strategy these same participants would have chosen if Prozorro did not have auction functionality (and they could not adjust their bid):

  • would the bidders immediately submit the lowest bid price at which they are willing to perform the contract?
  • would the contract amount be lower if the winner was determined by the initial bid?

On the other hand, the auction rule “the lowest price yields has the biggest advantage in the auction” seems to have an effect. In more than 98% of all these contracts for construction (including cases where the participant did not negotiate), the initial offer of the winner was at least one hryvnia less than the expected cost. An alternative explanation is that the bidders wanted to draw less public and regulatory attention to their procurement and simulated a price cut in their initial bid.

If we look at the progress of construction contracts where the winner lowered the price during the auction, in 83% of cases, either changes were made to the contracts (which did not always concern the prices) or these agreements were terminated. The larger the amount of the contract, the higher the percentage of such contracts.

However, if we look at the procurement of construction works where the winner did not change the prices at the auction, more than 85% of the contracts were either terminated or changed.

It appears that the participant’s price reduction during the auction has no effect on the performance of the contract. Further research is needed as to why there is a higher percentage of contract changes in purchases where the winner did not lower the price during the auction. One of the versions is that the technical terms of reference for the works were rather vague, and it was necessary to make changes at the execution stage. And this, in turn, could be one of the reasons why the participant did not lower their price during the auction.

If we look at the system as a whole, the execution of works contracts significantly deviates from the general trend. Change or termination of the contract in general were recorded in 56% of all contracts in the system, which is almost 30% less than this figure in construction.

The most likely explanations for this are as follows:

  1. Customers more reliably reflect changes in their construction contracts in Prozorro than in other types of procurement;
  2. It is much more difficult to execute construction contracts without making changes to them, which significantly distorts the results of the auction. Accordingly, it is necessary to think of a mechanism to minimize the number of changes/terminations of the contract or to make the process of making changes more understandable and transparent.


It should be understood that since this analysis is brief, the findings will also be short and raise new questions. We see that the problem with changes in construction contracts and their termination really exists. However, it does not currently appear to depend on the availability of auctions for such purchases. Moreover, almost half of the participants reduce the price at the auction, and this is slightly more than the system average. Here are some new questions to consider:

  • what strategies the business would choose if there were no auctions;
  • why there is a higher percentage of contract changes and cancellations in purchases where prices were not reduced;
  • how to minimize the number of contract changes/terminations or make the process of making changes more clear and transparent.

This publication has been prepared with the financial support of the European Union. Its content is the sole responsibility of Transparency International Ukraine and does not necessarily reflect the views of the European Union.