I am recommending the purchase of COSI (as a small position). COSI has a very compelling reward:risk ratio. If my bull case is right the upside is multiple times the investment, but if the worst case scenario plays out, the downside is not well protected—and I assume a 50% loss of capital in that scenario. With that in mind, I believe relatively binary outcome investments like these deserve a place in one’s portfolio (sized appropriately small) IF the upside is massive and one can see a reasonable path coupled with reasonable odds of the bull case being achieved.

COSI operates in the fast, casual dining arena. The closest public concept and competitor is Panera Bread (PNRA). Both companies have very similar business models—with owned and franchised locations. Very good, fresh bread (baked multiple times a day) is heavily incorporated into the menu and marketed as a major selling point to customers of both concepts.

As of Q3’06, PNRA had a total of 976 (including owned and franchised) restaurants or about 8.4x COSI’s 116 locations. Partly a result of this noticeable difference in company size and scale, in Q3’06 PNRA’s corporate G&A expenses as a % of total revenues were 7.6% compared to COSI’s 17.2%--a 960 basis point difference! Note that in FY2000, PNRA’s G&A expenses were 10.8% of revenues—this was with PNRA having about $125.5mm in owned restaurant net sales, which is about the same as COSI’s TTM restaurant net sales of $123.1mm. The difference is that COSI’s G&A as a % of TTM total revenues is 17.6%. Part of this can be explained by COSI’s minimal franchisee royalties/fess of $0.5mm while PNRA had about $13.6mm in net franchisee revenues—but that still leaves at least a 500 basis point differential not explained by a lack of scale. Management claims that it has built up corporate overhead in advance of major expansion plans. If one believes that in conjunction with the fact that it costs more to be a public company today (Sarbanes Oxley, etc.) than it did it six years ago, that can help explain most of the difference.

COSI’s restaurants (~$1.23mm sales/owned restaurant) do not generate as much net revenues per store as PNRA’s (~$1.86mm sales/owned restaurant). However, the average COSI store is 2,950 sq. feet compared to an average PNRA store of 4,600 sq. feet. Thus sales/square-foot for both companies is about equal in the $405-$415 range. At first blush it would appear that PNRA gains some expense leverage by having this larger store format since its 4-wall margins are 150+ basis points better than COSI’s with comparable sales/sq. foot. However, there is more to it than that.

Of the 108 company owned and operated COSI restaurants, 80 are “old generation” meaning that these were opened before the new management team came aboard a few years ago and were opened pre-2003. More importantly, these restaurants have an inefficient format at the front of the restaurant where customers order (lost revenues) as well as the kitchen (higher expenses), which is apparently based on the ordering area. The inefficiency specifically stems from separate lines for ordering sandwiches and drinks such as coffee as well as paying. As a result lines build up faster during peak times, which frustrate customers waiting longer than they should, resulting in them going else where and thus lost revenues.

So guess what? The new management team realized these two flaws in the design of the lay-out of the restaurant and began opening “new generation” stores that now total 28. These restaurants have one main line, but 2-3 registers at the end of the line where the customer orders everything (sandwiches, drinks, dessert) and pays at the same time. The customer is given a holder that has a number and a “runner” brings the order to the customer’s table typically about 5-10 minutes after the order is placed. With this new prototype, COSI has been able to deliver increased speed of service and better labor productivity, while dramatically improving customer satisfaction. New generation stores have higher revenues and lower expenses, thus yielding better 4-wall margins (24% v. 18%) and returns on invested capital (34% v. 18%).

One must understand that these concept expansion stories take time to play out. Many fizzle out, but some do make it and when they make it, they make it big and the numbers are staggering. I simply believe that COSI is at a major inflection point. The first restaurant was opened in 1996. Six years later COSI came public at the end of 2002 at just over $7. The concept did not have legs for the aforementioned reasons and it took a new management team to not only identify the problems, but adjust the model so that it could be franchised. Today, over 4 years later, the stock sits at $5.09, almost 30% lower from its IPO price. Yet, the major difference today versus four years ago is that it now as a very viable concept of which franchisors want to be a part. In 2005, COSI opened its first franchised location and now has 8. In 2007, it is expected that franchisees will open 40-50 restaurants. Currently, there are commitments from 27 franchisors to open 345 restaurants. At the same time, COSI expects to open 18-21 owned and operated restaurants over the next 12 months.

