Skip to main content Skip to Footer


From high risk to high reward

An Advanced Analytics Solution to Prevent High-Risk Procurements


In today’s budget environment, federal procurement organizations need to avoid high-risk procurements for a multitude of reasons. Not only can high-risk procurements mean higher costs and potential for fraud, waste and abuse, but they can also result in unwanted scrutiny from oversight bodies, legislators and the media—and the general public too. Clearly, high-risk procurements are a high-stakes proposition, and the time has come for organizations to start thinking differently about them to drive reform.

Accenture’s experience reveals an exciting new solution. What if your procurement organization could identify high-risk procurements before they occur? You can. Advanced analytics techniques can proactively identify risks associated with procurements; build near real-time visibility into these risks; and provide actionable feedback in the procurement processes and systems before the contract award occurs. Discover the benefits of rethinking the approach to high-risk procurements and the new value that advanced analytics techniques can bring for procurement excellence in today’s environment.


Agencies have historically monitored procurements based on the dollar value of a contract. While this approach makes sense in an environment with a high volume of procurements, it can permit risky procurements to proceed undetected.

Our research has found that

  • Up to 40 percent of the procurement awards have prices that exceed the historical value of the previous buys, even when accounting for inflation. This represents a significant opportunity for savings in cost of goods procured.

  • Suppliers who display a significant amount of risk for performance issues or fraud are included in the list of active bidders for government contracts.

  • There is a significant risk from companies providing counterfeit parts that support critical military and aerospace equipment applications (validated in recent congressional testimony and a Commerce Department report.)

However, even in cases where procurement actions are manually reviewed for appropriate due diligence, procurement officials find themselves constrained by time consuming and cumbersome research, most of which is performed outside of the agency’s procurement system. At best, this leads to a backlog in procurement workload and, at worse, a disregard for or limited mandatory reviews, thus increasing the likelihood of risky contract awards.


Accenture has found that a combination of advanced analytics techniques like data mining, predictive modeling and network analysis provide alternate approaches to identifying risk. These techniques can uncover the characteristics of high-risk procurements and develop models that capture the relative risk related to them.

Scoring models can then be built that “grade” the risk associated with each procurement. Once graded, the high-risk procurement information can be incorporated back into the procurement process, enabling further review that can reduce high-risk procurement exposure.

This advanced analytic approach counters the problems found in a traditional “business rules only” approach to identifying high-risk procurements. By using multiple risk factors through this hybrid approach, agencies move beyond simple business rules related to individual procurement dollar value and consider other factors for better visibility, control and decisions.

Further, these analytic solutions can be integrated with the organization’s procurement system, thus eliminating the need for “swivel-seat” procurement research.


Accenture has developed an approach using advanced analytics and data mining techniques to identify likely high-risk procurements.

  • Collect data

    Data is collected on previous high-risk procurements and specific characteristics and timeframes are encoded into the database. For example, were previous procurements problematic due to pricing, competition or compliance issues? Additional outside data from industry leading supplier and product content providers (such as Dun and Bradstreet) are also used to enrich the database. By codifying this information, the data mining and advanced analytics software engines can later use this prior experience to develop the model.
  • Build advanced analytic models

    The data mining and advanced analytic models are run and output risk scoring models are developed. Through our relationship with numerous analytical software partners, including our strategic alliance with SAS (an analytics software leader), we use commercial off-the-shelf software products with advanced analytics techniques. The data mining and advanced analytics process uses logistic regression, neural networks, decision trees, “Dmine” regression, gradient boosting, Bootstrapping, social network analysis and Ensemble approaches to test and build the risk models. The software identifies the factors that have the greatest predictive power and incorporates them into prediction models, which create a probability that a given procurement would be high risk.

The outputs from those risk models are then used in the advanced analytic process to build the risk scores associated with each procurement. When this process is complete, the chosen model (or models) can be adjusted based on how well they capture true risk and identify risky procurements.

  • Validate the risk scoring models

    The selected model is used to evaluate a “hold-out” subset of data that was not used to build the models. This protects the process against model over-fit, a condition where the model places more emphasis on a variable than might be found in the larger population of procurements. This step offers procurement organizations confidence in the model’s effectiveness identifying high-risk procurements outside of the statistical modeling environment.
  • Transform the procurement process

    Realizing the full benefit of this process means incorporating the scoring model into enterprise resource planning or other procurement systems. The risk identification approach is likely to be most effective when incorporated into an automated daily review of procurements. This review should result in workflows that require higher-level oversight and review of high-risk procurements, and track metrics around the number and dollar value of those procurements. This way, management can address pockets of risk where they are occurring and track trends over time to reach desired outcomes.