Trade promotions are like families: You can't do without them, but sometimes you don't know what to do with them.
Formulating and scheduling trade promotions for maximum impact is one of the toughest challenges in retailing. Many grocers leave this partly or wholly up to their CPG suppliers as part of continuous category management, but this presents its own challenges, for both sides.
The mission of the Promotion Optimization Institute (POI) is to foster collaboration between CPG companies and retailers by leading the emergent collaborative marketing profession through developing, advancing and disseminating trade marketing and merchandising knowledge and education (via POI's Collaborative Marketing Certification). To that end, the seventh biannual POI Summit, held April 6-8 in Chicago, hosted 250 attendees featuring presentations by CPG companies, retailers, academia and service providers about how to predict, plan and optimize pricing and promotions so that suppliers, retailers and their shared shoppers/consumers benefit.
The summit's last day was devoted to a unique exercise designed by POI with Gartner. Five different solutions providers were given historical sales data from actual retailers and CPG companies, which they used to predict and optimize promotional outcomes. The five were given wide latitude in interpreting the data, determining its relevance and deciding what kinds of promotions were indicated. The data were masked as to the identity of the retailers and products, but otherwise, they comprised sales information at every level, down to individual stores.
"Each of the participating solutions providers will have a different and unique approach," said Michael Kantor, POI's founder and CEO, in introducing the presentations.
Differences in the approach included which product categories to pay the most attention to, how to account for "outlier" numbers, how to develop algorithms and verify them with the data available, how to settle on a baseline price and how to determine price elasticity. Each presenter, however, worked with the same data set and was tasked with recommending the best approach to trade promotion optimization.
The presentations were mostly directed toward CPG companies, since they usually bear most of the responsibility for trade promotions and are expected to deliver insights to retailer partners. Promotion optimization has obvious benefits for retailers, and their cooperation is necessary to make it work, Kantor says: "Retailers as well are beginning to grasp hold of this and what is its meaning for them, far outside of simple trade dollars."
Nielsen prides itself on "Revenue Management Optimization." That means taking into account as many factors as necessary to yield the right solution, said presenter Rick Hall, Nielsen's senior vice president for trade promotions.
In evaluating the data sets in the TPO Challenge, one of the first decisions the Nielsen team made was to examine the data at store level, rather than look at data aggregated by market.
"If you don't have store-level data, you can't effectively model price," Hall said. "If you can't model price, you're going to tend to over-invest in features and displays, because you're really not going to understand how to price-strategize your business."
Examining store-level data, for both the test "client" retailer and its competitor, allowed the Nielsen team to determine price elasticity and optimal price points. The Nielsen team was able to identify two thresholds, $3.50 and $3.99; surpassing them caused sales to drop especially hard.
"What we see here is an enormous amount of volume being sold by the [product] manufacturer above the threshold," Hall said. As a result, "they're losing the effect of that threshold in elasticity" and pricing a large proportion of product in a way that will depress sales without enough compensatory increase in margins.
Everyday low pricing (EDLP) had been part of the existing manufacturer-retailer strategy presented in the data sets. But the Nielsen team determined that the EDLP at $3.35, had been too low to allow for adequate margins. The team adjusted the EDLP to $3.49, just below the first threshold, and combined year-long EDLP pricing with holiday-related sales and other temporary price reductions.
Under this scenario, the final numbers showed an increase in retailer profit of $18,000 for the year on the product category, while the manufacturer's profit increased $96,000 and trade spending increased only $48,000. By tweaking the timing and length of the promotions, Nielsen was able to optimize their effects while still keeping the benefits of the adjusted EDLP, Hall said.
"The key thing that we learned here is that by getting the right pricing, in addition to the promotional strategy, we're able to drive a scenario that will be a win for us and a win for our retail partner," he said.
The team from TABS Group had an unfortunate incident with a spam filter that prevented them from getting crucial data until barely a week before the TPO Challenge deadline, said CEO Kurt Jetta. But they turned it into a positive: "It was a nice stress test for [our] product, showing we can get things straightened out and turned around pretty quickly," Jetta said.
Due in part to the short deadline, TABS focused on one of the data set's four product categories, which represented 97 percent of the client CPG company's revenue. It found that a hi/lo pricing strategy was the least productive, actually yielding a negative ROI of 0.2 (defined as the ratio of incremental profit divided by promotional spend).
Jetta identified a metric that, he said, doesn't get enough attention: "incremental factor," which he defined as the percentage of a product's sales that are directly attributable to individual promotions. The three retail strategies that TABS tested were hi/lo, EDLP and a hybrid strategy combining the two.
One reason hi/lo came out so poorly is that it carried an incremental factor of 22 percent, well above hybrid (12 percent) and EDLP (3 percent). Combined with other metrics, like a high spend rate on promotions and a relatively low lift per event, this means promotional dollars are being used inefficiently in a hi/lo strategy.
In terms of ROI on promo dollars, EDLP came out on top by far among the three retail strategies studied, at more than $15 in revenue for each $1 spent. This, however, is due in large part to fewer overall promotional dollars being spent in the EDLP scenario.
One of the advantages of the TABS approach is that it can pinpoint the influence of promotions on sales with great precision, even across retailers, Jetta said.
"When you get that call from the Walmart buyer on a Sunday and [he asks you] why you're going so low on price at Walgreen's or Kroger, you can access your data and just immediately say, 'Well, it didn't hurt you. In fact, I see you have this buy-two-get-one-free display coming up that's also working pretty well,'" he said.
SAP developed individual models for every possible combination of the products, product categories and markets in the challenge data, resulting in nearly 1,800 separate models, said Colby Sheridan, SAP's global director for consumer product industry business solutions. This approach allows CPG companies using the SAP model to suggest promotional tactics and schedules that are highly specific to product-market combinations.
