Tyranny of Averages

Playing with one of my new toys, I rediscovered the "tyranny of averages". My wife's Toyota Sienna with AWD currently gets 16.4 mpg for the past few years (per a gauge on her dashboard), yet looking at the mileage in realtime with the scan gauge indicates much different information. It sucks gas at the lower speeds/higher rpm getting all that mass moving, but at constant speeds or going down hill gets very high mileage (+64 mpg or higher)

With one tank of gas for a set of measurements - I was able to get to 20.1 mpg. The experience is a bit like a video game. Managing the "low" mpg periods can dramatically increase the average by 25% (20/16). Not bad for a fairly simply device.

This leads to me to wonder where else are we assuming averages are true and missing a huge opportunity to improve business.

I once built a smart shelf for a retailer in Cambridge UK using RFID. As an aside to the project, I started tracking the inventory turns and how it related to day of week, stock levels, assortments (#of different products). Amazingly, you could radically improve turns (and hence) margins by just ensuring that the stock clerk fully stocked the shelf on Thursday afternoons. Fully stock meaning that they put all 3 versions of the product on the shelf and did it before the peak sales happened (Friday and Saturday). Simply monitoring that product on a Thursday afternoon would have more than paid for that stock clerk's salary by delivering additional margins.

Given sales results for this Christmas, I imagine many retailers (and CPG) companies will be searching their portfolios to avoid the "Tyranny of averages"

Great Case study on Consumer Goods Supply Chain

Wow -- nothing like winding down the year to get time to catch up on my reading and get time to post.

Here is a good article by my friend Amit describing the Consumer Goods Supplychain

http://punetech.com/supply-chain-management-in-consumer-goods-an-in-depth-look/

Downside of Visibility

I had the privilege of working at a couple of accounts outside the US this year. And while each was in a different industry, I noticed that same thing in their forecasts -- a declining downward trend.

In both cases, the statistical forecast was trending downwards, yet when I asked if team had spoken to the VP's about it, they said "no because sales team is not in agreement". Sales teams by nature tend to have a "positive bias" which is a good thing. However, running a company that way will lead you straight off the cliff. Never a good idea not to use all your data in executive decision making. Hopefully these companies will react soon enough.

Jason Busch over at Spend Matters, looks at the impact of having visibility in your supply chain.

Now that we are using all the data -- the next question is how to react to ensure survival AND prosperity.....

Forecasting --Continuous Improvement

Forecasting is key for companies who want to improve customer service AND margins, yet many companies don't spend the necessary time to do the analysis as it can seem overwhelming.

In even a 100 Million business there can be thousands of "intersections" (customer - product -location) that need improvement, where do you begin.

A simple way to start is to measure "MAD" --- "mean absolute deviation" (Forecast - Sales*)/Sales for every intersection. Most enterprise demand management tools allow you to do this -- in i2's Demand Manager users can set this up quite easily and run it every month.

You should get something that looks like after 3 months
Product Location M1 M2 M3
"Bat wings" "Portland" 50000 45000 30000

Assuming you sell Batwings in Portland and missed the forecast by 50000 in M1 and 30000 in M3.

Once you have this record, simply focus your efforts (or your teams) on the largest 3-4 intersections each month and do everything possible to improve your forecast accuracy

1. Clean up your data -
you have tagged all your promotions in the past 3 years and identified the lift?
if using shipments, make adjustments for stockouts in the history

2. Tune your algorithms and make sure the parameters "cover" your data set . (A topic for another post)

3. If part of a S&OP (Sales and Operations planning process), make sure to capture and listen to the sales folks on the promotions in those intersections for the next few months as well as any marketing promotions.

In fairly short order - 2 to 4 months you should see some dramatic improvements in these top intersections and you can start the process on the next 3 or 4 "top" intersections

This small amount of effort will dramatically improve your business results, make supply chain and sales life easier and even make your CFO happy if you explain to her the impact on the Cash Conversion Cycle.






*I'm assuming you have access to actual sales, you can also substitute shipments.

Forecasting vs Lean vs Theory of Constraints in Supply Chain Management

I've been thinking about different philosophies for managing supply chains. If you do a search on the web, you will see different "schools of thought" including

-- Planning
-- Lean
-- TOC (Goldratt's Work )

And many people seem to believe that each "school" has the best approach. I'm not so sure - I believe that there are a critical few pieces of each philosophy that can be combined to improve the supply chain.

Forecasting Principles does a great job laying out the options for forecasting in general and for demand forecasting in particular. From my experience, forecasting can make a very large difference in a clients business, particular when combined with an S&OP process.

