Project managers rigorously assess and plan the delivery of projects to avoid the nightmare of going over cost. But what if there was a more mathematical way to quantify this? This is a question global financial services firm, State Street, asked when it deployed predictive analytics.
Technology plays a major part in the delivery of State Street's products and services to clients worldwide. Being able to accurately quote clients and ensure quoted delivery times are feasible and won't result in cost overruns is crucial for the organisation.
As projects are usually complex, accurately estimating overall cost and time to deliver is difficult and could easily miss the mark. This is where applying mathematics to data comes in handy.
“There was a project where we avoided US$200,000 worth of cost. When we ran the predictive model through, we could see that it was going to take much longer than what they [the client] were hoping; 10 additional weeks it was going to take us to do this right,” said Scott Lancaster, VP of State Street at Predictive Analytics World in San Francisco.
“So we could work with the client by sitting down and negotiating with them. If you don't have data or a model to back it up, it's very hard to discuss that with the client,” he said.
State Street analysed historical data it has collected on its projects over the years, looking at the scope, team sizes, capability of the team, the amount of hours each team member spent, and ultimately, how well it delivered on these programs.
“It's also the tools they use, the processes being used, and technical difficulties,” Lancaster added.
Based on this data, State Street was able to build a predictive model to better inform clients of the cost and time needed to deliver on their projects.
“We want to model this and understand the non-linear relationship between this data. So we take that data and feed it into the predictive model, and we predict the effort, the cost, and the schedule.
“I need to be able to model this in real time when I'm working with the client as they list their requirements,” Lancaster said.
Predictive analytics is not only used to negotiate projects with clients but also to show different options or work around time constraints, Lancaster said.
“What if they say 'I don't care about cost, just time.' We'll it would take triple the amount of staff to meet that timeframe. And we are not going to ramp up that many resources in a short amount of time, it just doesn't work," he said.
“Using the predictive model, I can ask 'how much can I get done in this timeframe?' The model spits out that you'll get 80 per cent of your scope done. That's not terrible, but it's not everything. But maybe they are not too fazed if some stuff comes later. So we are able to sit down with the client and show the various options,” he said.