Our micro-level modeling service looks at the demand patterns of individual consumers, with a focus on Multi-touch attribution
Multi touch attribution (MTA)
Focuses specifically on the combination of online marketing touchpoints that contribute to online sales conversion. Measurement is based on the online journeys of individuals using various techniques, ranging from econometric choice modeling (Logit) and probability chains (Markov) through to economic game theory (Shapely values). Outputs are used to assign credit to and allocate budget between online media investments.
Whichever attribution technique is used, it is well known that the MTA process is subject to three key challenges:
- It does not incorporate offline mass media and pricing effects
- It cannot accommodate walled-garden data
- Selection bias and no baseline control imply that attributions are not necessarily causal
These challenges imply that current MTA practice is of limited value in evaluating the true incremental impact of online media on sales conversion. For clients that wish to focus solely on online marketing performance, we typically deal with these challenges through a combination of:
- Discrete choice regression modeling with latent instrumental variable techniques to control for observed and unobserved confounders
- Walled garden data A/B test results as prior inputs into the analysis
- Customer loyalty Modeling
Customer loyalty is central to strong brand equity and Customer Lifetime Value (CLV), where CLV is defined as the net present value of the current and future customer base. Through a process of customer acquisition, retention and expansion, firms may directly increase CLV and firm value. Customer loyalty models are designed to assist clients along each step of this process.
Companies grow in value by acquiring customers new to the market and/or by stealing customers away from their competitors. Acquisition models are designed to understand the characteristics of high value customer prospects. This information is then used to help set prices and acquisition budgets and optimally target marketing resources.
Any company that does not maintain a healthy acquisition rate in the face of customer attrition will see its market share gradually declining over the longer-term. Consequently, it is imperative that once acquired, firms need to work to retain customers.
Ensuring the longevity of customer relationships is critical to maximizing customer lifetime value, long-term firm value and profitability. Retention modelling plays a central role in understanding the factors that determine the likelihood and incidence of individual customer churn such as optimal price setting behaviour and marketing spending levels. These issues are of key importance in financial, contract or subscription based industries, where strategies appropriate for customer retention are often very different to those relevant for acquisition.
Add-on Selling Modeling
Add-on selling consists of increasing sales via a combination of cross-selling, up-selling and/or selling greater quantities of existing products and services. This increases the baseline lifetime value of the acquired (and retained) customer and is thus critical in maximizing potential customer equity. Models for add-on selling address a variety of business questions such as the range and quantity of product offerings, the timing of new product introductions and customer response to product offerings.
In circumstances where data are unavailable for demand modelling through econometric techniques, our experimental test and learn service focuses on alternative techniques to quantify demand response to price and marketing levers, with a focus on:
- Conjoint analysis
An alternative revealed-preference technique, allowing us to predict customer response to alternative pricing and marketing strategies. Key outputs include:
- Simulation of “what-if” scenarios involving new product introductions and deletions.
- Predicting the value of brand (equity) in terms of actual price premiums that a brand can charge when all else is equal
- Predicting share premium that a brand commands when all else is equal.
- Ad copy and placement testing
A/B testing and quasi experiments