How smart algorithms can help in informing customers correctly about parcel delivery times
Large companies usually have one or more challenges that can be fixed by using data, but are very complex and dependent on the organisation. The solution for fixing these can lie in Deloitte’s IDO approach.
Naser Bakhshi, Olga Hartoog, Paul Langelaan - 30 november 2017
When you order something online, you are often informed of the expected delivery time slot: “We expect to deliver your parcel between 9 and 11,” for example. You then arrange to stay at home for a few hours that morning to be able to accept the parcel. However, at 11 you really have to leave, but the parcel hasn’t arrived yet. When you get home in the evening, you find the little note on the doormat: “We did not find you at home and we will try again tomorrow.” They had probably got there just after you left. Annoying, isn’t it?
PostNL agrees, and finds it very important to give their customers correct information about delivery times. Especially in the period between Black Friday and the holiday season, it is vital to inform recipients correctly. Due to the enormous increase in volumes there will inevitably be challenges in the operation, possibly resulting in different delivery times than expected. Research within PostNL has shown that customer dissatisfaction occurs mostly when they are provided with incorrect information and when expectations are not met. Not only does correct information result in more satisfied customers, it also makes the organisation more efficient and reduces the demand on its customer call centres. The Time Slot Indication (TSI) algorithm they were using was not, however, predicting the time slots well enough. Customers nowadays expect excellent, personalised service; they already experience this with products and services offered by strongly data-driven companies. PostNL understands this, and therefore the problem of Time Slot Indication (TSI) had already been high on their agenda for a quite some time. Many internal and external experts had tried to tackle this problem, but their efforts had not yet led to the customer really experiencing an improvement.
The TSI problem is a complex one. Parcel volumes are increasing enormously due to the growth of e-commerce. Services such as “Ordered today, delivered tomorrow” and even “Same day delivery” make it hard to predict what will have to be delivered on any given day. Furthermore, traffic and other external factors add to the uncertainty of delivery times. At the same time, you’re dealing with an existing organisation involving tens of thousands of operating personnel - many of them subcontractors - who have a certain degree of freedom in their work. For instance, a parcel deliverer can decide in what order he or she delivers the assigned parcels in a certain area. This is why geospatial route optimization cannot be the ultimate solution. And even if you could let the algorithm decide the order of delivery, wouldn’t it be useful also to use the knowledge of the parcel delivery driver somehow?
Technology itself is not enough
Large amounts of operational data are available: up to a million parcels are processed every day and they are scanned at every move they make. Coming up with a sustainable solution to a challenge as complex as Time Slot Indication, however, will require more than just looking at the data. Technology alone won’t be enough. Just trying to change how people work in the organisation won’t do the trick either. The solution lies in a combination of the two. Apart from strong involvement of business representatives with decision-making power (mandate), an ‘elite’ team with deep data science knowledge and capabilities as well as business sense and understanding of the organisation is needed. Hiring external consultants for a short-term project isn’t really the answer, but finding (and freeing up!) this elite team inside the organisation has also proved to be very hard.
A perfect set up for tackling an organisation’s ‘TSI problem’ is teaming up the external expertise with internal resources in an alliance of longer duration. In May 2016, Deloitte Consulting helped PostNL kick off their transformation towards an Insight Driven Organisation (IDO) by founding Analytics Centre of Excellence (ACE). The choice of a centralised operating model was the result of an extensive Analytics maturity scan and analysis. ACE started off with a group of experienced Deloitte data scientists and analysts from within the organisation. In April 2017, ACE became an official PostNL department as the team had more than doubled in size with the transfer of five eager data scientists from various corners of PostNL.
Within the space of eighteen months, ten different business challenges have been picked up and addressed by ACE, from Parcels to Mail and from commercial to operations, and - with the help of data -, insights, actions, and predictions, even up to the implementation of models, have been arrived at for all of them. The new TSI algorithm is one of these solutions. After a period of brainstorming, data crunching, trying things out on parcel delivery drivers, and a great many simulations, ACE came up with this new solution that utilises the knowledge of delivery drivers. Simulations and small-scale pilots showed that up to 20% improvement in time slot ‘score’ would be possible.
In October 2017, a pilot with the new algorithm started across 25% of the Netherlands and resulted in scores that were even higher than expected, with an algorithm unique in its kind because:
- it uses scan information of the approximately 800K packages PostNL delivers each day
- drivers’ knowledge is used to steer the algorithm towards the best route
- it can be adjusted to unforeseen circumstances last minute
These facts, combined with results from simulations, persuaded PostNL to implement the solution nationwide. Millions of people in the Netherlands are now being given more reliable delivery time slots (16% fewer parcels outside their assigned slots) and having a better customer experience. And the design and implementation of the algorithm is only the first step: the project has also resulted in a data and simulations-based roadmap toward further improving the prediction of the time slots by changing the way of work and implementing additional functionalities. Fetsje Bijma, the business owner for the TSI project, on the co-operation with ACE and the approach for the project:
“I was pleasantly surprised by the speed with which the ACE team has mastered the problem. The genuine interest in the practical story behind the TVI problem has helped enormously in the search for the best solution. Together we explored elaborate as well as simpler models and weighed the pros and cons of each. Ultimately, the issue has been divided into several subproblems, among which several analytical issues and a number of operational issues. In the new roadmap, the solutions for these issues fit together nicely, so that we will have largely solved the persistent TVI problem within a year.”
The IDO approach
Every company has its own ‘TSI problem’: a seemingly insurmountable problem with a considerable impact. The Deloitte IDO framework and set-up can be the right approach for tackling these kinds of challenges. The journey with PostNL shows that with well-designed processes and methods, a balanced team in terms of hard technical skills and business sense, and a more extended alliance with external experts, even tough nuts like these can be cracked. Frank Ferro, ACE director:
“Data analytics is fun and walking through the process you try to optimize is a must!”
IDO is a structured methodology which helps organizations to become insight driven by utilizing the full power of data and data science in combination with deep business understanding. With the help of Deloitte’s IDO approach, PostNL has significantly advanced its analytics maturity over the past year, and continues to do so in order to keep living up to the expectations of the most demanding customers.
Is your organisation ready for IDO?