Modeling Marketing Mix Using Smoothing Splines

 Recall a similar set of problems occurring in the enterprise, this could be useful. Computing and statistical -Technical.

Modeling Marketing Mix Using Smoothing Splines  By Slava Kisilevich in TowardsDataScience

Capturing non-linear advertising saturation and diminishing returns without explicitly transforming media variables

The established approach among marketers for modeling marketing mix is to apply linear regression models which assume the relationship between marketing activities such as advertisement spend and the response variable (sales, revenue) is linear. Prior to modeling, media spend variables should undergo two necessary transformations to properly capture the carryover effect and the saturation effect of the advertisement spend. It is known, that advertisement spend is not linear with respect to the response variable and follows the law of diminishing returns. However, the functional form of the saturation curve is not known in advance. 

Therefore, the modeler should first hypothesize about the possible transformation functions that should be applied to each media activity channel to match the true spend-to-response relationship. In this article, I show an alternative approach to modeling marketing mix by using Smoothing Splines, which is the way to model the non-linear relationship between dependent and independent variables within the framework of a linear model. By following this approach, the model will establish the non-linear relationship between media activity variables and the response variable without the need to transform those independent variables to account for the non-linear relationships.  ... ' 

Marketing Mix Optimization Using Bayesian Networks

A long time challenge in the enterprise.  I note this seminar on the topic.  I am about to review.  Remember very well the causal assumptions applied and challenged..   Please do follow up if you have any comments, especially regarding practical use.

Stefan Conrady
Managing Partner at Bayesia USA & Singapore: Bayesian Networks for Research, Analytics, and Reasoning

Marketing Mix Optimization with Bayesian Networks

With over 130 participants, today's event was our most popular webinar yet. No wonder, marketing mix modeling remains a mystery to many! And, neither big data nor clever statistical techniques can solve this problem for us. One way or another, it's about the causal assumptions one needs to make (and be able to justify). However, Bayesian networks make this process much more transparent, and they allow for some brilliant shortcuts! Check out the webinar recording and see for yourself!
https://bayesia.wistia.com/medias/fhqe74avqp

On Marketing Mix

In Data Science Central.  Good detailed view of marketing mix, by Vincent Granville.  All the basics.  Surprisingly non analytical, mostly a look at all the concepts, context and variables involved.  A topic that all CPG companies analyze constantly.  I saw it the first day I joined the big CPG enterprise.

Bloomberg Marketing Analytics Tool

In Adage:  Interesting play.  I do wonder about the overall objectivity that is involved.  But a directed marketing mix play is intriguing.    Dueling tools among multiple advertising venues?   What other value propositions could a tailored analytical tool provide?  Integrated forecasting methods that include context expertise?  Cognitive wrapping to learn value drivers?   " ... Bloomberg Launches Marketing Analytics Tool ..... Application Allows Advertisers To Optimize Investments In Group's Properties .... "   

An Introductory Look at Marketing Mix

A simplified look by a student, at the marketing mix problem using SAS.  Nicely presented.  A classic problem solved by every company that needs to determine advertising and marketing strategy.  Good introduction, though most of the complexity is in the use of the results and their integration into corporate methods.

For more advanced blog posts here on marketing mix.

Thinkvine and Damon Ragusa

Have followed Thinkvine here for some time, out enterprise connected with them and have touched base  number of times since their creation in 2000 by Damon Ragusa.     They have done some very novel  work integrating agent based modeling simulation, marketing mix and related problems.  I like their new video on their offerings.  Simulation is a very powerful analytic technique to understand marketing operations.  In the news today, Damon Ragusa is back at the helm.  More details in today's Cincinnati Enquirer.  Hope to reconnect in the coming weeks.

Data Management Platform Vendors

Examining the breadth of  Data Activation: The Data Management Platform

One vendor, BlueKai,  defines DMP as:

" ... A DMP is a centralized data management platform that allows you to create target audiences based on a combination of in-depth first-party and third-party audience data; accurately target campaigns to these audiences across third-party ad networks and exchanges; and measure with accuracy which campaigns performed the best across segments and channels to refine media buys and ad creative over time.