By the end of 2010, it is expected that 2/3 of the store base will be franchised, up from less than 7% currently. I estimate COSI should go from posting EPS in 2007 of -$0.11 to $0.65 in 2011. I assume that COSI will have an average restaurant base in 2011 of just over 500 stores or about half of PNRA’s store base today. At that time, about half of the earnings stream will be from royalties earned from franchised operations while half will be from company owned stores. The franchised royalty stream would get a 20-30x P/E multiple given the low-risk nature of the earnings stream, the high margin/ROIC nature of that segment and the still massive growth potential to still double the store base. The owned segment should get a 12-18x P/E multiple given the riskier nature of that earnings stream, lower margin/ROIC nature of that segment, but with the potential to double the store base as well.

Thus about a 20x P/E multiple in this bull-case scenario would be likely and the stock would be some where in the mid teens or about 200% higher from today’s price level in 3-4 years. Assuming 50% downside in my worst-case scenario, this results in about a 4:1 reward:risk ratio. However, if COSI is successful, then it is likely that it won’t stop at 500 restaurants, but rather expand to the 1,000+ store level PNRA is at. This is when the EPS numbers get particularly big and as a result the upside is much greater than 200%, which changes the reward:risk ratio immensely to the upside.

The way I arrive at my worst-case scenario is that I look at what COSI would be worth based on its business in 2007 if it were to go private and/or be acquired by a larger competitor (such as PNRA). COSI had $21.7mm of G&A for the trailing 12 months. Of that total, 30% is tied to development growth for owned & franchisee stores and 21% is the cost of being a public company. So the way I look at COSI is that assuming it became a private company and did not grow the store base beyond what is committed for 2007, it would earn $0.20/share—using a 13x P/E multiple you get about half of where the stock is currently trading. The point is I do not believe the stock is a zero in a worst-case scenario even though the company does not generate any free cash flow currently nor will it in 2007.

All one has to do is look at the growth of PNRA. In 1999, PNRA had 81 owned stores, 102 franchised stores resulting in $171.4mm in sales and EPS of -$0.02. In 2006, PNRA is expected to earn EPS of $1.88 on sales of $828mm ending the year with about 1,000 stores of which 2/3 are franchised. PNRA was trading at $3.50-$4.00 in the beginning of 2000 and has since gone up about 15x in 7 years!

COSI had a comp decline in Q3'06 of -4.2%. It was a bad July and August and a better September, with business rebounding which coincides with some of the new items it placed on the menu. There was a decline in frequency (to 3-5x/month) among their heaviest users (normally visit 6+ times/month) while new customers increased by 3%. As a result they added a program, visit 3 times, and eat the 4th time free. Also, new restaurants don't come into the comp base until their 16th full month of operations—which will be 2007 for many of these new stores (~15) opened in the back half of 2005 and in first half of 2006. 

As far as insider selling goes: for the Chairman I was told that his selling is planned (25k shares each month) for diversification purposes. I was told that the VP sold half his stake last month because "he needed the money". 

Look at PNRA's insider selling in early 2001. The president sold a ton of stock costing himself $4mm in gains at today's prices—the stock tripled from his average sale price. He probably didn't believe the odds were 100% that PNRA would grow its sales by 5.5x in 7 years and take EPS from $0.52 to $2.31 over that same time period. It's not like he nor the COSI insiders have sold all their holdings, but they know as well as we do that there are risks to the growth and that it may not happen. This is why insider buying and selling are nice things to add to the mix of analysis, but I never rely too heavily on it because at the end of the day these people often diversify their holdings as much as we do—and rightly so. Many times these insiders don't know much more about their own business than we do.

If you’re looking for a speculative stock with a very compelling reward:risk profile, with a clear understanding as to how there’s a decent chance the upside scenario will happen, then a small position in COSI may be for you.

Catalyst