Case: Construction project risk
The customer, a major Nordic construction company, wanted to improve its project risk management system. They used a basic scoring process and wanted an integrated, online system that could handle up to several tens of risk factors simultaneously. Such systems are not commonly available for construction company use. Off-line products exist that are designed for standard processes and are hence of limited use in an environment where many essential factors change in each project. Continuing to trust on intuition would have been extremely risky. Therefore, BayesIT was called to design a risk management solution to suit their new needs. The development continues with focus on re-organizing the customer feedback process to support risk analysis.
The customer had a manual risk management system installed in the organisation. The system was developed over the years. It used static weighing factors and scoring done by seasoned managers. It provided sufficient functionality as long as the construction site was entirely in their own hands. With the increased use of outsourcing, the system was not capable of assessing all the risks such as subcontractors' use of workers of foreign origin with poor language knowledge. Statistics using accounting data from sales and logistics was no longer sufficient.
The customer faced huge cost pressure because of the downturn in the commercial building construction sector. The problem was not only the increase of labour cost and its impact on profit. The customer had also closed deals too optimistically because of out dated knowledge about company competence. The last point is extremely critical in an environment where there is a limited amount of new projects on the market.
Bayes Information Technology Ltd., being a leading provider of Bayesian modelling, was called in to solve the problem. BayesIT designed a risk model for semi-standard construction projects including a manual feedback form to collect data about the construction site activities. The accounting data was complemented with this additional data, and a Bayesian Network model was created. In the first phase the data was combined using a standard spreadsheet, and a new risk model was created when new data was available.
Initially the customer did not have an intranet data collection system in use, so the customer had to manually collect sufficient data from the construction sites. In this development phase, when the number of projects was low, this was a very cost-effective solution, as the forms were simply fed in the company database by a secretary. When the form and its questions were considered mature enough, the customer switched to Internet based site data collection. It took three generations of form development in order to focus in on the right variables and formulate the questions well enough
The solution was designed so that it can be easily integrated to an ERP as well as a CRM system for customer feedback data. Adding a new project to the knowledge base e.g. every day does not require touching the system solution at all. There is no need for special software, its updating or expensive training.
The Bayes Information Technology solution guarantees agility and quality of the risk knowledge model. Cutting edge technology ensures that risk identification is accurate also when data is sparse as it always is when making a bid for a new project.
With the risk management system designed by Bayes Information Technology, the customer could automate the risk decision support so that a managed decision could take place every day if necessary. The replacement of scarce expertise saved a lot of costs. During the first year some latent catastrophe cases were avoided saving approx. 1 million euros. This would not have been possible using the manual, traditional scoring approach.
The cost of the development project was approx. 100.000 euros and it took about two years to develop. The major part of costs was related to data collection development. Database etc. system updates caused only small additional costs. At the time of order, the payback period was estimated to be 2 years. In reality, the system paid back in less than one year as some disaster projects were avoided already in the beginning of the project. Additionally, the customers' business control was improved significantly.