Imagine a wildly complicated supply chain comprised of 400 vendors, six contract manufacturers, 31 distribution systems, 160 retail stores and thousands of resellers spread across 207 countries. While it might sound similar to the level of complexity most global fashion firms are dealing with today, it’s actually the structure of Microsoft’s consumer electronics business.
The tech firm is probably best known for software but it designs, produces and sells millions of game systems, tablets and augmented reality headsets each year. And to hear Darren Coil, director of category marketing for new products and servers at Microsoft, describe it, the process—and problems—associated with that business aren’t too far removed from the task of delivering on-trend dresses, scarves and jackets each season.
The biggest similarity between the industries, he said, is the ways in which painfully slow decision making can wreak havoc across the supply chain.
To attempt to solve this problem, many apparel retailers have recently opted to upend their corporate structures in an attempt to reorganize their way out of byzantine processes that have been holding them back from keeping pace with consumer tastes.
And while some have already seen results from tearing down silos and creating cross-functional teams, Microsoft has learned that gathering, identifying and analyzing “intelligent data” has had an even bigger impact on its business.
According to Coil, going digital has enabled the company to move faster, waste less, reduce inventory and ultimately become more profitable.
Though Microsoft is still in the midst of its three-part digitization initiative, it has already realized millions of dollars in benefits to the bottom line. It started, he said, with simply connecting the data that existed in individual databases throughout the organization. In fact, within six weeks of making the first connections, his team knew they were onto something.
Now as the company steps into the third part of the process, the gains are even bigger. In just one case, Microsoft was able to save $35 million a year by employing machine learning to solve a production problem with its Hololens augmented reality product. Prior to using machine learning, the yield rate for the three lenses used in the product was 50 percent at a cost of $50 million a year. Now, it’s up to 85 percent.
Here, Coil provides a compelling case for going digital and outlines a roadmap for companies that are just at the beginning of this process.
Sourcing Journal: What were the top challenges related to having so many suppliers and a complicated supply chain?
Darren Coil: The challenges are not unique to our world. They’re common across all manufactured goods type supply chains and that is there are so many different repositories of data that you need access to at different points of time to make critical decisions about your supply chain, your manufacturing, your quality, your customers, etc.
In our case, we have hundreds of databases that contain various critical pieces of information so it would take several days pulling data out of one system and relating it to data from a different system to make one decision. It’s first and foremost an exercise in just putting the data in one spot. In our world, much like in the retail world, oftentimes you’re making multibillion dollar decisions, and you have days, moments, hours to make them so you need the data in one spot to answer one simple question.
SJ: Speeding up decision making is a key focus for apparel retailers today. How has speeding up decision making helped Microsoft?
DC: When we enabled the team to make decisions in real time, we saw an impact on the order of several million dollars to our bottom line within months of going live. We see improvement in the amount of material we throw away because we were able to make better decisions. We also reduced the amount of inventory on hand so we were able to turn down our days of supply because we could be more comfortable with a more just-in-time approach. Also, like in fashion, after a certain period of time a style or technology is obsolete and you have to simply write it off. We notably reduced our obsolescence in our factories.
The cross-organizational pace also increases. In other words, the design engineers, the planners and the sourcing guys collaborate more because they can all see each other’s data so they can help make decisions faster. In the retail space, you could do the same thing. So, you could see a particular item is selling well in one region but not in another and you could quickly reallocate that inventory to the region that’s hot.
SJ: How did your team go about getting buy-in at the C-suite level to do this?
DC: This is a tricky one. In our case, this was an easy one because the corporate vice president decided he could no longer afford to make decisions in the way he had been. It was a top-down cultural shift to digital.
If someone isn’t at an executive level, and they’re trying to convince their execs to do this, I think it would be a tough sell. The executives will say, ‘I’m not sold and I see a lot of money to do all this work.’ And they’ll recognize that this will be a cultural shift in their operation and they may not be willing to do it. The technology behind all this is easy. The culture is the hardest part in this whole journey.
SJ: How has this transformed your relationships with your suppliers?
DC: In the old model, whenever there’s a problem the first thing that happens is Microsoft talks to a vendor and we start arguing. Microsoft has our data. They have their data. You spend all this time arguing about who’s data set is right. In our new model, we connect to the supplier, they supply the supply chain data—information about the product, when it shipped, the pallet that it’s on, how many, quality inspections, etc.—now we’re seeing the problem through their data so we don’t have to have an argument about the data. We can jump to our action plan.
SJ: How has it affected which suppliers you opt to work with?
DC: In general, we haven’t changed our sourcing decisions, however we have completed data that has shown certain vendors to have better performing product than others, so I know we’re paying attention to that and thinking about sourcing decisions based on what we see in the data.
SJ: You’ve said that Microsoft is undertaking its digital transformation in three steps, which you call connected, predictive and cognitive. Describe what these are and where you are in the process now.
DC: Connected is automating sources of data and putting it all in Azure [Microsoft’s cloud computing service]. We’re done with that but it’s going to be a never-ending journey because we’ll have new suppliers and data points.
Predictive is where we use statistics and data grading to elevate data that’s important and suppress data that’s not. We have that in a version 2.0 and we want to get to 3.0 before we’re done. We’re 80 percent there.
The third wave is where you use machine learning and artificial intelligence to make decisions for you. We’ve done several examples, and had it make real decisions for us and it’s had real impact to our bottom line. [For instance,] we had a yield problem that had us scrapping stuff that cost about $5 million and machine learning fixed that problem. That’s about 10 percent done. We’re in between the second and third wave.
SJ: What would your advice be about how to get started?
DC: What they want to think about is start small. Don’t try and topple massive amounts of connections all at once. Start with five or six things they do on a regular basis like looking at factory outputs, supply availability or some logistics thing—whatever they’re comfortable with and have really good data for and a really good way of getting the data even it comes from five or six sources. That way they can validate the data, how it’s being collected and reported.
Also, don’t let IT get in the way. I come from IT but it’s true. IT will have a tendency to say we have to do upgrades or we need to architect all this stuff. You don’t need to. You just have to look at the system from a left to right fashion: I’m going to connect here then there and it’s just going to keep going in sequence.
Start small with a small team and run fast. Once you’ve automated the first five or six reports, the business group that uses it will start driving the fly wheel harder. We went from five to 10 reports to 100 within 6 months from demand coming out of the supply chain. Once they saw how efficient it was and the data was true and they could believe it, we couldn’t build enough connections and reports for them fast enough.
And evangelize and publicize the victories. You need to say look at what we just did.
Finally, remember getting it done is more important than being perfect.