Home > News & Events > Exponentially Better: The Math Behind Our Shopping Platform š¢
Eugenia Cheng tackles combinatorics in The Wall St. Journal and indirectly helps me bury the axe with my old hosting company.
In 2007, I drew the process for sharing products between brands and retailers on a bar napkin, such as in this example. In this example, a brand is Syncing three products with four of its authorized retailers.
Moranās restaurant in New York City where I drew the Syncing service sketch on a bar napkin and called my hosting company. (Sadly, the restaurant, which had four fireplaces, is now closed.)
In 2007, technology was primitive as evidenced by the small size of mobile phone screens.
In a 2008 advertisement for the Syncing service, we can see the nine partner brands offered and three retail clients. Today, we have112 partner brands and 1,450 retail clients.
A merchantās web page with Synced product data. The Syncing relies on trillions of calculations to output the right brands and products on the right merchantās Online Store.
Retailers can Sync products with 112 partner brands. To compare this to the The Wall St. Journal article about the burger restaurant, itās as if each brand were a burger topping. Each brand choice exponentially increases the combinations a merchant may offer. Brand X may be pickles, brand Y is bacon, etc.
Merchants report selling up to 18% more due to the Syncing service.
Exponentially Better: The Math Behind Our Shopping Platform š¢
How a drawing on a bar napkin led to a software platform that processes trillions of calculations
In 2007, while sitting at the pub Moran’s with a beer, I drew a sketch on a napkin that would change our business. I drew a software model that would allow an indie store to instantly show (and sell) thousands of products online—without any labor. (The secret sauce: The respective brand would do the product update work for the merchants.) The immediate goal was to help my two retailers more efficiently load products onto their websites (and save me the boredom of doing it). In my excitement at this discovery, I called the owner of my hosting company from the bar. I pitched him my plan. He said it wouldn’t work because there would be too many possible combinations of products, and this would result in an impossible server load. I hung up and felt disappointed. And…I pursued my napkin sketch anyhow.
It worked. Today, we help 1,450 indie merchants each offer up to 3,000 brands. Of these 3,000 brands, 112 are premium partner brands, including Baccarat, Le Creuset, and Versace, that Sync 70,000 products with their respective authorized stores. The daily resulting combinations are in the trillions of trillions.
256k Reasons I Was Wrong
But this isn’t me saying I was right. The hosting company owner was: Mathematically right. Let me give credit where credit is due.
The reason my hosting company told me it wouldn’t work can be explained by a field of math called combinatorics covered in today’s The Wall St. Journal by Eugenia Cheng. In the article, we learn about an airport burger restaurant that promises 1 million burger combinations. Ms. Cheng looks at the math behind the restaurant’s claim and finds the restaurant would need to provide 20 toppings to offer 1 million combinations. Her math: 2 to the 20th power.
2²ā° = 2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2 = 1.048576 (e+6)
My hosting company said “no” because it estimated the number of product combinations a store may offer via our platform. The number of combinations from a given list of brands would become exponentially larger as we added more Syncing brands. To calculate the number of combinations, we use 2 to the n power, where “n” is the possible number of brands. If a merchant just offered 5 brands (and their products), that’s 2 to the 5th power, which is 32. The problem: Each additional brand exponentially created more server drag. The real-world byproduct of this is super-slow page loads. (What good is an online store with 200 brands and 100k products if it takes 20 seconds for your shopping cart to load?)
In 2007, each of our stores could potentially Sync products from up to 9 of our partner brands. That’s 2 to the 9th power.
2 to the 9th power = 2×2×2×2×2×2×2×2×2 = 512
Each of these brands had about 500 products, which is 256,000 products (512 x 500). The hosting company was saying “no” to 256k combinations running on each client’s site. At the time, websites and their processing were in their infancy, and 256k combos seemed a bit daunting for a small hosting company. To put the state of technology in context, in 2007 the first iPhone came out, and it had a postage stamp-sized screen compared to today’s phones and could not record video. Just a year earlier, Facebook had launched for the general public.
Moore is More
Thankfully, servers today can quickly deliver much bigger calculations with ease. While at Moran’s we just had 9 partner brands, today we have 112. That is 2 to the 112th power.
2¹¹² = 5.1922969 (e+33)
…That’s 5 with 33 numbers to the right of it.
On top of having 112 partner brands, today there are also other "multiplying" factors:
- Instead of two retail clients, we have 1,450 retailers. The server has to multiply that massive number by 1,450 clients.
- Each of those 112 brands has, on average, 500 products that can be turned on or off.
- In addition to our 112 partner brands, clients can also add up to 3,000 other brands to their account.
In sum, our server is conducting combinations that are trillions of trillions.
So, how did we get our platform to accomplish all this processing? How do we offer trillions of product combinations AND still deliver a shopping experience that is faster than Macy’s? First, we started with just two retailers—not the 1,450 we have today. That small base gave us a manageable size to build our beta platform and proof of concept. We then honed our code and process over 15 years as more clients joined us. We also added lots of hardware. Where our ingenuity ends is where Moore’s Law picks up the slack.
The problem isn’t solved. Each week, our clients add more products—thereby increasing exponentially the possible combinations and server load. And, each week, our team tweaks the code and platform to ensure the web page speed stays the same or improves. There is an equilibrium that takes place with more clients joining and adding more products and us recalibrating and adding more power.
Syncing Stores Sell 18% More
One may now ask, “But does offering exponentially more products truly save stores money—and help stores sell more?” Syncing products saves The Ivy House, a tabletop shop in Dallas, TX, $6k per month on labor. This adds up: Each year, we save indie stores $3.5m on website labor—money they can put towards their businesses, higher rents, and fighting Amazon. On top of saving money, our retailers, like The Ivy House, report that Syncing products help them sell 18% more.
Our team works hard and deserves a break: I should take the team out to lunch at a burger spot. If we’re kind, we won’t order too many topping combinations. And I’ll encourage the team to draw on their burger napkins.
Homework
Consider how many total combinations your business offers for a service or product. Can this number be promoted as an advantage? For example, if you offer a service or product with 20 choices, you may tout that you offer 1 million possible combinations.
……….
Read The Wall St. Journal article about combinations:
https://www.wsj.com/arts-culture/food-cooking/an-ad-boasting-a-million-burger-and-topping-combinations-seemed-ridiculous-actually-its-an-understatement-cca5acd7?st=8ybqwi1ovfhf8pv&reflink=share_mobilewebshare
Tags:
WSJ Eugenia Cheng Moranās syncing service
View Post on Shop Local
