Is it possible to accurately predict the success of projects? It is indeed, if you ask Aleksi Lattunen, Silverbucket’s Scrum Master. Lattunen has just finished his master’s thesis.
The thesis is called “Predicting project execution based on a resource plan by means of machine learning”. The thesis was approved in May in the Faculty of Information Technology and Communication Sciences in Tampere University. So, why is predicting so important?
Early bird catches the worm
From Silverbucket’s point of view, the most important findings of the thesis are related to data accumulation and various key figures. However, the data accumulated from the resourcing tool could be even more extensive. In his thesis, Lattunen pays special attention to how to better look back in project
”We do not have the data where we could see the status of a project at different times in the history of the project. When it comes to completed projects, we just know whether they were successful or not. In addition, we know what the project looks like right now when it comes to its key figures. But what did a successful project look like, let’s say, three months ago when it was halfway through?” Lattunen asks.
So, it is important to know what a successful project looked like when it was halfway through. Better yet if one knows what a successful project looked like in the very beginning. It is obvious to project leaders why it is important to recognize issues sooner rather than later.
”The reason for that is very simple. At the beginning of the project, issues are easier to correct than at the end of the project. If one notices in the beginning that the project is in trouble, there is still time to react”, Lattunen reminds.
Signs of danger are easily visible
It is easy to spot projects in danger when the key figures are interpreted correctly. So, what are the most typical signs of danger?
”A clear sign that a project is going astray is when no hours are being accumulated for the project. Of course, it means that nobody is working on the project. Some previous projects may still be going on and the employees are booked. Naturally, an excessive number of hours is also a sign of danger. It means that the project may not stay within its budget. It may also be that hours are being accumulated, but the project is done by people who were not originally planned for the project. There might be substitutes working on
the project, for example. If the experts originally attached to the project are not available, problems may arise”, Lattunen lists.
Next steps at Silverbucket – towards successful projects
Lattunen's major is Information Technology. Originally, his master's thesis was supposed to be from the perspective of hypermedia. However, things changed a bit. The result is more about machine learning and project management. The thesis is a great example of the importance of interdisciplinary theory.
Next, Silverbucket will start collecting data in real life. Data are collected from actual projects within Silverbucket. And of course, the customer data are fully anonymized. In practice, a database is set up within
Silverbucket to collect precise key figures for the needs of machine learning analysis.
“It is not possible to generate data afterwards. We have to start collecting it now. After six months of accumulation, the data are starting to become useful. The entire life cycle of completed projects must be taken into account. Only then we will see which projects really succeeded and which did not”, Lattunen
In addition to collecting data, Silverbucket will start working on an analysis tool. This is where machine learning comes into play. Once the analysis tool finally is ready and some data have accumulated, the results can be showed in the Silverbucket interface along with all the other resourcing information.
”In this case, the customer can see that our machine intelligence thinks that his project will succeed with a 78 % probability, for example. The analysis improves over time as more and more data are accumulated”, Lattunen explains.
“Thus, the benefits of machine learning can be revolutionary for the future projects. At one glance, the customers see how likely their project is to succeed. If there are several projects going on, the project manager will see immediately see which projects need a little more love and attention”, Lattunen smiles.
”Aleksi’s thesis handles an important topic for Silverbucket – how to make projects succeed. The challenging part is to gather all the necessary data. But when we succeed in that, it will be very interesting to see what kind of factors make up the success, and failure, of certain types of projects. In this way, machine learning allows our customers to learn something completely new about their projects and operations”, praises Toni Uimonen, one of the Founders of Silverbucket.