Lean and to a lesser extent TOC debate the need for forecasting - some lean components declare it is not necessary if you remove all sources of variation.

Unfortunately, in the CPG industry and for folks in working in supply chain or manufacturing the sources of variation are outside of their control
--- End customer desires for holiday shopping for example
--- promotional behavior of retail customers (holiday promotions)
--- and their own marketing departments desires to push more into the channel

In all 3 of these cases, I believe a better solution is to take key elements of each philosophy and work the problem to the "pareto-optimal" solution.

First, segment your business into two different areas --
1. from your DC's back into your suppliers business
2. From your DC's into your (retail) customers business

In segment 1
Implement a robust analytic process that includes both Expert Input (sales teams, etc) and Statistical Analysis. Include customers in the scope of your forecasting but aggregate them to the DC rather than to individual customers -- forecasting is always better when you allow the statistics to work for you than against (average out random fluctuations rather than try to predict them)

In segment 2 -- Implement a pull replenishment process using POS data if possible (e.g., use Lean Principles). P&G appears to be getting the results doing so according to the FT. But I'm willing to bet they forecast as well -- suppling retailers is a tricky business and the sell one pull one won't work if it suddenly jumps to 500....

Combining them together with a robust Sales and Operations Planning (S&OP) process will allow improve the organization and see where the constraints are that are blocking you from growth. The key to the process is having a common language and tool set where differing groups (Demand Planning, Supply Planning, Marketing, Finance and Sales) can work together on a problem rather than arguing about the data from a spreadsheet.

Once that happens, Goldratt's work in Theory of Constraints provides a useful framework for thinking about the problem -- what is the constraint in growing revenues by 2 or 6x? What is your "mafia" offer? And what is stopping you from achieving it? Many times a management teams perception is that the constraint is "the market" but in reality, they have many internal constraints that are blocking them from success. Besides, Eli would suggest to them that even the market is not a legitimate excuse - but rather how you serve the market.

Climate Change

Lots of comments from Plaxo and friends. Many of whom are far more certain then I am or so they think. In fact, I don't think I took a point of view other than to pick out an interesting video.

Here is another video that takes a different slant. From someone who is a Environmental Scientist, and is sure climate change is happening and knows how to solve it relatively cheaply. However he raises a very interesting question -- imagine you had a black box, that you could dial back carbon levels to anytime -say 1776 or 1865 or 1929 or whatever. Would you and more importantly could you? The important "Cui bono" question is now placed front and center. Who gets to decide - US, UN, EU, China, India? He (and I ) suspect each would pick different answers. Thought provoking....

Forecasting applied to Climate Change

Climate Change tends to get a lot of press lately and we tend to notice the weird weather events; however, just because there is more press and we've noticed more weird weather doesn't say much about our ability to predict temperature 5, 10 or 50 years in the future.

This is similar to me buying a new blue Honda a few years ago -- suddenly I noticed blue honda's everywhere. Did that mean that because I purchased a blue honda that everyone now had to buy one? Unfortunately not, just an example of confirmation bias in my thinking.

As Dr. Scott Armstrong (Prof. of Wharton) points out


"there are many scientists forecasting global climate change but he can't find any scientific forecasts of global climate change." Anyone have a scientific forecast of global climate change?

Dr. Armstrong is the owner of the excellent website http://www.forecastingprincipals.com and his new blog The Climate Bet . I'd also recommend his Forecasting Principals book.

Phase Transition -- time to rethink that business model

In physics there is a concept of a phase transition -- from the Wikipedia definition: "The distinguishing characteristic of a phase transition is an abrupt change in one or more physical properties. "

In the late 80's and 90's one of the dominate drivers of the global economy and geo-political events like the fall of the Soviet Union was falling and low oil prices. Oil price was underlying many decisions that were made in that time frame -- "extend my supply chain to China so I can save a bit of labor costs, no problem I'll build a plant and get enough volume to make good profits." Over the last two decades this has served companies pretty well.

Now the world is transitioning to another state -- Consumers are no longer satisfied with high volume of similar products that change slowly. The country is undergoing some radical "ag-flation" is driving up the cost of food. Long supply chains built to serve those old demands are becoming less viable.

I think companies will have to rethink their business models rather quickly. Take a look at the article and let me know what you think.

Trapped in JFK

Happy New Year - it has been a while.

Currently I'm sitting in JFK waiting for the flight to Boston using Jet Blue's Free WiFi.

Looking at some interesting links, eating sushi and feeling very tired.

I just finished a great book by David Maister -- this guy is brilliant and anyone in Professional Services or who deals with clients needs to at least follow his blog.

http://davidmaister.com/books.strategyFatSmoker/

More later