Recently Mal Walker conducted a warehousing design workshop at Smart Conference 2011. The following is an extract from his presentation.
Preamble
This module covers: Picking Rates, Capital Investment in Materials Handling Systems.
It will assist you to reduce investment by selecting the right level of picking technology for your distribution centre.
Picking Velocity
The most prominent driver of warehouse design is velocity! Velocity is the rate at which goods flow through a warehouse. In the ensuing case study, we consider ‘picking velocity’ which is normally measured in terms of ‘bin hits’ per day. By definition, a bin hit is a visit by an order picker to a picking location, whereupon a product is picked in single or multiple units.
Our exercise covers July 2011 data from Hogwartz Enterprises, suppliers of sporting equipment to Quidditch Players. Data was downloaded from the Hogwartz ERP system and is graphed below.
From the chart, for what quantity of bin hits per day would you design an order picking system?
I sincerely hope you said ‘need more information’ or ‘let me check that data more closely’. However, if you already have an answer, read on to check your assumptions.
Order Patterns
Mr Ellerby, who is the Manufacturing Director at Hogwartz, advises that most of the orders are picked at the end of the month. What’s your view?
To assist, I have added a linear trend line using the mathematical method of least squares. This method was developed by Carl Friedrich Gauss in 1794. Don’t worry, his work is entombed in standard Excel spread sheets and there is no need to crunch numbers to use it. But the results bring clarity and vindicate Ellerby’s statement.
There is a definite trend leading to higher orders at the end of month. So, returning to my question, what level would you design the system for? Middle of the trend line, top of the trend line, or do you still need more information?
More Analysis Please
For those of you, who are not comatose as yet, great! Stay with me to delve a little deeper.
From the original data the following statistics were derived:
Minimum = lowest number of bin hits per day.
Maximum = highest number of bin hits per day. Mean = Average number of bin hits per day (from whole data set).
Range: Maximum bin hits minus minimum bin hits per day.
No of Days = Number of days in the data set. Standard Deviation = a statistical measure of dispersion about the mean which was derived from the data set.
Of the items in the table, the most important are the mean and standard deviation. Together they reveal an intriguing story.
But first some background. The standard deviation was developed by English statistician Francis Galton in the late 1860s and is one of the most useful analytical tools a logistician can employ. Mathematically, the standard deviation is the square root of the sum of variances for the data sample with variance being the difference from the mean and each data point. The standard deviation is used to describe the spread of occurrences on a normal distribution graph (bell curve).
The bell curve is a predictive statistical model which describes the impact of standard deviations from the mean and is based upon the Central Limit Theorem. This theorem postulates that for endless aspects of life, including warehousing, the aggregate of random events tend to follow the shape of the bell curve.
By now, nausea will be setting in, so why I am telling you this?
Simple! This predictive model will save you both money and angst. Read on.
Cost of System
Suppose you wish to obtain a quotation for your picking system. You ask the supplier to quote for a system to handle your maximum bin hits per day i.e. 4,970 bin hits.
Its price is $3m. With this system you can handle any surge in picking up to your maximum. You proudly tell Ellerby, who turns white, and tells you in colourful language to find something cheaper, or find another job.
With wilting self-esteem, you ask the supplier to quote for a system to handle your average bin hits i.e. 3,359. Its price is $1m. You advise Ellerby and his colour returns, but quizzes you about system performance. You do some mental arithmetic and tell him that for 50% of the time it will cope without adding extra labour. But for the other 50% you will need to add resources to complete your daily picking task. Ellerby is not happy.
What do you do?
You review your mean and standard deviation and calculate the 84.2% level of operation. How? The mean is 3,359 hits per day. To this you add your standard deviation of 1,105 giving a total of 4,464 hits per day. You figure from the bell curve that at the 4,464 hits per day design level, your system will meet 84.2% of your daily picking tasks. See below.
For the remaining 15.8% of times, i.e. 5 days per month, you will need to add extra resources. You decide that this is an easy addition and will cause no financial or operational stress.
You get on the phone and ask your supplier to quote for a system to meet the 84.2% performance level. The quote is for $1.4m. With this information you compare the marginal capital cost for each design level against the marginal labour cost benefit.
Suddenly you see things in a new dimension. Alas, you will not achieve a marginal payback from installing the maximum level system, for over 11 years. But with the system designed to the 84.2% performance level, you marginal payback is within 1 year. This will meet both the capital budget of Hogwartz, and achieve the substantial productivity boost you need.
You are so elated, that you eagerly update your graph with the maximum, average plus standard deviation and average design levels, along with the quoted capital costs.
You approach Ellerby with confidence, tell him you are going to save Hogwartz $1.6m and are satisfied that you will achieve your daily picking with eight people plus an additional two to cover peak periods. Ellerby is pleased, knowing that delivery of Quidditch accessories is in good stewardship.
So what does all this prove?
If you are confused, dazed and delirious from such heavy deciphering, please allow me to conclude by explaining the lessons learned.
Firstly, use of averages for design purposes is perilous. (This is the most common mistake warehouse designers can make and should be avoided)
Secondly, check your data using the mean and standard deviation in order to really understand what is happening.
Thirdly, beware of the high capital cost of designing for peaks, as opposed to designing for the majority of occurrences.
Fourthly, check the marginal cost and benefit for solution comparisons, before making your decision.
I trust this has helped you to be more astute in your warehouse analysis and design. However, should you have any rebuffs, questions, or complaints, please drop me a line.
Best regards,
Mal Walker
Mal is Manager Consulting with Logistics Bureau where he leads the Warehousing and Distribution Centre Design Practice. He works with local and international organizations and has over 30 years experience in warehouse design and performance. He is a Life member of the Logistics Association of Australia, Member of the Council of Supply Chain Management Professionals, and holds qualifications in Engineering, Logistics and Business Administration.
Special Free Download Offer: If you would like a free download of the Warehousing Chapter from our new book, Supply Chain Secrets, just visit our free chapter on warehousing. This chapter has lots more great tips and advice.