Advertisement

Blog

Machine Learning Sets the Pace of Distribution

First, it is worth mentioning that over the last 40 years, the catalogue electronic componentsdistributor has enjoyed an impressive growth, from under a million dollars to USD 1,760,000,000 in annual global sales revenue, with staff growing from 14 employees to more than 3,400 employees globally, with double digit growth figures recorded annually. 

EETimes Europe: How is 2015 different?

Dave Doherty: In 2015, we've seen a relatively flat growth. But although it looks flat on the surface, we see continued growth. In fact, we've had good growth in Europe but it was tempered by the Euro to Dollars currency exchange. It is a fact that we'll report relatively flat revenues in 2015, but our fundamental business drivers remain stable.

EETE: How big a market represents the new wave of DIY makers, open-source projects and crowd-sourced startups for Digi-Key compared to engineers working within more established companies?

Doherty: From Digi-Key's roots some 43 years ago, there's been a continuous evolution from the early days when our customers were professional engineers by day and hobbyists or innovators by night. 
Often, they were the same people buying from a different perspective. With their professional hat, they may have access to more budget, while with their hobbyist, or maker, hat, they may be working on completely different one-off projects. The cost of innovation has decreased considerably over the years. Nowadays, through STEM initiatives, there are more science engineering projects at school, there is a renaissance of engineering interest. 

While growth used to be driven by discrete components, now it is often driven by board-level products. Take MEMS components which are fairly complex devices to design with, but more easily accessed through a functioning board level integrated module. There are numerous modular products offered by companies such as SparkFun or Adafruit, implementing specific one-off functionalities, making it easier for professionals and makers alike to start designing, hence broadening our customer base among innovators in general, not just professionals.

EETE: To what extend do distributors need to spoon-feed these new engineers with pre-built design kits and recommendations?

Doherty: Better than a spoon, a fire hose would be more appropriate. We can't get enough content online to satisfy them. No two makers are the same. Some are designing something unique for their own use, others want to incorporate a company. 
We put all the information material and tools to foster their innovation, with cost effective ways to do 3D drawing and CAD simulation. You get a headstart when you can get your hands on a board that does 80% of what you want, leaping off to innovate. By integrating our libraries within available EDA tools, we want to make it easy for our customers to access all the information they need from a single environment, regardless of their choice of platform.

EETE: Together with design kits, PCB CAD software and design environments, how important could become IP blocks or 3D printing CAD parts to support more prototyping and boost related component sales?

Doherty: Some years ago, there used to be only a datasheet attached to each part, now companies give complete reference designs with part libraries and in some aspect, there is a commoditization of hardware. 
Software blocks are becoming more important and more and more suppliers are looking at offering IP blocks where the hardware and software converge to deliver specific functions. The design chain is coming to fruition and pulling these elements together.

3D printing CAD is definitely on our radar, it is increasingly facilitated by our suppliers. We stock products from over 650 manufacturers and our aim is to normalize the information to aid engineers in their research. 
While we produce some of our own boards, we increase our offerings at a much faster rate by partnering with innovative and leading edge partners and use our site as a vehicle to reach engineers with these options to accelerate their time to market. We plan to address the 3D CAD component elements through the same strategy by supporting and not competing with elements of our customer and supplier base. 

As well as stocking parts, we are a curator of data. We get it from the manufacturers, digest it and restructure it so as to make it as easy to search as possible on our site. Our customers want 24/7 access, they want to be self-served and that's why we need to normalize the content across our websites.

EETE: It has become current practise among distributors to run design contests around specific themes, often to bring visibility to a new product line. What sort of feedback do you get on such activities?

Doherty: In a world that is becoming more open source, there is a growing desire for a segment of the engineering community to share their ideas. And putting new technical products into the hands of our customers for a design contest is akin to crowd-sourcing ideas. 
We provide an open-source environment where designers can share their solutions. It generates publicity, gives some product visibility as well as technical feedback to our suppliers, since they get to see what designers are trying to achieve with their parts.

