🤖 Do Androids Dream of Conversion Rates?

Let’s get some buzzword definitions out of the way, so we can really explore the studio space:

Machine learning (“ML”) is an application of AI, wherein a system is provided the ability to automatically learn and improve from large datasets, without being explicitly programmed.

Nailed it! Now let’s make music.

🕵️‍♀️ ML is standard issue

Think about your common digital experiences for a second - looking through Netflix, Googling something, firing up your Spotify or Amazon app, etc. Each of these experiences is curated to you. Your background, interests, and tastes have been measured, over time and via multiple devices, then fed into predictive systems (along with millions and millions of other data points), in order to inform and deliver a highly-relevant experience to you.

Relevance, in this instance, describes an experience that encourages a specific behavior; broad strokes, these targeted behaviors are buying more, listening more, and watching more.

~Rapid Fire Case Study~

You and your sister are bored during a long car ride. You pull out your phones, navigate to Google, and type in “Ross.”

Here’s the thing, though: you are a low-cost fashionista looking for bargain retail deals, while your sister is a diehard fan of Ross Geller from Friends. You want to find Ross Dress for Less stores nearby; she’s just looking for David Schwimmer memes.

It’s a good thing, then, that Google has invested massive amounts of money and effort into its machine learning, because they are hell-bent on delivering highly-personalized search results. Personalized = more relevant; more relevant = you find utility. More utility = you’ll keep Googling.

Without machine-learning-based personalization, Google might show Ross retail locations to both of you, or to everyone that types in that specific keyword. That would leave lots of Ross Geller fans out in the cold - and that’s the last thing we want to do.

🖥 Machine learning in eCommerce

ML is a cornerstone of eCommerce strategy, because it allows a digital retailer to feed huge behavioral datasets into a system, then learn to tailor a shopping experience to each individual shopper. Personalization of the journey is the holy grail.

Which components of the eComm funnel are influenced / augmented by ML?

Search

Search is the the bedrock of the internet. Google’s behemoth business was built atop their search algorithm - and when we say SEO, we really mean ‘Google optimization.’ Entire industries are built around enhancing website content to accommodate algorithms, because people often start their journey (shopping, information, or otherwise) with a simple search.

Setting SEO aside for a moment, because that deserves an entire post, let’s focus on on-site search - in other words, how you are yielding product results based on what a human is typing into the search bar on your website. 

Query Suggestions as Guidance

These results are based on ML, of course - millions of data points fed into a system that is learning which results are most relevant based on inputs. What have people ultimately 1) searched for, 2) navigated to, and 3) purchased when they start typing “Ross Gell…”? Turns out, a lot of people are buying Ross Geller funko pops. I didn’t know what funko pops were, so here:

The magic of ML is that it is predictive of intent; in other words, systems are trying to predict - rapidly and with a high degree of confidence - what you’ll do next, and what you actually want.

Consider, also, that every search performed by a user is a massively important and influential piece of data; people are telling you what they want. If you don’t stock Ross Geller Funko Pops and lots of customers are searching for them, shame on you.

Result relevance

Individual customers search in a very unique way. Some search queries are riddled with typos; some search queries focus on keywords; some search queries are exact product names.

On-site eComm search funnels must focus on determining intent and yielding appropriate and relevant results, even with the noise and nuance of individual searchers.

Bundles and Up-sells

What products do customers tend to purchase together? By feeding a machine learning algorithm with data on thousands of transactions, eCommerce systems can predict what add-ons to show you as you add things to the shopping cart.

Relevant item suggestions = happier you. Happier you = you buy things. You buy things = happy eComm provider.

Amazon sets the gold standard for suggested up-sells, as shown below.

Advertising and “Shifting Shelves”

eCommerce providers usually advertise throughout the web experience - natively via sponsored sections, or more notoriously through banner ads, etc. What ads should we show, when?

Same concept of relevance comes into play. Who is the shopper on an individual basis? What have they purchased before? What have they searched for before? What proclivities have they displayed based on third party information about them?

Collect data, inform algorithms with that data, then let the robots take a guess on what ad has the highest likelihood of encouraging the optimal behavior. ML!

Ads are one thing - but what about the way in which we merchandise our digital storefront? Should everyone see the same layout?

No! It’s 2021. Each individual should be served up a personalized experience based on what they like and what they are most likely to be delighted by. If they’re delighted, they buy more. In that vein, products should be merchandised on a session by session basis.

🌱 Machine learning in cannabis eCommerce

Personalizing an eCommerce experience is based on machine learning - and, to be explicit, the ultimate goal is maximizing conversion rate, average basket size, and return trips.

Everything we discussed is necessary in cannabis eCommerce, where customers are “voting” (performing searches) en masse across thousands of legal retailers. Most systems are egregiously antiquated, though - essentially spreadsheets thrown online with no personalization or smart search - and the conversion rates for those systems is predictably low.

All of the sophisticated ML components in traditional verticals are available, now, to smart cannabis retailers that choose the right software provider:

📚 tl;dr

• Machine learning (“ML”) is an application of AI, wherein a system is provided the ability to automatically learn and improve from large datasets, without being explicitly programmed

• ML is a cornerstone of eCommerce strategy, because it allows a digital retailer to feed huge behavioral datasets into a system, then learn to tailor a shopping experience to each individual shopper

• Personalization of the shopping journey is the holy grail

• Ross Geller Funko Pops are popular

• It is Tuesday