For instance, SAP was able to determine that shoppers for product category No. 2 (out of four) were the most price-sensitive, but there were wide variations within that category, with some individual products or markets having double or triple the price sensitivity of others. Conversely, product category No. 3 was deemed "moderately price sensitive" overall, with 156 of the possible 489 product/market combinations judged as price-inelastic.
SAP used a straightforward way of testing its predictive model: It developed a model based on all of the challenge data except data from the last six weeks, then input the pricing and promotions data from those weeks and compared the model's predictions to the actual sales. The result was an accuracy rate of better than 91 percent, Sheridan claimed.
SAP's metrics also include a process called "advanced volume decomposition," which breaks total sales volume down into five components: tactic, price, holiday, seasonality and baseline. Bar charts that show this breakdown by month can be generated for any possible category/product/market combination. Doing so can reveal more useful information than examining the data for an entire product category. For example, the advanced volume decomposition of category No. 3 shows that most of the sales depend on baseline volume. But drilling down to a particular product and market shows that a majority of the volume there is tactic volume (i.e., resulting from short-term promotions), with seasonality also playing an important role.
These kinds of insights can help a CPG sales associate draft promotional plans that are highly customer-specific, Sheridan said. SAP's software can also spell out the history of a specific tactic for a particular customer.
For instance, interpreting a particular product's sales trends, "we know that promotions are effective for this kind of product, but displays are not," Sheridan says. "So don't spend your money on displays. Spend it elsewhere."
T-PRO Solutions prides itself on giving clients real-time information, integrated into a single database.
T-PRO ran a real-time demonstration for the POI Challenge audience, first examining the sales history as outlined in the data. The data for the challenge set an everyday price of $2.96. When the product price was discounted to $1.67, the ROI was not acceptable because the deep discount eroded profit. When the price was set at two for $4 with ad and display support, once the merchandising fee was factored in, the ROI came in at negative 22.9 percent even with the higher price, said John Weller, T-PRO's chief knowledge officer.
This kind of historical data puts T-PRO in a position to accurately examine different scenarios for trade promotions, Weller said. When the price point was set back to two for $4 for one week, mimicking the earlier sales event but without the merchandising fee, the T-PRO model predicted a volume lift of 3.31 times normal sales volume. The historical data reveal that not only do special displays not help, the sales volume lift actually is lower with them than with the price reduction alone, suggesting that the (unspecified) product in the challenge's data sets is commodity-based and not sensitive to price, Weller said.
When an alternate scenario is run through the model, with a price of $2.19 and no special merchandising, the volume lift is lower, amounting to an extra 900 cases of product, compared with an extra 1,200 cases under the two-for-$4 deal. But because the promotional spending is lower, the ROI works out to better than 12 percent, Weller said.
This kind of data can be used to build a trade promotion calendar using another module of the T-PRO software. This allows users to predict the effects of alternative trade calendar plans over the year, tracked against budgets for sales, promotional spending and other factors.
Weller ran a promotional schedule that consisted of several two-for-$4 deals spread throughout the year. The results showed increased revenue, but this was offset because the trade marketing budget was exceeded. A new schedule, replacing several of the two-for-$4 deals with discounts to $2.19, yielded similar results with promotional expenses below budget.
Accenture concentrated on the top 10 of the 51 retailer accounts in the data sets, since those accounted for 80 percent of total revenue. Its recommendations, extrapolated across all retailers in the challenge, would result in $30 million in additional revenue, said presenter Kimberly Bryant, senior manager for strategy.
This narrowed-down approach reflects two aspects of Accenture's general strategy on TPO: establish a hierarchy, and simplify. The hierarchies, for both products and customers, were not in the data sets, but if this were a real account, "we would work with you directly to understand your true hierarchy," Bryant said. As for simplifying, the first of Accenture's recommendations was to consolidate promotional activities and reduce the number of promotional events, to cut costs and increase impact.
One of the first steps for the Accenture software is to determine which product/retailer combinations are the best candidates for automatic modeling, based on how well a model developed from the first 18 months of data predicted the results in the last six. Of the top ten accounts, two scored 91 percent in "uplift accuracy," the predictability of changes in sales volume.
The next step is to determine the baseline price. Accenture took the baseline data provided in the challenge's data sets and "smoothed it out" by filtering out factors like EDLP fluctuations and promotional events. "We want to have a smooth baseline to reflect seasonal trends in a much more accurate way," said senior product manager Martin Burgard.
Once the baseline is determined, Accenture has three modeling methods to predict promotional uplifts: geometrical, which analyzes shapes in charts of the promotional data; neural network, which analyzes patterns in the data; and linear regression, which determines correlations among variables. The uplift prediction was run using the promotional tactics that were furnished as part of the challenge. The result was an average uplift factor of 1.02 and additional revenue due to promotion of $4.5 million for one of the two selected accounts, Burgard says.
Accenture then developed a promotional plan for the coming year. The result was an additional $30 million in predicted revenue, even when cannibalization is taken into account, Burgard said.
Here's how the POI defines Trade Promotion Management and Trade Promotion Optimization:
Trade Promotion Management (TPM) is the process of planning and accounting for the funding of events and activities at retail including brand management, budgeting, account management, demand planning, integrated sales and marketing, retail execution, back-end processes (including settlement), and analytics.
Trade Promotion Optimization (TPO) is the process of utilizing integrated goals, factoring in promotional (e.g. price, duration, seasonality) and supply constraints, plus predictive analytics to create continuously improving trade promotion strategies and results. TPO requirements include: