Editor's note: Mark Travers is an insights specialist at Burke, Inc., a Cincinnati research firm. The author wishes to thank Megan Nicollerat for helpful comments on an earlier draft of this article.
It’s no secret that offering the right mix of products to consumers can make or break a business. Whether it’s PepsiCo trying to decide which soda flavors to put in vending machines, Apple grappling with which laptop models to put on its display tables or U.S. Bank determining which credit card plans to offer customers, product assortment is a critical component of business success.
But how should businesses go about solving this critical question?
Although getting to the right answer may seem like a daunting task, it shouldn’t be. The goal is straightforward enough: select the product assortment that maximizes profit. And the variables to get us there are straightforward, too. We need to decide which products, and how many, to offer our customers.
To put this in more concrete terms, imagine you’re an insights manager at a national juice manufacturer and you’ve been asked to review the company’s product assortment. The company’s goal is simple: put together the best line of juices that maximizes gross profit. Gross profit, here, is defined as total sales minus the cost of goods sold.
The good news is that there are really only two moving parts to this equation. First, you need to determine which flavors your company should put on shelves. Naturally, you start with the most profitable flavor and move down the line from there. A second and related concern is how many flavors the juice company should offer before capping the assortment. In other words, you need to find the “cut point” – the point at which offering additional flavors is no longer profitable for the business.
Now that you’ve clearly defined the problem (i.e., maximizing gross profit) and the variables at play (i.e., which products, and how many, to include in the product assortment), how do you solve this problem? Simple. You experiment. As Amazon’s founder Jeff Bezos famously said, “If you double the number of experiments you do per year, you’re going to double your inventiveness.”
In this case, you’ll want to find a representative sample of juice consumers and ask them to evaluate, in one form or another, the different juice flavors the company is considering for its juice lineup.
Before going merrily down the path of experimentation, however, there are a few important points to keep in mind. First and foremost, make sure your sample is representative of the population you are trying to understand. My philosophy here is simple: always get the best sample money can buy (within budget, of course). The business decisions that will be guided by this research are too important to roll the dice on sketchy convenience sampling. If budget is a limiter to your research program, I would advocate scaling back your analysis dollars before cutting cost on sample.
Second, stay true to foundational research principles. For instance, you always want to randomize the evaluation of products to guard against order effects and other confounding variables.
In terms of the experimental design to leverage, there are many that would be appropriate. One of the more common designs for this type of research is a product-by-product purchase intent evaluation. This is where consumers are asked to evaluate how likely they would be to purchase each product, one at a time. Another design often used in these situations is a choice exercise. Typically, a choice exercise would ask consumers to select their preferred product from a set of products, and would ask them to repeat this process a number of times.
As mentioned, both of these approaches would work (and debates over which is superior are largely academic in nature). Ultimately, what you want to get from your experiment are quantified measures of product appeal. Once you are able to quantify, you can compare products relative to one another to determine which sets of products are most appealing to which consumers and, in turn, most profitable for the company.
Now that you’ve designed a fundamentally sound experiment, go and collect the data needed to optimize your juice portfolio. Before doing so, however, there are three additional factors to account for in your portfolio optimization research: variety-seeking behavior, product profit margins and subgroup preferences. These are the components that separate good-enough research from excellent research – and can easily mean the difference between stealing market share from key competitors or simply treading water. Considerations and best practices on how to handle these important factors are addressed below.
Variety-seeking behavior
At a high level, there are two ways to think about the task of optimizing a product assortment. On one hand, you want to offer products that appeal to a wide swath of potential consumers. Returning to our juice example, this might mean offering orange juice to appeal to adults seeking a morning drink to pair with cold cereal, apple juice to appeal to moms with young children and, finally, lemon iced tea to appeal to adults looking for a tea beverage. As you can see, this juice lineup appeals to a wide range of non-overlapping consumer groups. This is beneficial as it increases your brand’s reach – the raw number of consumers interested in buying one of your products.
On the other hand, you want to offer flavors that stimulate repeat purchasing. For example, maybe you still offer orange juice and apple juice, but instead of offering lemon iced tea, you include grape juice in the product lineup. Sure, you lose the tea-seeking consumer group but you gain purchases from moms who cycle back and forth between buying apple and grape juices for their children.
Ultimately, the answer to whether you should offer the lemon iced tea (bringing in the tea-seeking consumer group) or the grape juice (fortifying offerings for moms with young children) is mathematical in nature. Recall that the objective, as always, is to offer the product lineup that maximizes gross profit. If the lemon iced tea sells at a higher velocity than the grape juice, you’re likely better off offering the iced tea; if not, it’s probably best to go with the grape juice.
As a general rule, it is the degree of variety-seeking behavior and repeat purchasing in the category that tips the scale in favor of brand reach (appealing to a wider net of potential consumers) versus product duplication (offering multiple products that appeal to the same consumers). For example, imagine you’re in the business of selling lawnmowers. Because consumers buy lawnmowers so infrequently, it’s probably best to offer a product lineup that maximizes brand reach. Offering multiple lawnmowers that appeal to the same consumer isn’t all that important as this tends to be a one-and-done type of purchase. Rather, offering a suite of lawnmowers that includes something for everyone would be better.
