The most advanced and adaptive approach to modeling -- we focus on the increasingly complex marketing landscape from a top-down perspective. We leverage the best modeling approach for each client, not one size for all.
The latest in a sequence of timeseries modeling advances, NextGen Marketing Mix Modeling integrates several advanced modeling techniques into one unified model estimation process. Specifically, we integrate high order Bayesian pooling (5+ hierarchies), with high frequency data (daily, hourly, etc) and Dynamic State Space estimation of long term effects. While none of these approaches is unique in themselves, unifying them into one estimation approach is exclusive to the Truesight and Marketscience Partnership. Furthermore, this isn’t simply a technical advancement, but it allows us to better estimate the increasingly complex consumer demand equation and therefore better measure the underlying economic theory that underpins all modelling approaches.
The benefits of this approach overcome several shortcomings of traditional Marketing Mix Modelling:
The various impacts of marketing can be modeled out in one common estimation; specifically, we are able to isolate short-term impact programs (promotional and campaign based) from long-term programs (typically brand based)
The high frequency and high order pooling allow us to drill down to greater levels of granularity and detail for short term effects. This provides greater clarity into day-part and placement level details to further optimize media investments
Higher-order pooling allows us to measure all these effects at the appropriate geographical and product hierarchy level and thereby overcome issues of aggregation bias
The Dynamic State space allows us to measure the subtler impacts of the Brand Experience (BX) and Customer Experience (CX) on sales thus integrated perceptual and experiential metrics into the common estimation thus informing communication strategy
Higher-order pooling allows us to model at zip code and/or customer micro-segment level and therefore better align with test and control experimentation and programmatic buying approaches
A flexible ‘adstock’ approach that leverage’s Polynomial Distributed Lags (PDLs) allows the data, rather than an a priori transformation, to specify the advertising response and capture’s potential wear-in, decay and diminishing return effects
A prerequisite to fully leverage this approach is that we have a complete and detailed level of data. The pay-off however is in a greater ability to optimize and driver greater ROI improvements.
The scope of measurement under NexGen Modeling is very comprehensive:
Price/offer and promotions short term lift
Traditional, digital and social media in the short term
Key environmental drivers (economy, competition, weather etc)
Seasonality estimated within the timeseries estimation process
Long-term brand impacts of changing consumer sentiment and experiences
Long term impact of pricing changes and policy changes
Interactions of short and long-term drivers
Ultimately, the measurement has to inform decision making. Our obsession in building the most complete model is so we can derive reliable and accurate estimates of marketing impact and response. It is these estimates that will inform our optimization & simulation tools and identify real ROI improvements for our clients.