Professor Richard Eglese, Department of Management Science, Lancaster University Management School examines Choice-Based Demand Management and highlights a missed opportunity for retailers – as well as a chance to improve relationships with customers.
The future success and viability of the boom in e-tailing is dependent on finding solutions to fundamental, practical issues around delivery. Broadcasting an offering of goods and services via the internet is a powerful platform. But delivering to customers – who might be ordering anytime and anywhere – at a time which suits them creates complex problems of logistics and costs which eat into profits.
The growing demand for these kinds of services can be illustrated by the British online grocery sector which grew by 17.2% in 2011, with sector sales being forecasted to rise by 79% over the next five years relative to levels in 2012 (Mintel 2012). Businesses must get the trade-off right between sometimes substantial order fulfilment costs and maintaining customer satisfaction and the principles are exactly the same for any kind of e-tailing transactions other than grocery.
Small delivery time windows appeal to customers but cost money. The importance of the trade-off has been dramatically demonstrated by business failures such as Webvan (bankrupt in 2001) or Publix Direct (shut down in 2003). Other companies learned from these failures and successfully provide delivery services with larger delivery time windows, using appropriate scheduling and vehicle routing software and, at the same time, often have become very good at collecting valuable customer information that is used for customer segmentation and targeted marketing.
The sales process typically consists of collecting orders including delivery time requests (subject to available capacity) until a certain cut-off time, and subsequent planning of the delivery schedule using appropriate routing software. Relying on customers to request their own preferred delivery times has a potentially large impact on scheduling and routing efficiency: for example, the given delivery time specifications might require much longer routes than if demand were geographically clustered for each time window; and, some delivery time slots might be much more popular than others, so that a large vehicle fleet would be needed for peak times and leaving many vehicles idle in offpeak times.
A reservoir of great profit-generating potential is being missed. What needs to happen is for demand – in the form of online bookings – to be managed by setting incentives so that customers are steered towards selecting time slots that maximize the firm’s expected profits. Incentives could take a variety of forms: the size of the delivery charge for different time slots, and rewards such as discounts or ‘shopping points’ for choosing unpopular slots, or even by encouraging customers to consider the environmental impact, an approach already used by Ocado . The objective of maximizing profit is dependent on clear knowledge of the expected fulfilment costs, linking demand management to vehicle routing. The customers’ choices of delivery times directly impact on delivery costs so that steering their choices could cause significant profit increases.
In our research we have looked at how this would work in practice. We considered the problem of servicing requests, arriving at random points in time, for a fixed delivery day. We put together a schedule/routing plan for the delivery of the collected orders, based on the use of a number of delivery vehicles with a fixed capacity, and data which specified customer information (including post code), a revenue figure for the item/s and a figure for ‘capacity consumption’ corresponding to the number of standard-sized transport boxes required to accommodate the goods that this customer has ordered. The study is based on the creation of a more advanced customer choice model linked to a monetary incentive or disincentive, and the firm’s task is to set these incentives in real-time so as to maximize profits over the entire sales horizon leading up to a specified delivery day.
Through a detailed simulation study we have been able to show that by including the impact of future expected orders (a “foresight policy”) on the decision of what delivery fee to charge for each delivery time slot, we can obtain higher profits than only using orders accepted-to-date in this decision (a “hindsight policy”). Using real data from the online grocery sector we found that the foresight policy results in improved profits of on average 2.6% over the hindsight policy, and on average 3.8% as compared to the typical industry practice of using two-tier static delivery prices that only depend on whether or not the order value exceed a certain threshold. In an industry that operates on very small margins (for example, one etailer reported operating margins of 0.55% in H1 2012 according to Mintel), this profit potential is remarkable. The model shows that our approach can outperform the static two-tier delivery pricing policies that are often found in practice by up to 6% in profits. We also found that dynamic pricing without taking future expected demand into account can produce even worse results than static pricing when dealing with situations where delivery capacity is scarce.