Just Throwing It Out There is a 2x/month newsletter that provides deep thoughts on shallow things: fashion, luxury, eCommerce and the future of retail. If you enjoy this issue, subscribe below:
At some point you may be blessed/cursed to experience an event that forces you to reconsider your deeply held beliefs. This happened to me circa 2015.
I was working at my first eCommerce job for a brand owned by a private equity firm. Their game plan for growing the online business profitably conformed to all of the “best practices” I had learned up until that point: invest in high ROAS performance marketing, use CRM technology to scale retention efforts, build this year’s plan based on last year’s plan.
That approach worked very well for the three years leading up to my start date in 2014. My boss created a daily sales forecast before the year kicked off, and she beat it consistently. It was easy for me to assume that the strategy and tactics were working and management was in complete control of the situation.
From my 26 year old perspective, company management and the PE guys were all very impressive people. They all had MBA’s; I went to fashion school and was pretty much completely self-taught. They spent their entire lives “on the path”; I spent the first half of my 20’s studying fashion design and seeking out open bars…I mean “finding myself”.
I hope you can tell where this is going by now. In 2015 the strategy that had been working so well stopped working. First gradually—our bi-annual sale missed expectations—and then suddenly, dramatically—we had trouble comping last year’s sales consistently.
This was the point where I expected the much more credentialed and experienced people around me to spring into action and solve the problem. But that wasn’t what happened. In fact, they struggled to effectively identify what was going wrong. The numbers were getting worse, but none of the metrics or tools the team was familiar with could tell us why, or what to do about it.
There was missing information and no one knew where to find it.
In an emergency situation you always regress to the level of your training, and no one had the right stuff.
And training or not, the most important thing you can do in an emergency is make the conscious decision not to panic. That’s not what happened either. Things got hilariously ugly—another story for another day.
It wasn’t hilarious at the time though. I knew that I did not want to be put in the same situation ever again.
I had to find that missing information. I had to develop the training to assess these kinds of situations so I could at least try to counter them effectively in the future.
While I was still in the thick of it, I was reverse-commuting from Manhattan into White Plains every day, about a 45 minute train ride each way. I started reading a lot about data analysis and decision making, hoping something would help point me in the right direction. Maybe someone on the internet had this same problem before and wrote about the solution.
One of the few books I read at the time that I still think about today is Information: The New Language of Science by Hans Christian von Bayer. I have no idea where I got the idea to purchase it. Maybe I literally Googled “how to find missing information”. But this book is filled with galaxy brain insight, and distills very complicated ideas into simple narratives.
The book is all about information: What is it? Why is it? How do you structure it? How do you measure the amount and the value of it? So instead of my usual brand-focused newsletter, I’m going to share some concepts on Information, and how I used them to eventually find my missing information.
The Value of Information
“My daughter announces that there is no school today, and I recall that it is Sunday, I know the probability of her announcement being true is 100 percent, so I ignore her. If it is an ordinary weekday, the probability that she is telling the truth is practically nil, and I don’t react either; but if I notice that it is snowing, the probability that she might actually be right rises to somewhere around 50 per cent and I take her very seriously. The value of the information she offers me depends on what I know, and what I know is expressed in the form of the probabilities that certain propositions are true.”
If the information I needed was missing, why? And who left it out?
The big data AI hype machine will tell you that data is the new oil, and big data will solve all of your problems. But information without context is close to worthless. Fortunately there is a formula you can use to assess the value of new information.
Don’t worry, this isn’t going to turn into a math lesson. Think about the cliche “where there’s smoke, there’s fire”. Many cliches survive through the centuries for a reason. Imagine that you really were unsure if there was a large fire burning a mile away. Then, the visual evidence of smoke rising from the location was introduced.
The formula simply flips the question: given that you’re seeing smoke, what is the probability that there is fire? And then it uses that probability to measure how much closer you are to certainty on the original fire question.
All this means is that information has a measurable value based on what you already know. You can quantify how much closer it gets you to the truth and usually assign a monetary value to it.
The dialogue in eCommerce and digital marketing is centered around the conversion—how do we increase conversion rates, and how do we identify the users most likely to convert and target them?
Digital marketing practitioners and digital advertising vendors are obsessed with identifying and tracking signals of purchase intent. Retargeting vendors tell you their algorithm can find the people most likely to convert, and prospecting vendors (including Facebook/Instagram) tell you their algorithm can find people most likely to purchase your product.
The algorithms work by taking a big pool of digital user activity (where I go online, what I click on) and mining it for information—some sequence of activities is a good predictor that a person is going to buy something, so we should target that person with an ad.
If you run the numbers through the formula, these purchase signals really do bring you closer to certainty that a single conversion will occur. Within a conversion event, these signals are high value information.
