We apply models flexibly, and can structure analyses to suit the needs of a particular project and the confines of the data available. Some prototypical applications of marketing models that Gradient can apply are below.
Segmentation is the process of dividing the market into separate customer bases, each with demographic, behavioral, and psychographic patterns that makes their purchasing behavior unique. Which segments are right for you?
How are your customers reaching a purchase decision? What are the steps they take? Where do they get stuck and fall out? With individual-level CRM data, we can model this process directly and produce richer insights than are available with generic path-to-purchase models.
How well did your campaign perform? When customers are seeing ads in many different online channels, or in offline channels (like print, outdoor, or TV), getting reliable data about the impact of any one media spend can be difficult. New causal inference methods can shed light on the performance of your media spend.
How likely is it that a given customer will purchase? Or renew their subscription? Powerful predictive models can be used in these cases to score your customers and decide which customers should receive an intervention.
How much should you be willing to spend to acquire a customer? Are your customers becoming more or less valuable over time? How much residual value do your existing customers have in their future purchases? Advanced probability models can model the underlying process of customer purchase behavior and retention directly, which give substantially more accurate results than traditional Excel-based modeling.
Which feature bundle should we choose? How much more would our customers be willing to pay for a particular feature? What are the price sensitivities of different market segments? Pre-product launch, choice-based-conjoint surveys (also known as discrete choice experiments) can simulate customers’ choice among alternatives and produce estimates of relative market share, relative price sensitivity, and clusters of feature preferences.