During discussions about budgets and resource allocation, I often hear push-back from line managers of the form: “What we’re planning next year is different from the history you’re looking at, so your results don’t apply,” or “The market environment is different now, so all that history is irrelevant.” Leaving aside clichés (“those who ignore history are condemned to repeat it”), is there truth in these arguments?
The answer is, not really. Time series marketing mix models are commonly based on about three years of recent experience. For marketing drivers for which strong statistical significance was found, models derived from the history should be applicable to future decisions while working within the range of historic experience. That is, if media delivery levels have varied during the modeling period over a range 1X to 10X, we can have high confidence for simulations in that range. A simulation at a 20X level, on the other hand, would be subject to lower confidence.
For more qualitative changes in marketing, some rules of thumb are helpful. Minor changes in media, creative execution, packaging or detailed product specifications are unlikely to cause meaningful forecasting errors – the modeled history very likely contains comparable variations of each of those. A major change in product positioning or quality relative to competition may lead to some shift in advertising response relative to prior history. A change in communication strategy or messaging, if outside historic norms, may also result in a modest change in response. Thus, it can be helpful to triangulate measured response from a product-specific history with benchmarks or norms from comparable brand studies.
In one recent situation, a client planned to broaden its target audience by reaching an additional customer segment, and wanted to determine what impact this might have on advertising effectiveness measures. To answer the question, we looked at factors including the relative size of the segments and their different media consumption patterns, as well as the benchmark advertising response projected for comparable products. Combining all of these inputs, we agreed on a low risk solution that met the new communication strategy needs while remaining within the range of media mix that projected as optimal based on prior history.
A general principle for marketing models is that a balanced integration of quantitative analytical projections and business insights is more reliable and accurate than either in isolation. In other words, neither method should be used blindly. Instead, triangulating recommended budget allocations using the objective inputs outlined above, along with the experiential wisdom of brand management and ad agencies, will commonly yield the most informed (and pragmatic) decisions.