In order to properly design and size a logistics system, information is needed about the distribution of the load on the system over time. A large number of different individual analyses are carried out, e.g.
- the daily and hourly quantity of orders, order lines and pieces,
- the number of different (unique) products picked per day,
- the overlap of picked products per day (i.e. the stability of demand per product),
- the popularity of individual days of the week, and
- the profile of different calendar weeks,
to name just a selection.
Essential in the analysis of the time profiles is to avoid a common mistake, namely the (sole) use of average values. Almost always will using averages in time profiles lead to incorrect planning assumptions. This applies both to values such as the typical quantity of orders or order lines per day, the distribution of which over time must be taken into account (see, for example, the histogram below), and to the frequently encountered breaking down of daily values to hourly values on the assumption of a uniform distribution, which in most businesses is wrong.
The assumption that daily order volume can be evenly distributed over the daily operating hours is rarely ever true and almost only works for long order lead times. In many other cases, and especially in food retail, orders to be processed are highly unevenly distributed not only across different seasons, months and weeks, but also within a day. For example, many retailers offer same-day shipping if orders are placed by a defined cut-off time. As a rule, therefore, a peak in incoming orders can be seen shortly before this order cut-off time, which must be taken into account in the system design (see the following figure with peak in incoming orders shortly before 15:00). An evaluation of order quantities on an hourly basis is therefore usually the basis for dimensioning the logistics system.
Even within a working week, there are often highly different loads on the logistics system. Often, individual days stand out on which demand is particularly high, while load is significantly lower on other days. The bar chart below shows, for example, a particularly heavy order load on Mondays. It also shows that the data set is “contaminated” with individual left-over orders on weekends and that the annual profile is highly uneven.
The bar chart below on the popularity of days of the week highlights the uneven distribution of load on the warehouse. On the one hand, it shows that the order load drops over the week. On the other hand, it also shows how misleading the use of averages for system design is, since they are massively distorted by outliers. With regard to the date in the specific example, extraordinary effects must also be taken into account for the design of a future logistics system: In 2020, many retailers, especially online retailers, experienced a Covid-induced demand boom, which in many cases is not representative for subsequent years. Data sets from years with such special effects must always be treated with caution.
In addition to an aggregate representation of the popularity of days of the week, the week-by-week comparison can provide additional insight, again to rid oneself of potentially incorrect assumptions from the aggregate representation.
The polar diagram does show that Mondays are often particularly strong for new orders. However, it also shows a wide range of variation. In the present example, it would therefore by no means be the case that Mondays always turn out to be particularly strong. The level of variation suggests that the cause of the variation should be clarified.
All in all, the time profile analysis for the conception and dimensioning of a logistics system, whether manual or automatic, is not witchcraft. However, it goes far beyond the still frighteningly widespread formation of arithmetic averages over daily values of order lines and the like. There is always an interpretative component to time profile analysis as well; the analyst must understand the customer’s business and interpret the data set accordingly in order to analyze it correctly and come to the correct conclusions. This applies to all types of logistics data analysis, but especially to time profiles. We would be happy to support you!
