Analytic Strategy Partners

Improve your analytic operations and refine your analytic strategy

News

Claude Hopkins: An Early Advocate of Test-Measure-Refine

by

Claude Hopkins, Scientific Advertising, 1966 edition.

Claude Hopkins (1866-1932) wrote a book in 1923 called Scientific Advertising that was only 95 pages long but contains a large number of important insights about analytics that are just as available today as they were in 1923. 

I have the 1966 edition by Crown Publishers that contains an introduction by David Ogilvy, who writes: “Nobody, at any level, should be allowed to have anything to do with advertising until he has read this book seven times.”

As Ogilvy points out, not everything in the book has stood the test of time, and some parts of the book were misguided when the book was published in 1923, but the majority of his insights are still worth following.

Hopkins was one of the earliest advocates of the Test-Measure-Refine (TMR) Loop in advertising.  He was a strong advocate of coupons since coupons are not only an effective technique themselves but could be keyed, and the different keys could be used in the TMR Loop to measure outcomes of different variants of the ads.  In 1923, the TMR mail-based loop required weeks, today we can do it in minutes to hours. It’s also important to note that the Test-Measure-Refine loop is also just as important in building analytic models as it is in advertising.

  • [I spend] far more time on headlines than on writing. [I often spend] hours on a single headline [page 34].
  • Almost any question can be answered cheaply, quickly and finally, by a test campaign. And that’s the way to answer them — not by arguments around a table [page 77].
  • It is not uncommon for a change in headlines to multiply returns from five to ten times over [page 35].
  • The best ads ask no one to buy.  This is useless.  Often they do not quote a price. … They offer wanted information.  They cite advantages to users [page 23].

All quotes are from: Claude Hopkins, Scientific Advertising, Crown Publishers, Inc., 1966.

Hopkins wrote outstanding ad copy, including the slogan that Schlitz is the “beer that made Milwaukee famous.” His advertising and branding proficiency also created the brand equity for Palmolive soap and Pepsodent toothpaste.

In addition to stressing the fundamental importance of testing copy, Hopkins also pioneered the use of split testing, using coupons to provide samples and to track outcomes, loyalty programs, brand images, and product demonstrations in advertising.  He was always testing and experimenting with new advertising ideas, which seems to be the reason why he titled his 1923 book Scientific Advertising.

Hopkins expertise and ability to execute was well appreciated. Indeed, he was hired in 1908 at the age of 41 to write ad copy for Lord & Thomas,  at a salary of $185,000, the equivalent of over $5 million in 2019.  

Hopkins influence and the influence of Lord & Thomas is still present even today. Lord & Thomas was founded in Chicago in 1873, and grew under the leadership of Albert Lasker to become in 1908 one of the largest general advertising firms in the US. When Lasker retired in 1943, he sold the firm to the executives who ran its three offices: Emerson Foote in New York, Fairfax Cone in Chicago, and Don Belding in Los Angeles and the firm became known as Foote, Cone and Belding. It’s still around today as FCB and still headquartered in Chicago.

The post is adapted from Chapter 4 of my book The Strategy and Practice of Analytics, copyright 2020.

Using the Analytic Diamond Framework to Manage Your Analytic Projects

by

One of the most useful frameworks that I have found looks at a data science, machine learning or analytics project from four different perspectives:

  • Analytic modeling. The analytic model team consists of statisticians and data scientists that develop, test and validate the statistical or machine learning model.

  • Analytic infrastructure. The analytic infrastructure team consists of data engineers and those knowledgeable about databases, data science frameworks, enterprise IT and related areas and are responsible for the analytic and IT infrastructure required for managing the data required to build the models and managing the systems required to deploy the model into products, services and internal operational processes.

  • Analytic operations. The analytic operations team deploys the model into a product, service or internal operational process. Importantly, it is also responsible for integrating various actions associated with the outputs of the model and improving the actions so that they bring the required value. For example, a model might output a score from 1 to 1000, but depending upon the value of the score, no action may be taken, a standard offer may used, or a special offer may be used.

  • Analytic strategy. The analytic strategy team selects appropriate analytic opportunities and works with the product team to develop an appropriate business model for the product or service developed.

I’ll be talking at Predictive Analytic World about analytic operations:

Title: What are Analytic Operations and What Are Some Frameworks to Improve Them?

Abstract: There is a lot of information and best practices available so data scientists can build analytic models, but much less about how analytic models can best be integrated into a company’s products, services or operations, which we call analytic operations. We describe three frameworks so that a company or organization can improve its analytic operations and explain the frameworks using case studies.

Event: Predictive Analytics World, Las Vegas, June 19, 2019.

Running a Successful Analytic Project

by

There is a lot of information available about best practices so data scientists can build analytic models, but much less about how to manage analytic projects and analytic teams and how to coordinate with other departments within a company so that analytic models can be deployed and integrated into a company’s products, services or operations.

On May 23, 2019, I gave a talk at The Data Science Conference (TDSC) in Boston about managing analytic projects and their teams.

I focused on two themes:

  • How to quantify the risk of an analytic project and work to reduce it over time.
  • Understanding how to recruit a balanced data science team, which makes it more likely that your project will be successful.

Improving Your Analytic Operations

by

I gave a talk on analytic operations at The Data Science Conference (TDSC) that took place on September 20-21, 2018 in Chicago.

One of the topics that I covered was the Develop, Deploy and Extract (DDE) Framework for analytic operations. This is a framework that covers the life cycle of an analytic model that spans the development of the model by data scientists, the deployment of the model into an organizations products, services or internal operations by data engineers, and the quantification and management of the value created by the model by the product owners.

Title: What are Analytic Operations and How Can You Improve Them?

Abstract: There is a lot of information available about best practices so data scientists can build analytic models, but much less about how analytic models can best be integrated into a company’s products, services or operations, which we call analytic operations. We describe four frameworks so that a company or organization can improve its analytic operations and explain the frameworks using case studies.

Event: The Data Science Conference, Chicago, September 20, 2018.