Slice and dice your own online and offline first-party data to create audience segments for display, search, video, and social campaigns; apply it to ad campaigns targeting audience across the purchase funnel; and get visibility into the ROI each campaign delivered for each segment. ...  Marketers can pull in and analyze all audience data presenting a 360-degree view of how campaigns perform against specific target audiences. ... " 

Cross Device Marketing

In Mobile Marketer:    Not unexpected.    " ... Cross-device advertising commands growing portion of media spend: report  ... Interest in cross-device campaigns is growing
One-third of digital media budgets will be allocated to cross-device campaigns this year, up from 24 percent last year and 19 percent in 2011, according to a new report from ValueClick Media and Greystripe .... " 

Marketing Mix Criticized, Defended

Defense In AdAge:  We used marketing mix since their very earliest inception.   It is a relatively simple statistical technique that suggests the best investment of money to media and marketing channels.  It has become a classical technique in common use in may companies.  It had been getting criticism for dealing well with new channels like the Internet and Social .  Yet it had been adapted reasonably to these channels as well.  Here is the article detailing with the attacks.  Companies are expanding the capabilities and accuracy for marketing mix, by adding newly available detailed data to tune marketing systems to many channels and dynamics.   See, for example. the company Thinkvine,  who we worked with and have covered here a number of times.

Measuring Advertising

Measurement of outcomes is always interesting to me, so this led to the announcement below:

The Wharton Customer Analytics Initiative (WCAI) and the SEI Center's Future of Advertising Program (SEI FoA) are pleased to announce a conference ( Thursday, May 16, 2013) focusing on new ways to measure advertising effectiveness. This fast-paced, one-day program will feature presentations on research projects that were selected to receive grants from WCAI and SEI FoA last year, along with commentary from a number of well-recognized practitioners. Presentations will cover a wide range of cutting-edge approaches to gauging advertising’s effectiveness including:

Field experiments
Advertising models: going beyond marketing mix and attribution
Combining experimental and non-experimental data
New measurement techniques: eye tracking and emotional response

To encourage a lively discussion on what these cutting-edge findings mean for today’s marketing analytics professionals, the program will feature short research presentations combined with panel discussions. The highlight of the day will be a keynote from Deb Roy, an expert in data-driven methods for analyzing and modeling human linguistic and social behavior, and founder of Bluefin Labs (recently acquired by Twitter).

The full agenda and registration information can be viewed here.

Wharton Customer Analytics Initiative
t: 215-746-4161 | e: wcai-mail@wharton.upenn.edu 
http://wcai.wharton.upenn.edu 

On the Rise of the Swarm

Our enterprise did a number of analytics projects based on swarm technology, also called Agent Based Modeling.  We brainstormed warehouse compliance swarms.  Now,  A new survey of work underway, empowered by swarm robotics.  The components of the swarm are simple with minimal intelligence, but as a collective swarm they can do impressive things.

Gamification as Goal Concept Waning?

I was involved in a number of applications in the enterprise that would today be called gamification.  Involving people to engage with systems and data to compete for better value.  Using the concepts used in online games.  Since then I have given several talks on the subject, advised people on its use, even interpreted specific marketing examples and written about that here.   I detect somewhat of a decrease in its  use in the hype cycle of technology.    That is not uncommon.  Should we go into a project with the major goal being to gamify an interaction?  Or should it be considered one technical tool that can be added to the mix as needed?  Too much of the former has occurred.   More here.  In an E-Commerce Times article. " ... Using gamification to enhance the user experience must take different user motivations into account. For example, if the goal is to drive user to donate to a cause, a reward system should engage them to return more frequently and feel emotionally connected with the process. A ranking system might be implemented to encourage friendly competition between users.... " 

Cleveland Indians Using ThinkVine

I just got a press release from marketing mix optimization and simulation modeling company Thinkvine saying that the Cleveland Indians baseball organization is using their technology for marketing mix type applications.  More here.  We tested and used their methods in the enterprise from their early days.    ... “Sports marketers need a high level of flexibility to adjust their plans during the season based on league standings or other factors that are specific to their business,” said Mark Battaglia, CEO of ThinkVine. “With ThinkVine, it’s now possible for sports marketers to forecast how varying marketing plans, along with variable factors such as team performance, influence targeted sports fans to purchase game tickets. This enables the Cleveland Indians to respond quickly and more accurately to maximize their budgets and keep revenue up.”  ...  

Shopper Insight

In Retailwire,  IBM/DemandTec sponsors an interesting paper: Shopper Insights: Actionable or Academic?  An overview of a survey on the topic.   Our innovation centers were established in part to fine-tune an understanding of the shopper from a number of different resources.  From our own experimentation, from the knowledge of our consumer researchers, and from leading edge research done by internal and external vendors. DemandTec was an early visitor,  where we talked about the effectiveness of planning promotion and assortment, using analytic methods.    The paper suggests, as we saw early on:

" ... Shopper insights are here to stay, according to RetailWire’s recently conducted study of industry  professionals and practitioners, and are quickly becoming a “must-have” feature of decision-making relative to category management, merchandising, marketing and in-store activities.In particular, shopper insights derived from specific retail accounts are seen as beneficial for fostering strong collaboration between supplier and merchant. Said one respondent: “It is the new currency to drive retailer/manufacturer collaboration to work jointly to improve the business/category.” ... " 

So we clearly saw the early need for learning that was more than academic.  And produced results that could be applied early to our own mix of products and could take our understanding of the consumer, for new products and new initiatives, and combine that with operational retail knowledge.

This paper studies insights from a survey in late 2010:  " ... The study sampled information and opinion from 593 executives and managers.  Supermarket/grocery retail and manufacturer companies were most strongly represented  in the sampling, and nearly half of respondents are in upper management positions. ..." 

Worth reading for its results.  As an excerpt I always like to look at future actions, since my role was always of a futurist and technologist, and this was interesting:

" ... both retailers and suppliers expect that manufacturers will partner relative to shopper insights with more than the number one retailer in a particular market. Indeed, only 2.5 percent of the  total number of survey respondents say that a supplier should only partner with the market leader and  go no further. From the supplier viewpoint, 20 percent answer that going beyond the market leader  should be done to enhance more retail relationships, while 69 percent say that working with more than one retailer in a particular market should certainly be done for relationships, but also for gaining  additional knowledge for HQ planning.... " 

I am incorporating some of this data in a study I am doing.  Please send me any further insights you have.

IBM DemandTec for Marketers

From IBM DemandTec, a good view of the need to construct valid shopper segmentation.  Why?  So we can use targeting to make sure the right consumer gets the most influential message.  The link goes to an infographic and further  information on this concept.

New techniques are now available to make it easier to do.  Relationship and social marketing allow the integration of new and precise marketing methods.  Linking this further to marketing mix methods, minimizes   cost to achieve a particular return.  To take this further.   And more at the link:

1. Use advanced science and shopper segmentation to better understand shifts in consumer demand and connect with individuals and segments rather than broad markets. Merchants can now surface those insights right at the point of decision of pricing, promotion and assortment activities. By recognizing that each shopper is unique and responds differently, revolutionary merchandisers can tailor strategies to focus on their most valuable shopper segments – providing differentiation at the shelf and a real path to category growth.

2. Stay relevant with intelligent targeting
Invest in capabilities designed to manage the entire promotion lifecycle – from analytics to deployment. Target promotions not with a broad brush, but by shopper segment through localized and automated ad versioning across virtually all customer touchpoints. An integrated planning environment can keep your marketing counterparts on the same page as your team throughout the planning cycle, translating into greater visibility, improved quality, and ultimately more time to focus on strategic issues.

