THANK YOU FOR SUBSCRIBING
As long as I can remember working in Supply Chain, BI tools have been my best friends. Without them, impossible to perform specific analysis on new issues, add a KPI, special report to the Directors etc. Actually, in all the industrial companies I have been working for, we were dependent on BI to run daily operations. Why is that ? I have 2 examples in mind.
1/ For one warehouse in charge of preparing thousands of orders per day, we were given precise instructions by the VP of merchandising to manage product shortages (particularly during product launches), taking into account country, margins, store size, store ownership… There was absolutely no way we could execute that in the WMS without making mistakes. So, with the sharp minds in the team we had develop an algorithm to extract orders lines from the ERP where they were placed, download them in PowerQuery, apply the priority rules we were given before re-injecting them in the WMS. Dangerous ? Absolutely. We could have made huge mistakes. However, we were able with this method to execute business priorities in serving our customers and own stores.
2/ In charge of supplying raw materials for several factories, I didn’t want to use standard MRP functions in SAP, generating too many alerts and changes. I decided to use automatic replenishment based on thresholds. And guess what, to set up the thresholds I built a alrogithm in Excel (based on past variability of consumption, delivery lead times, MOQ etc..). And here again, extracting data from SAP into Excel, running the model, tweaking it to adjust the targets, and re-importing it into SAP became a bi-monthly routine. With this I reduced inventory from 9 to 2,5 months of coverage and never went out of stock.
I can remember dozens of situations like this, when the Supply Chain teams had to be smart, and under the radar of IT and finance used BI or desktop tools to significantly improved Service or inventory levels. It even started in 1993, when I started my career at Renault. Renault was clever enough to have developped a BI environment where production data could be accessed by the SC teams to build reports and KPIs. I realized then that learning SQL could help me tremedously in my work. With the years I learned how to use MS Access, then Pivot tables, then PowerQueries, etc and then I was not clever enough and delegated these tasks to our precious BI teams, working with more and more advanced tools.
Now start the questions :
Why can’t we have these queries and optimization possibilities in our APS ? APS are done to optimize supply chains, right ?
To whom should the BI team report to? To IT or in Supply Chain ?
Does the rise of Data Science change what I have experienced for more than 25 years ? If yes, in which direction, and who should actually master Data Science techniques ?
Here are my tentative answers to these 3 questions
1/ What about these fancies software ? I believe any one who already went through an APS implementation process has seen their limitations. What are those ?
Speed. 12 to 18 months is a minimum for implementing an APS. During this time, chances are that you business has changed. The result you get is not covering your new needs… not even sure they were actually overing your initial needs.
Resources. During these 12 to 18 months and of course after implementation is finished, you have to mobilize your most experienced and brightest people. Their tasks : understand the software, describe the actual process, constraints… in a way it can be modeled by the software, extract and clean data, tests, retests, train, understand the limits of the software. In other words, you need to mobilize your teams on new problems… before starting solving the problems you had before. Instead of working on optimizing your Supply Chain, they optimize the model, the software.
Costs. For implementing an APS you need to spend money on the the licences and support / maintenance for the software, on the consultants helping you to overcome the complexity of the project, on the IS/IT integration, on training… Don’t forget 3 other things : opportunity costs (what you could have done with the same money, on BI solutions for instance…), the time and efforts to actually get the funding from your CFO and general management… and the time you spend proving that you actually do better with the new tool.
Yes, extrem difficulties to prove that you are actually better off with the systems than without it… especially if you consider you could have spent all this money differently.
Fixed solution : APS vendors will tell you their solution is super flexible and can perfectly describe your Supply Chain and help you optimize it. From my experience, all Supply Chains are unique. What makes your company’s offer unique is also what makes it’s Supply Chain unique. Integration of the value chain, diversity of suppliers, markets, customers, constraints, opportunities… make a complex system which is absolutely unique.
Data : most APS are quite rigid in the way they require data. Don’t forget that your Supply Chain is visible only through your data. How many errors and approximations are you going to make trying to fit your data into the data model required by the APS ? Only part of the data you provide is exactly what the APS needs. As a result, the model of your SC (digital twin as it’s called now) can be uncomplete and fairly wrong.
Flexibility : Once you get your APS implemented, don’t you think you’ll need to update it ? Tweak it ? Adjust it ? How do you run continuous improvement through a fixed IT solution ? Yes, exactly, you use BI solutions on top of it, because you are probably sick and tired already to go through the change request process to close the gap between the tool you have and the tool you need.
2/ Where should your BI team sit ?
The first question is already : do you need an BI team at all ? Or do you need « BI trained » Supply Chain experts ? I have seen both work. The critical success factor is that these people now your Supply Chain. And as I explained it earlier, a large part of your supply chain is actually the data coming out of your ERP, ordering systems, WMS, TMS, APS… It means that the team / the experts need to perfectly know how and when each piece of data is generated. What does this date mean ? This quantity ? A IT expertise is definitely needed in building the back-end of the data cube, lake… Data must be made quickly available in a semi-structured form allowing fast queries. At the time you build this cube, you don’t know which queries will be developped later on… because optimization needs will come only later on. There is actually no limit as you don’t know which ideas you’ll have in 6 months, which request you’ll get in 12 months from top management to improve your procurement, production or sales. That’s the all point. There is only a short list of predefined queries that you can run every day for the next 3 years. Most queries get outdated super fast as your business changes. I’m sure Consultants can come up with 5 maturity stages on BI solutions and teams set up (and you’ll probably sit between maturity stages 1 and 2 so that they call bill you hours to help you reach the next stages !). I’m also sure you’ll need to fight with IT to keep BI competencies in your SC teams, or in direct control of your team.
And by the way, how do BI solutions compare to APS ? In my view, the do better on all aspects : costs, speed of implementation, adaptation to the uniqueness of your SC, focus on actuall issues and opportunities vs creating new issues, and most importantly they support continuous improvement, as long as they are run in an « agile » way.
3/ How does Data Science change the picture ?
My view is that progress Data Science makes APS even less relevant.
Data Science brings two benefits to SC professionnals on top of the current BI strengths.
a/ Possibility to manipulate even larger unstructured data sets
b/ Sophisticated statistics. I will develop this benefit, that most APS actually don’t bring. Mastering a Supply Chain consisst in reducing, anticipating and absorbing variability. The way we have done it for the last 25 years has been quite « empirical ». I will not develop further the limits of those approaches (one is for instance the used and abused assumptions that most variability is « normally » distributed, which, we all know, is quite wrong) but will insist on how modern « data science » actually fits to SC needs in the sense that Data Science tools are mastering analysis of variability.