- Tags:: 📜Papers, Demand management
- Authors: Niels Adrianus Hugo Agatz
- Link:: RePub, Erasmus University Repository: Demand Management in E-Fulfillment (eur.nl)
- Zotero Link:: agatzDemandManagementEfulfillment2009
- Source date:: 2009
- Finished date:: 2021-09-01
General notes
Not sure about the name of the broader scope of the problem. Traditionally, Revenue Management is Demand management when marginal costs are negligible.
Demand management has greater potential in profitability than supply-side optimization:
Demand management has an equally strong impact on profitability, affecting both costs and revenues, and may hold a much greater potential for many of today’s Internet retailers than further supply-side optimization. (p. 1)
Since supply is usually relatively inflexible in the short term, dynamically managing the slots and corresponding fees as demand unfolds provides promising opportunities (p. 22)
Campbell and Savelsbergh (2005a) show that actively managing demand in Internet direct delivery can produce substantial cost savings (p. 2)
Especially in the case of low-value items, such as groceries, transportation costs are a key determinant of the business viability. (p. 34)
Ojo, con cuidado:
overly complex pricing policies may leave customers confused and distrustful (Garbarino and Lee, 2003) (p. 29)
Again, maintaining a certain level of transparency may be important for customer satisfaction. (p. 33)
In any case, it is going to be hard to model users behavior:
The time slot offering has an immediate impact on the attractiveness of the service for the customer, and potentially on sales. Customer behavior modeling is one of the most challenging aspects of demand management. The fact that customers order online, however, gives the modeler an advantage, because it facilitates monitoring and analyzing individualized customer behavior. (p. 139).
How much we can save from differentiating the number of time slots statically?
The use of service requirements to differentiate the number of time slots in different delivery regions results in significant cost savings as compared to simply offering all 2-hour time slots in all zip code regions. The results from our numerical experiments suggest up to 10% savings (p. 141)
The optimization of the assignment of the delivery time slots to zip codes yields only moderate additional savings over the manually constructed schedule (p. 141)
There are clear economies of scale in the delivery operation. The results show potential cost benefits from increasing the number of stops within the current delivery area. This illustrates the importance of growth in order for an e-tailer to become more profitable. (p. 141)
However, there are gains to be had by dynamic adjustments:
Taking some measure of the remaining effective available capacity (e.g. an order limit per time slot) into account when determining the time slot offering in real-time results in substantial improvements in profit. The results from our numerical experiments indicate gains of about 9%. (p. 142)
Taking into account the differences between customers in terms of delivery costs yields substantial improvement in profit over simply accepting demand first-come, first-serve. Our numerical results indicate potential improvements in profit of up to 75%. (p. 142)
Several e-grocers do some form of it: Tesco, Peapod, Ocado (p. 2). It is particularly amazing the idea of Ocado: highlighting the slots that would minimize fuel consumption because a delivery van would already be near the neighborhood.
Almost all e-grocers use demand management in some way (p. 22)
Chapter 2. E-fulfillment at Albert.nl: An Illustrative Case.
Chapter 2: “E-fulfillment at Albert.nl: An Illustrative Case” is particularly interesting, since it describes a high view of the whole operation of the e-grocer. Note that this is what they did at the time, but the whole thesis proposes demand management actions both in offline and online planning.
2.2.1 Offline planning.
Demand
Determining the specific time slots for each zip code involves a careful trade-off between marketing and operational considerations. (p. 19)
Assigning specific time slots to a zip code should not be done in isolation, but should be considered jointly for neighboring slots. This results in a complex planning problem. The time slot schedule employed at Albert.nl is created manually (p. 19)
Supply
Off-line supply planning tasks range from long-term (e.g. distribution network design) to mid-term (required staffing levels) (…)
The main physical capacity constraint at Albert.nl is the available room for the storage of orders between order-picking and transportation.
The company plans its required staffing levels (order-pickers and delivery couriers) six weeks in advance** based on sales forecasts, which are based on past sales developments, seasonalities, and sales promotions** (same as us).
2.2.2 Online planning
In supply, this is what they say about quick adjustments of personel:
well-trained delivery personnel. This essentially limits the options for using temporary employment and thereby short-term flexibility. At this point, supply and corresponding costs are essentially fixed.
3.3.3. Inventory and Capacity Management
One of the challenges in retailing concerns seasonal demand fluctuation. In a traditional retail store these fluctuations affect decisions on order quantities, shelf space allocation, markdown pricing and sales force levels. In e-fulfillment, demand fluctuations, with respect to the moment of delivery, also affect the utilization of the delivery capacity and therefore tend to have an even stronger impact than in traditional retailing. In addition to annual demand patterns, demand differences during the day (morning - evening) and during the week (mid-week - week-end) are particularly important in e-fulfillment. Staffing levels need to be adjusted to these demand fluctuations. This includes both delivery and order picking capacity. Since delivery requirements tend to be more variable and more interrelated across orders than picking requirements, capacity management of the delivery process is considered to be a greater challenge. (p. 41).
On dynamic pricing
Asdemir et al. (2009) propose a dynamic pricing model for the delivery windows of a grocery home delivery operation. (p. 30)
…discounts can be used for matching a delivery with a visit to a nearby customer, and for moving demand to temporarily underutilized delivery periods, thereby enhanching capacity utilization (p. 62)
With just a little:
The experience of Peapod indicated that even a small discount (e.g., $1) can change the customer’s slot selection (Campbell and Savelsbergh, 2005b) (p. 62)
May be seen as unfair, but people get used to everything.
… customers may perceive unexpected price changes as unfair (Kimes and Wirtz, 2003, Xia et al., 2004) (…). However, the fact that it seems normal that our seat neighbors on a flight pay a different ticket price than ourselves illustrates that acceptance of dynamic pricing may be a matter of habituation (Wirtz and Kimes, 2007).
But if people get used to everything…
If discounts follow a regular pattern customers will learn to anticipate them and thereby limit the directive effect of the pricing tool (Talluri and van Ryzin, 2004).
Note that they have a section on modelling the effect of changing the delivery slots assortment (5.2. Response to Static Time Slot Management)
Other notes
They were already using “Panal” in 2004:
In the hub-and-spoke system, a large truck transports up to 100 picked orders to a hub where they are transferred to regular delivery vans (p. 12)
We don’t need to think of physical stores:
optimizing web-channel prices without changing store prices often provides a reasonable heuristic for maximizing total profits. (p. 31)
Order picking costs account for the largest part of warehousing operating costs. (p. 39)
Further interesting research
simulation-based analysis of different delivery strategies for e-groceries (Punakivi and Saranen, 2001, Punakivi et al., 2001, Punakivi and Tanskanen, 2002, Yrj¨ol¨a, 2001) (p. 27)
Geunes et al. (2007) model the delivery pricing problem when both the size of demand and the demand frequency is price sensitive. They focus on the question of which customer regions to serve, at which price, in order to maximize profitability (p. 28)