Trends and Challenges of ROI Measurement: Tip #5

NON LINEAR ADVERTISING RESPONSE:

 Practitioners in media modeling world often use complicated non-linear functional forms to estimate media response curves. For instance, a common approach is to assume that the media response curve is S shaped, i.e. has an initial convex and subsequently a concave section. This S shape is a result of two different factors: (1) threshold effects of advertising: the phenomenon that marketing efforts are not effective until they reach a certain minimum level and (2) diminishing returns to scale: the fact that too much advertising leads to a reduction in its effectiveness at the margin. This has led to the use of a non-linear S shaped media response curve by practitioners. Apart from these effects, there are a large variety of different factors to take into account including various types of media, advertising lags and multiple brands / markets / stores, all of which increase the complexity in the relationship between advertising and sales often leading to the use of complicated functional forms to get a good fit.

Guest Contributor Pat Bhattacharya, Managing Principal, Thinkalytics

Trends and Challenges of ROI Measurement: Tip #4

HALO EFFECTS and MULTICOLLINEARITY: 

Several marketing and media efforts often take place at the same time (e.g.: coupons may coincide with magazine and TV ads). Since combinations of marketing activities often work better than individual campaigns, marketers frequently use multiple campaigns simultaneously to take advantage of halo / interaction effects. However, this also means that the influence of one variable may be difficult to separate from that of another. This is known as the problem of collinearity and the two variables are said to be correlated with each other. When this happens, we may end up estimating only a joint effect for two or more media variables that are very highly correlated. Similar difficulties arise from the common practice of the various different media efforts coinciding with seasonal peaks making it more difficult to separate the effects of seasonality from those of marketing. Some concurrent occurrence of marketing campaigns is, of course, essential, in order to take advantage of, as well as measure halo effects. However, since we also want to measure the effects of each campaign separately, care should be taken during implementation to avoid simultaneous occurrence of two or more campaigns over an extended period of time.

Guest Contributor Pat Bhattacharya, Managing Principal, Thinkalytics

Trends and Challenges of ROI Measurement #3

Difficulty Measuring Advertising Content:

The content and creativity of an ad is an important determinant of the effectiveness of advertising. However, it is very difficult to measure the content or creative qualities of a particular advertisement and compare it to any other ad. The usual practice is to use marketing expenditures or gross rating points or impressions to measure advertisement. However, two advertisements with the same gross rating points or impressions may have completely different responsiveness based on the emotional appeal or creative genius of the ad and this should be factored into the analysis.

Guest Contributor Pat Bhattacharya, Managing Principal, Thinkalytics

Trends and Challenges of ROI Measurement #2

lack of good data:

Accurately measuring the effectiveness of marketing campaigns requires data on the different variables that may have historically affected the firm’s performance. However, corporations may not have data for several of the important marketing variables that have affected their sales in the past. An example is an electronics equipment manufacturer who may have data on its past media expenditures and large campaigns but have no information about the local newspaper and TV ads by individual retailers or local promotions offered by these retailers, all of which drives a great deal of traffic to their products. Building the tools and processes for the collection of such data is a difficult and time consuming process and it may be years before this information is tracked and collected to a standard required for analysis. However, not having the data on so many of these important variables makes it very difficult if not impossible to account for the effect of these and other media variables on their sales.

The emergence of the new social media channels further exacerbates this problem since the data from these channels are still unstructured and have yet to be accurately captured for analysis.

Apart from missing variables, the quality of the data that is available may also be suspect.

Missing variables and measurement errors introduce bias in the estimated impacts and care should be taken to minimize these kinds of errors.

-- Guest  contributor Pat Bhattacharya,  Managing Principal, Thinkalytics

Trends and Challenges of ROI Measurement #1

No standards of measurement:

One of the greatest challenges that the industry faces is the lack of consensus about how the effectiveness of marketing should be measured. In July 2004, Forrester Research conducted a survey of 54 members of the association of national advertisers (ANA) who were asked for their definition of marketing ROI. 70% of those surveyed said gaining agreement on the definition of ROI was somewhat/very difficult. Moreover, 78% cited measuring the sales impact of marketing as somewhat/very difficult.  

Since how well a campaign works depends on so many different factors including the product, the company, the competitors’ actions, the overall state of the economy, government regulations etc., it is difficult to quantify how much the campaign itself adds to the outcome. Moreover marketing affects Sales in the short run as well as in the long run and can also lead to increases in brand equity. While it may be possible to identify and measure the short term incremental sales due to a marketing campaign, it is more difficult to estimate the long term impact arising from that campaign and virtually impossible to figure out the brand equity that results from it.

In practice, most measures only take into account the short term gains from marketing which means that marketing returns are often underestimated. This can lead to underinvestment in media spending which in turn may have detrimental effects on long term profitability and brand equity. 

