Forecast trade spend ROI for promotions

The economics of trade spend is often brutal because it’s hard to both measure past performance and predict future outcomes. Each type of promotion serves a different purpose and has different time horizons, so that further complicates the process.

From working with hundreds of different companies, we’ve seen that almost everyone has POS data (syndicated like SPINS/NielsenIQ/IRI), retailer portals (Walmart, Whole Foods, Ulta, Target), and promo calendars. Even with abundant data, forecasts can be much more rigorous than what's being done in spreadsheets today.

Last minute decisions about running steeper / more discounts also eat into margin and therefore growth. A good trade spend ROI forecast will allow brands to spend less and keep more margin, especially if the category is competitive.

What defines forecasts for trade spend and promotion ROI?

A trade spend ROI forecast is an estimate of incremental dollar sales for trade spend, and range for how uncertain that estimate is.

Trade ROI = Incremental sales for the event / incremental trade dollars spent

The incremental sales calculation should account for baseline sales that would have happened anyway, discount depth, trade mechanics, and cannibalization. Otherwise almost every promotion will look good if baseline and hidden costs are ignored.

Minimum viable forecast model

1. Define the exact promotion event with as many attributes as possible: retailer, banner, region, weeks active, SKUs included, discount depth, and type.

2. Build your own baseline based on recent trend and last year same period, then adjust for ACV changes, price changes prior to the event, and known disruptions (OOS, weather, resets). A baseline is foundational to any forecast, so it’s important to get this right.

3. Compare against past similar events where there’s same retailer/SKU/depth combination, then separate lift from price and type of promotion (i.e. feature, display)

4. Forecast a range instead of a single number. Have a base case (most likely), downside (account for execution risk), and upside (good timing). It’s impossible to get a forecast exactly right for a wide variety of factors, so it’s important to give a range so the team can get more accurate expectations.

The final output should be able to answer expected incremental units, expected incremental dollars, ROI range, top risks, and backup plans in case of underperformance.