# A Small Matter of Trimming

Prepared by Grant Russell and Duane Kennedy

Tracy looked at the graph that was slowly winding its way out of the plotter on her credenza. With a sigh, she examined the plot of the observations and the line that was carefully placed through them. It was clear that this would require more drastic surgery. The slope of the regression line was just too darn obstinate. Well, she thought, we’ll just have to trim off more Moose Factories........

The project had come in a month ago, and the two partners of Dentrac Consultants, Inc. were very excited about it. Tracy and her university "bud", Denise, had decided to go it alone after graduating from University. While both had interesting co-op placements in major accounting firms during their university years, the idea of working for a large organization had become increasingly less attractive. When Denise broached the idea, Tracy had responded quickly and the new consulting firm of Dentrac Consultants, Inc. had been born.

Tracy had graduated with a significant number of mathematics and statistics courses completed, and had worked as a statistical analyst for a major oil company on one of her workterms. Denise, on the other hand, had taken a significant number of organizational behaviour courses, along with communications and policy. They both laughed that Denise was the "country club" partner and Tracy was the "drone". Denise was responsible for bringing in the clients and, to date, had only modest success. Tracy, on the other hand, was responsible for doing any quantitative analysis that needed to be done. While there were enough billings to live on, it didn’t seem like there would be a BMW for either of them in the near future.

The project was a bit of a surprise. It was a subcontract, with the prime consultant being a large, Toronto-based consulting operation. The contact had been another school friend who had recently joined that firm. Denise had done all the upfront negotiating and, when she was finished, the firm of Dentrac Consultants had a $15,000 consulting fee for approximately one week’s worth of analysis. Denise figured that Tracy could probably finish it off in a couple of days. The BMW dream started to look more practical. Denise had described the project this way. Raditech, the Toronto firm, was undertaking a major cost management project on Beer Unlimited, the joint marketing arm of the Ontario breweries. As a government licensed monopoly, Beer Unlimited operated all the Beer Stores across the province under some restrictive legislation. Beer Unlimited had to operate on a breakeven basis, with all breweries being allowed to market their product through the 350 stores located throughout Ontario. In the 1960's and 70's, the brewers had been only a few national operations. However, in the 1980's and 90's, the market had dramatically changed. Many microbreweries sprang up producing a multitude of different brands, packaged in different quantities, and often with different sized and shaped containers. These microbrewers tended to supply beer stores in their own local areas and in the major cities but, given that their distribution system often consisted of the owner delivering cases in his own station wagon, their range of distribution was not very large. As a result, the beer stores, designed to stock only a few brands with standard sizes and dealing with large volumes, were required now to stock many brands in many different sizes of packaging. Since many of the brands were not well known, many cases of product became stale dated, and would have to be held by the store until the manufacturer returned to take them away. Costs had risen dramatically as a result, and the consulting firm had been called in to find a solution to the problem. Denise made it clear that this was not an insignificant problem. "The guys at Raditech tell me that they can improve the handling situation pretty easily, by installing different stocking techniques, and by installing appropriate bar coding for the inventory and sales record keeping. However, the big issue is that the costs will still be high and the big brewers are upset at having to pay all these costs for the small micro-brewers who are competing with them. At the present time, there is no way to change the legislation to restrict access to the retail stores, wherever and whenever a brewer wants. Can you imagine how that would go over at election time if the party in power tried it? You can do a lot of things, but you can’t mess with John Q Public’s beer. However, Beer Unlimited has the right to charge back appropriate costs for stocking the beer. At the present time, this is based solely on volume, so that the big breweries take the hit. What they want to do is to institute a stocking fee for each brand, so high that the micro-brewers will be priced out of the majority of stores, and be restricted to only a few stores where the brand is well known and their volume of sales is large enough to cover the stocking fee. They’ve already determined that the fee should be$700 per brand. That should be good enough to wipe out the 2 cases per week sales of some of the micros."

Tracy broke in. "Well, sounds like they’ve got this all wrapped up.....where do we come in?"

