February, 7 2018
Balaji Raman, Associate Vice President, Cogitaas AVA
The term marketing analytics is quite broad, it covers a gamut of problems concerning consumer segmentation, e-mail marketing, pricing, brand positioning, brand equity, fighting competition, spends optimization and allocation. In all cases, business decisions are taken based on rigorous data analytics done through statistical modelling. Here, I will focus on different issues related to a brand, particularly in the context of CPG industry based on my experience.
I work with Cogitaas, a boutique consulting company providing recommendations and suggestions based on statistical and econometric models for CPG, retail, media and e-commerce firms. Using “Data for Decisions” involves statistical methods which are customised for different data sets.
Marketing –it’s All About Communication
Amazon, Apple, Netflix, Coca-Cola and BMW are some of the popular brand names that owe a lot to dedicated marketing campaigns involving innovative ideas in attracting consumers. Some important questions faced here are:
- How do consumers perceive my brand? Can it be quantified?
- What are the key brand attributes that drive demand?
- What should be the message conveyed to drive the next phase of growth?
Traditional channels of communication are TV, print, radio and out-door hoardings. TV has been the dominant channel and has received majority of spends from campaign budgets. However, with the popularity of internet and mobile phones, “digital” media is challenging the dominance of TV.
Given that resources are finite and limited, it is necessary to:
- Measure Return on Investment (RoI) from different platforms.
- Optimize media spends and allocate budgets across platforms.
Data analytics plays a very definitive role over here, as consumers have exposure to multiple media channels, and complex models are used to separate out the effects of each.
Brand Growth – Driving Market Share and Fighting Competition
Determining strategies that help in driving brand growth is a continuous exercise for any brand manager. Some of the questions which we have addressed through analytical models are:
- What are the key growth drivers of my brand?
- What are the opportunities in different geographies?
- Are there any preferences for SKUs (Stock Keeping Units)?
- What is my brand equity and how does it compare to competition?
- How do I fight competition?
Pricing and Promotions
Decisions on pricing are strategic. Poor decisions can lead to irreparable damage to a brand’s equity. Pricing decisions are taken at SKU level as each SKU caters for different segment of consumers. Some questions posed to us related to pricing are:
- What are the optimum price for SKUs?
- How to quantify the impact of pricing on brand equity?
- What will be the impact of any price increase?
- Are the SKUs cannibalizing?
- How to quantify the impact of promotions or discounts given on SKUs?
- Should the discounts be deep or shallow?
- When to provide discounts?
Trade – link Between Consumers and Manufacturers
Retailers either online or brick and motor, serves as a marketplace for consumers to interact with brands. Manufactures provide discounts and incentives for retailers to stock products. These incentives, commonly referred to as trade promotions account for the biggest spends for a brand, even more than advertising budgets. Brands also compete to buy prominent shelf space (visibility) in a retail outlet. Some problems related to trade are:
- How to quantify the impact of number of retailers on volume sales?
- What are the opportunities in different geographies?
- What is the impact of visibility in a retail outlet?
- How are retailers reacting to different incentive schemes? What are the RoIs of schemes?
- Which incentive scheme is appropriate for each retailer?
The part of marketing that involves statistical models to solve above problems is known as marketing science. I have been working on this field for the past 5-6 years. It gives me immense joy to see “applied” statistics in action and in seeing recommendations from models developed by us being implemented in practice. Cogitaas is very particular about monitoring the business benefits accruing from, “Data for Decisions”.
There is an extra element of challenge, when dealing with emerging markets like India, because of issues in data and the presence of cheap local competition with a strong distribution network.
Typically, our work is done via the following steps:
(1) Data processing and cleaning as well as data transformations;
(2) Statistical modelling: Generalized Linear Models (GLM) and time series are the most commonly used tools to solve the above-mentioned problems. Over the past couple of years, we have shifted more towards Bayesian methodology (especially hierarchical models). There are certain advantages in using Bayesian methods, which we have observed and experienced;
(3) Creating appropriate User Interfaces (UI) tools like Tableau, Shiny or Qlik.
Data processing may seem to be less attractive by modellers. However, I believe spending time on data processing would enable a modeller to know about sources of various datasets available, procedure behind data collection and in general a feel for data even before modelling
Statistical rigour and high levels of testing standards are important even with relatively poor data sets. As a statistician, I would emphasize three critical points:
- identifying a “suitable” model for a given problem,
- focusing on modelling ethics, and
- interpreting the message given by a model for use by business managers in different functions in the organization.
After this work is done, there is more work in terms of cascades and “buy-ins”. Sometimes model results will be against a client’s conventional belief system. The conviction process involves further supporting models which accept or reject implications of conventional hypotheses.
Sense of achievement comes from seeing large global corporations accepting and implementing “data driven decisions”.