EETE: Concerning consolidation in the semiconductor industry, is this a boon for distributors (more efficient chain-supplies and better volume deals) or is it a business killer (fewer distinct contracts, hence lower margins)? Do you see this driving further consolidation in the distribution industry?

Doherty: I think there are elements of both. We've seen consolidation across the distribution industry over the last 30 years. It has been necessary and it benefits companies with economies of scale and to offer lower cost products. One might see the beer industry as somewhat analogous to ours with the largest breweries continue to expand through acquisition. 
Is it stifling creativity? Quite the contrary, you see the emergence of micro-breweries representing and increase share of the overall market. Through chip foundries, the cost to innovate even at the device level has been lowered and should continue to foster nimble and responsive companies with new ideas and a higher tolerance for risk.

Much consolidation has already occurred in the Americas, Europe still has a small handful of viable regional distributors while in Asia, there is a longer tail of distributors.
We do business somewhat differently, we don't have a large brick and mortar infrastructure (all stocks are centralized in a large warehouse in Thief River Falls, Minnesota) and we don't need to wait for extra infrastructure to expand on our 81 websites or to do business in any local language. 

The web allows us to be very responsive and scalable. The impetus for mergers and acquisitions often comes from shareholders or from Wall Street, pressures that we are immune to by being a private entity.

EETE: Would you say Digi-Key could remain independent because it doesn't serve OEMs and EMS so much or precisely for this reason, could it become a complementary target from high volume distributors such as Avnet or Arrow? Alternatively, could you ever see synergies with competitors such as Farnell, or Mouser justifying a merger? 

Doherty: We believe our model offers unique advantages in both the breadth and availability of inventory. Some other entities have followed a different path focused on lower inventory availability, relying more on sophisticated logistics modelling. 

We serve a customer segment whose needs don't fit that model, they are looking for a very simple, perhaps even “old fashion” distribution tenet of having in stock inventory to meet their unforecasted needs. There are fairly significant underlying financials assumptions and requirements that cause those two models to differ fairly significantly. 

Thus to best serve engineers and purchasing, it might be the best case that separate companies exist to optimize their processes around the uniqueness of each model.
Competition is healthy for any industry and drives value to the customer. As a private company we appreciate never being at odds in serving the sometimes differing needs of shareholders and customers. 

EETE: Either for inventory management, or to offer better online design support (maybe based on cumulative CAD designs validated on the cloud), is machine learning gaining grounds in distribution? 

Doherty: In distribution, we see literally millions of transactions and predictive analysis is the future. Unlike tradition distribution most of our shipments are directly from stock with very little visibility through backlogged orders. 
This factor alone has caused us to use elaborate forecasting algorithms trying to anticipate future demand and trends to ensure high service levels even before we experience hard demand.
Another use of the analytics is to leverage algorithms based on shopping cart activity, search patterns or even devices used together in our own EDA tools or those from partners to the benefits of our engineering customer base. 

Our engineers want to know of solutions in the form of parts that were designed to interface with other parts, again, aimed at accelerating their time to market. Our goal is to do this in a way that is very protective of the confidential nature of any one engineer's IP but to look for more common, natural association between the large number of products we offer. For example, what lens, drivers and thermal dissipation products best support a specific high bright led.

It's not just Wall Street looking for math majors to develop these algorithms. Just like Billy Beane in the book “Money Ball”, we are all looking for ways to transform data into information to benefit our customers. 
The consumer industry is ahead of us in this area today. Netflix, Amazon and a host of others tailors recommendations today based on our patterns. These algorithms will only get better and offer increased potential value. 

Many distributors operate in a model in which they want the manufacturer to demonstrate velocity of a product before they stock it. We want customers to see us on the front end of their NPI cycle, offering the newest technologies and trying to create new demand for our suppliers allowing for the greatest level of innovation by our customers. 

Story originally posted on EE Times Europe.

0 comments on “Machine Learning Sets the Pace of Distribution

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.