Denise smiled. "They can’t just institute a fee like this without justification. What they need is a sophisticated analysis of costs relating to the number of brands demonstrating that the cost of each new brand is $700 or greater. Then, they’ll be protected when the microbrewers run to their MPP for help. However, the real plan is to bamboozle the microbrewers with an unassailable argument based upon as much data as possible. They figure that if it’s done right, the micros will simply fade away." Denise continued. "All we have to do is run the data that they have provided on the 350 stores. The data is total annual cost of each store and the number of brands sold at each store. Keep in mind that the number of brands sold at each store is different depending upon the local microbreweries. The regression equation that you derive (shown in Appendix A) needs to have a low intercept and high slope coefficient—preferably greater than$700."

The data had been conveniently provided on a disk, and Tracy had begun to analyze the data, using a SAS program. First of all, she had plotted the data, and had observed that while the majority of the observations seemed well behaved, there had been some that clearly differed from the rest. She could identify four such groups, which she described to Denise in a memo. Denise, with help from Raditech, quickly explained why these observations were different. The resulting information is shown in Appendix B.

Next she ran the regression equation shown in Appendix A, The results had not been promising. The regression had an intercept of $200,000 which reflected the fixed costs of operations, and a slope of only$300. When Denise had seen the print out, she was clearly upset.

"This will just not do, Tracy, this will just not do. I told those guys that we could do this, that we could get the numbers to work properly for them. If we can’t deliver for them, we’ll never get another contract."

The resulting silence was broken by Denise again. "You were able to do miracles on this type of stuff in school, I mean, can’t you fix the data?"

Tracy spoke slowly. "Well, the data does have some ‘irregularities’ in it. I could do some ‘data-trimming’ on it and see what effect that has. But if we do this, we’ll have to supply both equations and provide a rationale for the second equation."

"Well, see what you can do. Our whole future is riding on this one!"

"Oh, and Tracy, it’s probably not a good idea to supply two different equations or any kind of crappy statistics rationale. The guys at Raditech said they wanted it simple and powerful. The senior staff at Beer Unlimited are expecting our results, already. I told them that, based upon my own observation of the data, that the slope would be at least \$700. Besides, they probably can’t figure it out anyway. The only ones who would challenge this are the microbrewers and, with luck, they’ll never figure it out."

Tracy returned to the data. After carefully examining it, she decided that she would eliminate the stores at Moose Factory, Timmons, Sioux Lookout, Dryden, Grand Bend, Kingston, and Collingwood as being not representative. She reran the regression again.

If you were Tracy, what would you do?

## Appendix A

Plot of Outliers

Regression Equation

Y= A + BX + E

where Y = Total Operating Cost

A = Fixed Cost of Operations

B = Cost per Brand

X = Number of Brands

E = Error Term

## Appendix B

Observation Characteristics Representative Stores Rationale
Low number of brands, high costs of operations Moose Factory, Timmons, Sioux Lookout, Dryden These stores are all in remote, northerly locations. As a result, they have fewer brands since transportation costs (borne by the brewer) stops delivery of all but the national brands. At the same time, costs are also high because of the remote location, the high staffing costs, and the energy costs.
Large number of brands, high costs of operations Toronto, Hamilton, Ottawa, London, Windsor, Thunder Bay These stores are all in major metropolitan centres. As a result, they have many brands (many microbrewers are located there or can obtain enough sales to make it worthwhile to sell there). At the same time, costs are high because of the big city lease costs and the high staffing costs.
Large number of brands, low cost of operations Grand Bend, Kingston, Collingwood, These stores are all in major tourist areas. As a result, they have many brands (since these are prime locations to introduce new products to potential customers). At the same time, costs are low because the site is seasonal, and the rental and other operational costs are extremely low during the off season.
Small number of brands, low cost of operations Peterborough, Listowel, Smith Falls, Welland These stores are all in small towns and villages. As a result, they have few brands (unless a microbrewer is located right in town, the stores handle only national brands). At the same time, costs are low because of the small town setting.