December 22, 2019

Progress

If you don't like running through the history books, take today off ... heck, take the next two weeks off and come back in early January.


Now, if you noticed, I ended last week's discussion about catalog circulation with discussions about five-year ad-investment simulations. This would be the part of the program where my comments are met with groans. But progress happens. And there will be a time when you are running simulations of your business, so that you can make decisions based on the outcome of myriad simulation runs and not based on random strategy changes implemented in real time. It's unavoidable, progress is coming.

Progress doesn't always happen on the timeline we'd like it to. 

When I started working at The Garst Seed Company in 1988, I was introduced to the concepts of Experimental Design. Sure, I dabbled in it a bit in college, but in the real world we conducted a ton of experiments. We had a whole "tournament" set up each year ... various corn/sorghum hybrids were tested against our existing "winners". If a hybrid showed promise, it advanced to another round of experiments against existing "winners" and promising new hybrids. Eventually, a handful of hybrids achieved commercial status ... they passed the testing gauntlet and were commercially viable.

Our testing program was the end result of progress made in the 1930s and 1940s. We leveraged the work of Snedecor and Cochran ... I still own their Statistical Methods book from 1938 ... yes, 1938. Their ideas were made famous at Kansas State University and Iowa State University, and were fully embraced by the statisticians I worked with in 1988.

By December 1990 I was a newly minted Statistical Analyst at Lands' End. At first I created logistic regression models and ordinary least squares regression models ... borrowed from my predecessors, then fully implemented from my own work ... with techniques borrowed from Norm Draper. Folks leveraging "machine learning" today are just building upon stuff that Lands' End (and many others) did in the mid-1980s. Progress.

In 1991 we were wondering why we kept adding catalogs to the mail plan and sales didn't increase linearly as catalogs were added. This is where the work of Snedecor and Cochran was applied to catalog marketing. I'm sure we weren't the first to do this, but we took the concepts far, probably farther than most in the industry at that time. Over the next four years we learned a ton about the true (incremental) impact of catalog marketing by employing the testing tools that Snedecor and Cochran applied nearly sixty years earlier. Progress.

It was no different at Eddie Bauer. We used the techniques described by Snedecor and Cochran to test the effectiveness of email marketing, to test the effectiveness of discounts/promotions, to create matched market tests to measure television advertising effectiveness and to test the appropriate price for a pair of blue jeans. Progress.

It was no different at Nordstrom. We used testing methodologies to demonstrate that we could eliminate $36,000,000 in catalog ad cost and move into the future. I fielded the calls from trade journalists, they thought we were insane ... and they printed their critical thoughts. But we had testing data (they didn't). We ran simulations (yes) to demonstrate that the business would be healthy in the future. We knew more than trade journalists and industry experts knew. We had progress on our side.

And we were proven right. Online/Phone sales increased even though $36,000,000 of ad cost were removed from the ecosystem. Hint - that results in A LOT of profit!!

In the Spring of 2007 I spoke at the Spring NEMOA Conference in Boston. I was also asked to participate on a panel. I was asked about the future of catalog marketing, and in particular, how catalogs would be measured in the future. I shared the fact that catalog key-codes / source-codes were already irrelevant, and that the future was all about experimental designs ... stuff that had already been employed in other industries for 70 years ... and that the results of these tests would tell us whether catalogs possessed any meaning in a digital world. I told the audience that we executed these ideas at Nordstrom, grew online/phone sales, and greatly increased profitability.

The audience laughed.

They heartily laughed.

If you've spoken in front of an audience, you know when the audience thinks you are a moron. You see the condescending looks, the glares, the dismissals. At that moment, the full menu of industry derision was on display. It's always amazing to gain hard-earned knowledge and have folks who haven't earned the knowledge laugh at you and criticize you. Ha!

Progress doesn't happen on your timeline.

You may not like the fact that your future is a fusion of Experimental Design (i.e. tests) and Simulations. I've been running simulations (not complex ones, but simulations nonetheless) since the mid-late 1990s at Eddie Bauer. The insights offered by simulations enable the Business Leader to side-step years of trial-and-error in the real world.

I mean, if you simply A/B test your catalogs, A/B test your email campaigns, test any spend with Google, and test any spend with Facebook (and if you are a cataloger you A/B test your co-op investments as well ... right ... RIGHT??) ... if you do that and then build a simulation tool around your investment levels, well, you're going to know EXACTLY what you need to do in the future to make sure your business achieves the potential your Ownership Team or Shareholders or CEO or Executive Team (or you) believe it should have. You'll know exactly how much you have to spend, and you'll know exactly where to spend it, to achieve your potential.

I know, I know, you are going to ask around ... "is anybody else doing this???" ... and you'll feel reassured when you learn almost nobody else is combining Experimental Design with Simulations. You'll say, "see, I'm not crazy, that silly blogger, he's the crazy one."

This is how progress works.

Remember the NEMOA reference? Source-Codes / Key-Codes are meaningless as we move into 2020. Mailing decisions via the "organic percentage" and mail/holdout tests are common now. 

Progress happened.

You are going to leverage a combination of Experimental Design (testing) and Simulations to figure out what direction to take your business in. It's coming. This is how progress works. 

Heck, just A/B test your way into understanding your organic percentage ... do the tests we executed at Lands' End ... in 1991. The outcome of the tests is more valuable today than it was back then. Go learn something!!! Have fun!!! Make progress.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.

A Post-Tropical Cyclone

A tropical system doesn't stay a tropical system forever. Take Milton for instance. A Category 5 hurricane at two different periods earl...