-- by guest contributor Pat Bhattacharya, Managing Principal, Thinkalytics

The importance of Customer Life Time Value (CLTV) -- my mouse trap story

Being an analytics consultant, I always preach the value of taking a comprehensive view of the customer to my clients. It makes a lot of sense to weed out risky customers during acquisition or to put in greater effort to retain customers who are likely to spend more in the future.  This is the basic premise of CLTV. Further, for this measure to be really actionable it should be tied to the CRM, so that tactical and strategic  decisions can be influenced by the CLTV score. 

Recently I was very thrilled to see a real life illustration of this by none other than amazon.com. The winter being what it was in New Jersey, we had a somewhat unwanted guest in our house seeking refuge from the harsh outdoors-- a mighty mouse!  Once I discovered the guest, I was determined to get it out, but being soft at heart, I researched the Internet for "humane traps". Found quite a few on amazon.com, from where I buy almost everything that I need that does not grow on trees or has feet. Leaving nothing to chance, ordered about 6 of them to be delivered express. Well  for some reason the package never arrived. After a few days, I sent a complaint  email to amazon.  I expected to have to duke it out with the  retailer, supplier and the delivery service  to get my money back--- especially because the product status was showing delivered. Imagine my surprise, when within 10 minutes I get an email stating that 5 new "humane traps" were being shipped, and the money was being refunded for the sixth variety as it was out of stock.  The shipment arrived the next day. 

If amazon was to  do this for all customers, without  even asking any follow up question to verify the facts, it will very soon be inundated with frivolous claims. However, the strategy makes very good sense for customers who  do a lot of business with amazon.    And speed to action is everything here, because instead of complaining about a bad experience, I am actually praising them for the way they remedied the situation. And to do this at the speed they did, the score has to be tied to their CRM. 

For those curious, the mouse was caught and released in a nearby reservation area.

-- Datamining_guy

Neuroeconomics

A very interesting article which compares Humans ability to predict future events, based on past trends, with those of rats and pigeons; and Humans do not come up on top! The author then goes on to explain it based on the the way the brain functions.

Very fascinating!

This reminds me of something we used to joke about in our college days in India. Our hostel was right next to a wooded area and if you forgot to shut your window, monkeys would come in and rummage through your stuff and go off with your food, especially bananas. Now, there was a banana vendor right next to the gate of our hostel. Monkeys being held sacred by a large section of the population in India, a lot of people would buy some bananas and feed them to the monkeys. What was amazing was that, never once would a monkey attack the open cart laden with ripe bananas, but would patiently wait for someone to buy one to feed it to them. Now let the same monkey spot an open window in a forgetful student's room ---different story. Those being the days of "rational expectations" theory, we had our own rational expectations joke about the monkeys optimizing their steady flow of bananas, and not driving the the vendor out of business by attacking him!

Well we might have been correct! Looks like animals brain are better wired for rational actions.

Enjoy the article.

-- datamining_guy


The Durango Herald 02/16/2011 Rats can teach humans some investing savvy

Clever Algorithms

This is a very interesting book for data mining enthusiasts. The author Jason Brownlee was leading the Hearst Challenge, before he decided to quit and focus on his book. Just browsed through the online version and it looks very impressive. Looking forward to reading it in details.

Clever Algorithms

--datamining_guy

Perone wants business data analyzed - Norwalk News - The Hour - Norwalk's Newspaper

A state senator in Connecticut introduces legislation which will require government to setup a business analytics division to make better decisions --- very interesting development. I am optimistic that we will see a lot more of this as everybody grapples with the amount of data they are collecting/have available and try to make sense of it all.

Perone wants business data analyzed - Norwalk News - The Hour - Norwalk's Newspaper

-- datamining_guy

Data Mining to Prevent Infections

A very interesting article on using data mining for better infection prevention. from the article it appears that a lot of useful data is collected but the Infection Prevention professionals are not able to use it effectively. A common theme across several industries/applications. However, am very thrilled at the diverse areas where analytics is being considered an useful ally:

Electronic Surveillance Systems Data Mining Can Yield Rich Results for Infection Prevention

-- datamining_guy

Location Based Marketing on New Year's Eve

This  New Year's Eve, saw a lot of  location based marketing initiatives. It has always made sense, but technology  has made it a lot more  easier now. Here are a couple of examples:

American Eagle teamed up with SCVNGR at Times Square to raise money for Big Brothers and Big Sisters.

Bravo channel also tied up its inaugural  New Year's Eve special with FourSquare.

Then there was the tie up between Time Square Alliance and Facebook Places, using a billboard in Times Square to urge attendees to log on to Facebook.

What is exciting is that for the first time, there was a concerted effort by several marketers to link the online and offline worlds around a special event/location.

Looking forward to how this plays out during the Super Bowl!

-- Datamining_guy