Some thoughts on David Simchi-Levi, Kris Timmermans, A Simpler Way to Modernize Your
Supply Chain, Harvard Business Review, September – October 2021,
p133–141
Metrics.
KPP: key performance predictors: “metrics that indicate what the state of the supply chain will be in the next {}” (in the case of consumer packaged goods example firm: three to six weeks; in the case of makers of fashion products: “selling seasons last no more than 10 or 11 weeks.”)
Model:
Apply either a strategy of efficiency or a strategy of responsiveness to your supply plan according to which segment your products belong to:
Part your products into low volume / high volume on a y axis.
Part into low demand volatility / high demand volatility on the x axis.
Divide the low demand, low low volatility products into 2 segments:
low-margin products.
high-margin products.
Products with high volumes, low volatily: source efficiently (to save cost)
Products with high volatility: source responsively (to assure high service levels)
Products with low volumnes, volatility:
prefer efficiency for low-margin products;
prefer responsiveness for high-margin products.
Operations.
Get rid of consensus forecasts and traditional S&OP (sales and operations planning). Get a data scientist. Get data from multiple sources (e.g. customer demand, shipments to retailers, macroeconomic data, events, holidays, competitors’ promotions) to create a “unified view of demand” and create detailed supply plans. Allow for the data to be described easily – by showing how they came into being as a fusion of weighted variables. Compare actual demand with the plans to follow up and find out how big and why there are differences.
Externality facts.
Variability in customer demand is lower than variability in retail orders.
(Similar to: workshop – purchase orders!)
Traditional S&OP takes up to one month to complete a plan for a 50 to 80 weeks horizon.
Semantics and curiosities.
I think you have to dissect these kinds of articles on supply chain management techniques and describe them on a high level in order not to get lost in different terminologies that, in the end, all describe the same (like an earlier adoption of the S&OP framework in Demand Driven S&OP by the DDMRP movement), without highlighting the actual change that may have taken place.
I suggest we recall the work of Agraval, Gans, Goldfarb – Prediction Machines. Thus when Simchi-Levi and Timmermans, in essence, just propose to increase order cycles (by performing a weekly, “smart execution” S&OP based on additional “KPP – key performance predictors”), their process just calls for a review of automatically generated data by “prediction machines”, to be okayed by humans, to cut short discussions about data, and to shorten the time horizon at which managers are looking at – thereby decreasing the stakes.
The difference is in readily available, trustworthy data and relative ease of making statements and arriving at conclusions about those data. Nowadays there seems to be less wriggle room in presenting and interpreting the data as opposed to 1980s ur-style S&OPs taking up to one month, where “[t]raditional approaches employ consensus forecasting, in which (...) all the functions get together and hash out a compromise uniform forecast”.
I assume: How organizations design dataframes and receive data has changed fundamentally. While in earlier times data sets may have been somewhat magically gathered and let’s say “chauvinistically” presented by managers, it seems managers may now oftentimes be coupled with (trained!) “data scientists” (p137) to design dataframes, and when the data are presented, with everone knowing that data scientists were involved in the design and gathering of data, they are also more likely to have veracity to them.
