The Neustar MarketShare Modeling Approach

 

Today’s marketing environment has exploded in complexity. The opportunity to keep up with the proliferation of online and offline customer touch points, direct and mass channel vehicles, and a multitude of delivery devices, makes marketing measurement and attribution more challenging than ever before. Traditional mix models, test/control experiments and judgmental attribution methods are not comprehensive enough to provide timely and credible answers to questions regarding marketing allocations, impact, and trade-offs.

The Neustar MarketShare solution deploys a modeling approach that uses analytical techniques, such as log-log multi-regression models, Bayesian approaches, diffusion models, and the like, based on the particular challenge at hand. Our econometric method identifies the causal relationships between outcome (e.g., consumer purchase funnel and sales) and marketing and other business drivers based on observed behaviors. This method enables measurement of a range of stimuli (e.g., media, sales promotions, incentives, economy) that may be occurring simultaneously.

The most important differentiator of this approach is that Neustar MarketShare models use actual behaviors and impacts that are not simply theoretical aggregates of self-reported media research data. While more basic techniques can work fine for traditional, low-involvement consumer goods purchasing, they may introduce error factors into higher involvement, more considered purchase categories.

 

Modeling That Identifies Causal Drivers and Formulates Outcome Hypotheses

These hypotheses include an expectation of the direction of impact; the magnitude of the impact; and the lag between the cause and effect. Coefficient signs, standard errors, and t-statistics provide the basics for evaluating whether specific variables are retained in a specification. In addition, we assess coefficient magnitudes and appropriate lags based on expectations derived from past experience and a comprehensive library of over 6,000 prior studies and results from academic research.

Among the many marketing variables included in the Neustar MarketShare marketing effectiveness analysis some—like TV, print, paid search, YouTube video ads— do not necessarily exhibit an immediate sales transaction. Neustar MarketShare tests for lags on such variables independently, by channel, to measure the short-to-medium term persistence of response using either simple lags or ad stock-type measures. These results provide valuable insights into how the marketing drivers are translated into changes in consumer engagement behavior.

 

Control Media Spend and Optimize Marketing Mix, with Confidence

MarketShare provides a forward-looking ability to evaluate diminishing returns as a basis for optimizing the marketing mix. The forward looking ability of MarketShare models has two important implications:

  1. There is a saturation level beyond which further investment yields minimal incremental benefit
  2. Deciding how to spend the next marketing dollar depends on where you are on the marketing response curve, which may not always favor the tactic with the greatest overall lift factor.

Media Channels Work Harder Together Than Independently

Consumers are constantly absorbing advertising messages from multiple media outlets. Each of these messages and platforms complements one another, reinforcing the overall message. The result is that advertising can be more effective if these channels are used in tandem.

Mathematically, this means that advertising resources work multiplicatively with each other, and that their impact is not simply the sum of those efforts, but the product. The chart below on the right demonstrates the multiplicative benefit of spending the same budget across multiple touch points.

 

Understanding the Benefits of Direct vs Intermediate Effects

Modeling Only Direct Relationships Between Outcome(s) and Marketing Stimuli Results in Sub-Optimal Mix and Budgets

The Neustar MarketShare approach accounts for the influence of marketing on intermediate outcomes— organic search queries, own-site web traffic, online video viewing, social media exposure, brand awareness, and so forth—thus revealing the full effectiveness of a brand’s different marketing tactics.

By identifying potential drivers of outcomes, we obtain data series’ that allow us to develop models that trace and quantify the consumer decision journey through indirect pathways such as Google organic query volume, social media mentions, brand metrics, and website metrics. While controlling for the impact of external forces such as the economy, competitors, seasonality, and the direct impact of marketing on sales, we are able to measure the full effective of marketing

This conversion from information seeker/receiver to purchaser occurs along the intermediate (or “indirect”) pathways. The impacts of these pathways, which include brand consideration conversion pathways, comprise the intermediate effects of marketing. These effects can be quite substantial relative to direct effects. By modeling both the direct path and the multiple indirect paths to purchase, Neustar MarketShare models uniquely capture the full effect of a brand’s marketing investments.

We have found that by modeling these indirect pathways, true marketing effectiveness is as much as 30% greater than measured by modeling direct effects alone. Failing to account for these indirect effects in optimizing marketing budgets can lead to substantial resource misallocation. The chart on the previous page illustrates the importance of the indirect intermediate - path effects play (shown in green) in contributing to overall marketing response for a consumer electronics company.

The Neustar MarketShare Applied Modeling Approach

Neustar develops models incorporating potential drivers using regression techniques to find the mathematical relationships between marketing tactics and sales. We do this by relating the week-to-week shift in sales volume to the presence/absence of multiple marketing tactics and finding which tactics/combinations explain the sales variance.

In the analysis, we also test for any lags on the impact of a driver outcomes across all of the variables included in our modeling.

The final form of the equation is multiplicative, capturing the synergies between marketing vehicles: Model form: Y = β0ΠXiβi

For example, a two driver model might take the form of:
Weekly Unit Sales = Base Demand x PS InvestmentPS Lift Factor x TV InvestmentTV Lift Factor

Where:

  • Base Demand = sales not attributable to marketing activities
  • Lift Factor = driver coefficient or measured impact on sales of a change in driver

The models estimate the lift factor (or “elasticity”) for each driver in the model. Below are two charts illustrating the market response curves (with diminishing returns) calculated from the elasticities measured by a regression analysis. The TV lift factor is higher than the paid search lift factor, which creates a steeper slope for marginal returns.

Once we have developed econometric models that accurately reflect the impact of the ecosystem of drivers that impact sales (or other outcomes), Neustar is then able to provide clients with actionable insights and forward-looking recommendations. Relating to marketing, the models and the underlying lift factors provide clients with the ability to identify optimal allocations and the greater marginal return on incremental investments.

The optimal mix is informed by the diminishing returns of marketing and the relative impact of a given marketing vehicle on sales. By reallocating marketing investments to the point where the marginal return on investment is equal across all marketing vehicles, our clients are able to achieve the efficient frontier for their marketing investments. This is accomplished by evaluating the sales response for each marketing tactic given a level of spend and finding the balance across vehicles. Two important implications of this methodology are:

  • Identifying the saturation level for each tactic, and
  • Determining which marketing vehicle to spend the next dollar on based on where you are on the response curve.

For example, while the marketing response curve for Medium A is steeper than for Medium B—a dollar spent on Medium A would generate more sales than a dollar spent on Medium B—the marginal return of Medium B is greater. Reallocating $80 of spend from Medium A to Medium B would result in a net increase in sales of 2X.

Download PDF