But these purchase intent signals are also relatively commoditized—everyone knows about them, they apply across brands and industries, and they’re easy to track. So smart advertising platforms have attempted to gobble up more and more digital real estate in the hope of acquiring rare signals that will set their offering apart.
Facebook bought Instagram for a billion dollars because they were purchasing information real estate—now, no one else has access to those signals. Whether or not the signals really predict purchasing behavior is another question entirely, but it seems to be working out well for them.
Conversion propensity signals are valuable to advertising marketplaces precisely because they are easy to package, measure, and market to businesses. Conversion is easy to see, and it’s easy to understand. But a single sale doesn’t reveal a lot of information about the probability that a retailer will “hit plan” aka achieve their financial goals.
And yet…a decade of research, analytical tools, best practices and frameworks for selling things online have been built around the individual conversion. Our tools were built by software engineers and product managers who never worked as retailers, and our line of sight was biased by what those people thought was important...for selling ads.
The missing information lies beyond the conversion.
“Most of the time, when we quite a probability, we are referring not to an easily replicable and thus quantifiable situation, but instead to a unique occurrence that could not possibly be repeated. Consider [this] example: An economist announces ‘There is a high probability that the economy will pick up in the next quarter.’ Does she have a hundred copies of the national economy in its current condition tucked away in a back room, so she can watch them evolve, and measure their performance three months from now? Clearly not.”
When traditional approaches to digital marketing and eCommerce optimization fail, why do they fail? Because we treat our business as an easily replicable and quantifiable situation. We fail to acknowledge or measure what is shifting beneath the surface of the things we choose to see.
In eCommerce, what is always changing beneath your feet? The merchandise assortment and its price, and the customer and his/her wants, needs and life situation.
What is the number one thing impacting conversion rate on your site? The merchandise you’re selling.
What is the primary factor that determines if a new customer will come back and purchase again? Whether or not you’re selling something they want to buy, or can afford to buy.
But in traditional approaches to digital marketing, we barely take the assortment into account at all. We rerun the same playbook over and over again whether we’re selling winter coats at 70% off or bathing suits at full price.
The complexity of the assortment amplifies uncertainty. If you’re selling one thing at the same price, the situation is easier to understand and forecast. If you’re selling multiple styles and categories with a complex markdown/promo calendar and an assortment that turns over multiple times per year, you might as well be predicting the weather.
The composition of your customer file also amplifies uncertainty. If it’s your first year in business, returning customers do not contribute that much to your overall sales, so you really only have to manage acquisition.
If you’re a mature business, some significant percent of your sales are coming from returning customers. Each of those customers is essentially in a box—90 out of 100 times, they would never spend more than a certain amount for a certain item, or purchase outside of a certain product category.
If you don’t know the “boxes” your customer file sits in, you may fail to provide them with enough product that checks their boxes. And then your conversion rate goes down.
A strategy that relies on moving customers out of their box—like acquiring lots of customers during a promo and then “leveling them up” to full price—is not going to scale.
Think about it: would you ever spend $20 on a cup of coffee? Are you going to suddenly start buying smooth peanut butter if you really like chunky, just because you saw a sick Instagram UGC story ad? As marketers in the post-shared traffic era, we need to generate traffic and sell the dream. So strategies that work with human nature are essential in padding out the bottom line.
As digital marketing and eCommerce practitioners, we should be measuring the impact of the merchandise on the conversion rate. If your marketing gets 10% more effective but the assortment gets 10% less appealing, what is the net result going to be? Nothing.
At face value, this sounds like I’m asking you to measure something that is fundamentally subjective. But we need to apply the same approach as the digital media vendors we love so much—we need to find meaningful signals. We need to find the hidden information that will bring us close enough to the truth to make meaningful decisions.
I’m not going to go into detail about how to do this. I don’t give away all my secrets for free on here. But if you’re interested, this is a good place to start.
Objective reality is an illusion we construct for our own comfort. The best we can do…is to create a coherent model of the world that reproduces its measured properties without claiming to describe reliably what actually is.
Merchants, investors and gamblers are all looking for asymmetric information—information that brings a decision closer to certainty that no one else has bothered to decode. Merchants seek to decode consumer trends, investors seek to decode trends in the prices of assets, and gamblers seek out cause and effect relationships in whatever outcome they’re betting on.
Caveat here—only applies to good merchants, investors and gamblers, not dopamine chasers.
Conversion signals = commoditized information. Your assortment, your customer file, your channels, and how they interact = asymmetric information. Because it’s unique to you.
What I Wrote Since Last Time
People I work with who are reading this will be thrilled to hear I have been working too hard to write anything but the newsletter 😭
It’s been almost a year since I started writing online. Enjoy the piece that started it all.
And One More Story I Think You’ll Like
In case you couldn’t tell, I have been pondering the big questions lately. Here is a great Twitter thread on the future of retail:
And here is my new guiding life philosophy: