Oh my! The blog post, What About Them Apples, raised a lot of heated discussion on LinkedIn. Quite the dust-up. Why? While I am advocating rethinking supply chain planning, for some consultants the only path forward is the adoption of DDRMP. The recommendation is a broad-brush approach. No matter what I write on demand planning, the response is to blindly deploy DDMRP. Note: DDMRP stands for Demand-Driven Material Requirements Planning, which was introduced and is promoted by the Demand Driven Institute (DDI).
To address the topic, let me digress. There were two great blog posts on the topic of DDMRP on Linkedin last week. I would encourage the reader to study both:

• The Good the Bad and the Ugly of DDMRP from Stefan de Kok from ToolsGroup. If you don’t follow Stefan, you should. He is a great thinker on inventory planning and demand management. I like Stefan’s quote, “However, all the boasts of value it (DDMRP) brings are in comparison to the value of MRP. That is setting the bar so very, very low. It is like boasting to a college math major that you are better at math than a third grader. Similar and even greater value than DDMRP has been achieved for decades compared to MRP. “
• DDMRP – A Repackaging of Lean and JIT from Shaun Snapp . Shaun has the courage to closely examine SAP’s market claims. Shaun debates the launch of DDMRP in SAP. The good news is that he has the capability and experience to do this. (I admire Shaun. It is not easy to take a contrarian approach against the big marketing machine of SAP. They hate contrarian views.)

Can We Change the Discussion?
For all of the DDMRP believers, and technology sales people wanting me to endorse products and methodologies, can we change the discussion this morning? In the words of Stefan, can we have a discussion like a group of college math majors that realize most of today’s supply planning implementations are less than ideal? Today’s implementations cannot be our baseline. I get the fact that today’s forecasts are not good enough to drive replenishment, and that rules-based consumption to translate monthly demand to daily demand was a mistake. I like the use of deeper math found in E2open, OM Partners, SAP (now in IBP HANA from SmartOps), and ToolsGroup to translate market signals outside-in into a better replenishment signal. (I would argue that not all of these math engines are equal. I see the best success with E2open (previously Terra Technology) and ToolsGroup, but I do not have enough client references to judge OM Partners’ product.) I also get the fact DDMRP believes in the calculation of a replenishment buffer in the operational horizon using orders. They calculate the buffer based on what is close to a naive forecast based on incoming orders. (In my simple mind, I think of this as a forecast.)
For the record, I also like the fact that DDMRP calculates buffers across the enterprise. (I call this demand translation, and the market has been slow to see the need for this.) The recognition of this need is a great gift to the industry. I appreciate Carol and Chad’s work in this area. My issue with DDMRP is the rigor of the math, the misnomer of the brand, and the overhyped market promises. (My favorite DDMRP technology is Orchestr8. I like BT’s use of the technology to establish the buffers. In the words, of Brian Dooley, DDMRP needs to be applied where it fits. It needs to be selectively used versus a broad-brush approach.)
So with this said, I would like to have a discussion about some promising technologies I am seeing in my travels. For those who don’t read my blog regularly, there are a couple of truths about Lora Cecere:

• No one pays me for my ink. I write from my heart.
• I have a passion to help business leaders, especially early adopters.
• I like best-of-breed innovators, especially those with deep math and a passion for what they do.

Cool Technologies
Driving supply chain innovation is hard. This is why I have a soft spot in my heart for technology innovators. As I sit with my coffee, I am reviewing a response by a large system integrator to a client that is just bad. Really BAD. The large system integrator is blindly recommending the technology they know and not the full range of products. The client is looking for innovative solutions, and the consultant is hawking yesterday’s approaches. Buyer beware!
The promising technologies that I see in my travels are:

• Lokad. At the recent Supply Chain Insights Global Summit, Spairliners presented on their use of Lokad to deliver better forecasting using probabilistic forecasting techniques. The Lokad approach assumes that demand data is not a normal distribution. (Skewed demand data is the new normal.) The company uses a schema-on-read approach to assemble data based on patterns and apply probabilistic forecasting techniques to drive better outcomes. In my recent dinner conversation with the founder, he stated, “50% of our 100 clients use the product for new product launch, and less for trade promotion management. Trade promotion management forecasting is far and away more difficult than new product launch.”  The forecasting of long-tail, sporadic demand patterns, seasonality, trade promotions and new product launch requires deeper math. I like Lokad’s use of machine learning using open source database techniques (think about attribute-based machine modeling in a modified Apache Spark-like architecture.) To get there, the change management issues as outlined in Figure 1, are intense.

Figure 1. Change Management Issues of Probabilistic Forecasting

I am working with a client now that has run out of patience with SAP APO. I am recommending the client test the use of Lokad on top of a Point-of-Sale Data Repository from Retail Velocity. The client sits on ten years of point-of-sale data.  I cannot wait to see the results for new product launch and promotions.

• PINC. The highlight of my trip to CSCMP, which was largely disappointing, was meeting Matt Yearling of PINC. We discussed the use of machine learning with drones for continuous cycle counting in the warehouse. PINC was an early adopter of drone technologies for fleet and yard management and has now moved into the warehouse. While the promise of real-time inventory data from RFID still looms large as an ideal state for the industry, it has not been practical. The industry has not been able to overcome the issues with reading tags on liquids and metals, and the overall cost of the tags. The PINC approach uses drones in the warehouse to fly by racks and use machine learning to capture the number of items and condition of the items in the racks. Love the approach. I have asked Matt to come and share his technology approach in the digital showcase at the next Supply Chain Insights Global Summit, and I am looking for a case study to showcase the technology as well.
• ThinkIQ.  When I was visiting a food manufacturer recently, I found that it took ten days for the company to identify lots to recall. Lineage in food manufacturing–track-and-trace of lots and specifications–is problematic. If this is your problem, consider ThinkIQ. The company uses the Internet of Things (IOT), blockchain, open source database techniques, and machine learning to track patterns and trace products. I am writing a couple of case studies of clients using this technology. As a gal that cut her teeth in the food industry, this approach has a lot of appeal.

Conclusion.
So, if your boss asks you for three use cases of machine learning and open source technologies in the industry today connect them to Lokad, PINC and Think IQ.  I also like the work in cognitive computing at Aera, Enterra Solutions, and Transvoyant. These are all very different solutions but promising. I strongly believe the future of the supply chain lies at the intersection of outside-in processes, better math, cloud, open source, and machine learning.
Let me know your thoughts. I look forward to hearing from you! Do you have any new innovations that you would like to share?