3. Capture value, measure results
Increase marketing ROI by aligning assortment, price, promotion and marketing decisions in a unified planning environment. A common platform also allows you to work more effectively with your merchandising team counterparts – helping teams to plan for category growth, new shopper attraction and behavioral impact on key shopper segments while allowing you to tailor your marketing investment to achieve financial goals..... "

Strategic Marketing Mix Decisions

I have been recently been taking a look at the broad concept of marketing mix decisions.  The problem is old:  how do we decide how much promotional dollars to allocate to what marketing medium?  How much are we wasting?   And how do we time the allocation to make sure it optimizes the results achieved?   It links to the general problem of forecasting demand.   Forecasting depends upon both specific company decisions and external uncontrollable contexts we need to consider.  When I taught forecasting I always included the   phrase 'The forecast will be wrong, but how will it be wrong is the more important question'.  

In recent years too the number of possibilities for forcasting methods has expanded.   Online and social methods, still in their relative infancy, also need to be allocated funds to make them work.   How do you react to rapid changes by competitors?  And the question of how much and when immediately emerges.

I have been taking a look at IBM DemandTec's approach recently to understand how it it integrates with other methods on the market.   My enterprise tested their approaches early on. Their holistic approach based on optimization methods bears a close look. They write:

" ... Holistic marketing mix decisions.  One continuous, actionable plan..... Simply using average marketing investment returns doesn't tell you much about what's driving volume, share, or profit.  If your assumptions are stale, you can't truly optimize your marketing mix in a rapidly changing marketplace. Now you can use the most dynamic predictive tools available to optimize marketing investment decisions throughout the year.  It's all delivered through one comprehensive system.

IBM DemandTec Strategic Marketing Planning transforms traditional marketing mix insights into one continuous analytic system that spans all of today's consumer touchpoints – media, trade, social, and more.  Empower CMOs with sophisticated tools for allocating portfolios.  Give brand managers precise up-to-date data to optimize decisions. .... 

Quickly adapt to marketplace shifts

Optimize marketing spend

Efficiently allocate marketing funds

Fully leverage your existing marketing mix analysis investment   .... " 

Advancing Analytics to Machine Learning

Things are changing in  advanced analytics.   Simple statistics  is not enough anymore.  An example from Orbitz.  Read the whole article:    "... I lead the traditional statistical modelers as well as the chief scientist and the machine learning (ML) crew. At Orbitz, we have found value in incorporating both types of data mining professionals (machine learners and statistical modelers) because many problems are well-suited for both camps. For example, the statistical modelers effectively address areas such as marketing mix analysis, predictive models across online marketing channels, customer lifetime value models, churn models, credit card fraud models, etc. Similarly, the machine learning staff deploys their algorithms in areas leveraging Big Data, where system feedback is leveraged to quickly learn from patterns in order to self-improve – areas such as the Hotel Recommendation Engine and Hotel Sort  on the Orbitz Web site. ..." 

Swarming Drones

Boeing demonstrates swarming drones.   Which reminds me of long ago projects that looked at how simple,uni-purpose, cheap and even disposable robotics could be used to perform relatively complex tasks by aggregating their efforts.  It also points to to the idea of agent based modeling, which we used for such corporate problems as trying to determine the predicted value of new initiatives.  And another designed to understand how complex supply-chains could be improved.  We never attempted to make the agents fly, though.

Adstock Model of Advertising Effectiveness

The use of the Adstock Model was just brought to my attention.  Have never utilized it, but wonder of there are others out there who have and have some comments on its broad use, especially beyond its usual use in TV ads, here is the abstract,  from the full technical paper.

" ... This paper is intended as a review of existing models of Television Adstock transformations that enable the inclusion of dynamic and nonlinear effects of Television advertising within linear sales response models.Television advertising is one of the largest investments for consumer marketing and companies invest a lot of effort in measuring the impact and ROI of TV advertising. This is typically done using either individual response(e.g. Discrete Choice models) aggregate response (e.g. Marketing-mixmodels). For the purposes of this paper, we will restrict our inquiry to aggregate response models. These models are linear in parameters but canaccount for non-linearity through variable transformations.... "

Agent Based Modeling

Gib Bassett sends along this interesting link about agent based simulation.  We actively worked in this space for years.  Agent-Based Modeling and Large-Scale Simulation with Shared Memory