However, decisions on whether to maximize brand reach versus product duplication aren’t as cut and dried in categories where there is a high degree of variety-seeking behavior and repeat purchasing – as is the case in the consumer packaged juice market. Fortunately, you can identify the correct answer by adding a few additional questions to your survey that measure variety-seeking and repeat purchase behavior (see “Using TURF to find something for everyone” in the June 2017 Quirk’s for details on a technique specifically designed for this type of analysis).
From there, you have the ammunition needed to optimize your product portfolio to match the purchasing behavior that underpins your category.
Product profit margins
One of the main responsibilities of the insights manager is to keep a finger on the pulse of all issues related to consumer demand – and to share important findings and learnings with colleagues in other functional areas of the company. In the case of our fictitious juice manufacturer, this might mean providing other business units with a detailed understanding of consumer preference for different juice flavors.
There’s no reason, however, the insights manager can’t offer a deeper layer business intelligence.
In the case of our juice lineup optimization, one aspect of this problem is figuring out which flavors are most appealing to consumers. This analysis typically falls squarely in the purview of the insights department. However, another equally important aspect is incorporating flavor-specific profit and cost margins into the optimization equation. Here, insights managers often defer to the finance minds to round out the analysis.
But they need not. In fact, the optimization functions generally work better when consumer preference data and product profit margin data are addressed simultaneously.
To see why this is the case, imagine that as the insights manager, you collect the data needed to determine which juice flavors have the most consumer appeal. The results of your experiment reveal that orange juice is the most appealing product, followed by apple juice, lemon iced tea and grape juice, respectively.
You then send this information to the finance team for further analysis. They decide that, even though the orange juice is the product with the strongest consumer appeal, its distribution costs are simply too high given that it requires refrigerated transportation while the other flavors don’t. The finance team also worries about the company’s capacity to produce three flavors of juice. So, in the end, the decision is made to offer a two-product lineup of apple juice and lemon iced tea.
The problem with defaulting to the “next-best” option(s) when one or more products are deemed problematic from a business standpoint is that products are optimized relative to all other potential products in the choice set. So, taking a product out of the running due to a sourcing or distribution concern can change the entire complexion of the optimization function. In our example above, it is entirely possible that removing orange juice from the choice set would cause grape juice to jump ahead of lemon iced tea in an optimized portfolio set.
This is why it makes sense for the insights department to own both facets of the optimization equation. Had the finance minds simply taken over the analysis, they might not have realized that product rankings can change as products are included/excluded from the analysis set.
Granted, it’s possible that there’s more to the profit/cost margin equation that you have available to fold into your analysis. But being able to offer a point of view on the matter to your finance or revenue management team will only add value and strengthen your department’s position as a critical hub of information in your company.
Subgroup preferences
A carefully thought-out plan for subgroup analyses is another facet of portfolio optimization that can mean the difference between stealing market share from competitors or simply treading water.
By subgroup analyses, I refer to any analysis that is conducted on a subset of your data – for instance, figuring out which product assortment is optimal for Millennials, females or Walmart shoppers in your sample. Although this may seem like more of a curiosity-satisfier than an action point, there are many cases where effective subgroup analyses are absolutely critical to a company’s success.
Let’s return to the case of our juice manufacturer. Recall that you identified the optimal product lineup of orange juice, apple juice, lemon iced tea and grape juice, respectively. The finance minds then nixed orange juice due to its high distribution costs and also capped the assortment at two flavors due to capacity concerns. Re-running your portfolio optimization removing orange juice and with the two-flavor constraint, you identify apple juice and grape juice as the optimal product assortment.
If you weren’t interested in exploring subgroup differences, this would be as far as you would need to go. However, might there be more profit to be found if you took the time to explore subgroups?
It’s entirely possible that, while at the national level apple juice and grape juice are the strongest juice lineup, other combinations perform better at the regional level. For instance, it may be the case that U.S. consumers in the south drink more iced tea than in other parts of the country. If this were true, more profit would be found by offering the apple juice and lemon iced tea lineup to southern U.S. consumers and the apple and grape juice lineup to everyone else.
One consideration to keep in mind when planning out subgroup analyses is to make sure you have a sufficient sample size to draw reliable conclusions from you data. Recall that subgroups represent a subset of your data. You need to make sure you’re not cutting your data down to a size that’s too small to draw reliable conclusions. This will be based on the incidence level of your subgroups of interest: the lower the incidence for your subgroups of interest, the larger your overall sample needs to be.
A complicated problem
Optimizing a product lineup is a complicated problem with big implications. Putting the optimal product mix on the shelves can easily mean the difference in millions of dollars of lost, or gained, sales.
My preferred approach to solve these problems is a modified TURF analysis (as described in previously referenced June 2017 Quirk’s article). This procedure, called TURF-War, allows for the seamless integration of variety-seeking behavior, product profit margins and subgroup preferences into the optimization equation. As such, it tends to outperform traditional purchase intent evaluations or choice-based analyses.
However, this is not to say that other techniques aren’t reliable. Most importantly, having a well-designed experiment, a sound sampling strategy and an analysis plan that accounts for variety-seeking behavior, profit margins and subgroup preferences is the surest way to guarantee that you’ll arrive at the profit-maximizing solution.
Then, it’s just a matter of sitting back, sipping some of your favorite juice and watching the profits roll in.