CASE STUDY
How Lights Online increased CVR |READ NOW
How Sentiment Based Shopping can revolutionize eCommerce.
By
Michael J. Prichard
May 18, 2022

I began my career in e-commerce, believing the best product recommendations would win. Personalization was about turning shoppers (aka human beings) into a point on a vector drawn by a machine learning algorithm using historical data of others’ behavior, and that this was the way to maximize conversion rates. Hence, I fell into the trap of trying to create the better “AI” to beat out the other “AI” that was designed to better the previous “AI” and so on and on.

Here is what I have found.

We made it to “good enough”

When compared to nothing, any recommendation (including random) will achieve alpha. Historical data-driven recommenders such as co-bought and co-viewed can achieve more alpha, maybe, as compared to random. Add some sort of profiling (until Apple and Google completely kill pixels and cookies) and we quickly reach a state of diminishing returns.


Despite the marketing hype from personalization companies claiming their algorithms are better than their competitors, the underlying challenges of predicting what people want is limited to the same data math and, hence, the same guesses; give or take.

Don’t believe the hype

Now you are thinking, “OK, boomer! <insert some company>’s AI says they can get me 100% CVR increase because they have super genius data science engineers on the team.”

I guess, maybe.

However, ask yourself, how much more would it outperform a random set of recommendations? Maybe a little more? A bit more? None? At some point, good enough is good enough.

We know product recommendations work

The stat I see over and over is that 35% of Amazon’s revenue is generated by its recommendation engine. You would be hard-pressed to find an e-commerce merchant with the data, the best minds, and, frankly, the money to apply these resources to solving product recommendations once and for all. Why is it that even they can’t make it perfect? For example, Amazon still shows me vacuum cleaners for weeks after I bought a vacuum cleaner.

Where does this leave you?

We all know that to compete in today’s e-commerce landscape, a brand needs to develop a close relationship with their customer. Typically, especially in apparel or luxury goods, brands that stand out tell their story well (and authentically) and truly work to know their customer. Customers are unique. They should not number on a spreadsheet but rather repeat customers that buy into your brand, story, products, and all.

Ask your customer what they like and dislike

At Skafos, we are on a mission to develop and nurture humanity in digital shopping through conversation, interaction, and joy. The keywords in that mission is “humanity” and “conversation”. Shoppers are people just like you and me. We are social creatures who exchange information through conversing and emotional queues. Many times emotional queues are not so obvious and motivations as to why we like or dislike something are easily transferable to words. In other words, “I know it when I see it”.

Engaging customers directly is great when you are in a retail store with a few people, but it’s practically impossible when you have hundreds to thousands of shoppers hitting your online store daily.

This is where the next revolution in e-commerce will play out. How we engage our customers and drive conversation at a more personal level but at scale. The only way to do this (aside from hiring hundreds of customer service reps) is to re-imagine your online shopping experiences in that you are asking your customer what they like and dislike and bringing them delight and joy as they interact with your brand.

Whatcha’ ya talking about, Michael?

You ask them what they like and dislike; literally.

We learned that user interactions such as these empower shoppers to quickly sort products in the order of what they are most likely to buy. We are able to determine context (why they are there) and sentiment (how they feel about products) which we then mix with traditional recommenders to drive a very unique personalization experience.

We call this Sentiment Based Shopping and we believe it is the way of the future for e-commerce. More dynamic, rich experiences that engage with your customers that really require very little data about them.

This is the way

Considering that privacy regulations are getting tighter and pixels and cookies are being deprecated, your third-party data solutions are going to lose fidelity. The way to really understand your customer is to build zero-party data for yourself and use that to personalize experiences directly for your shopper. The only way that it can be done is to ask them. Sentiment Based Shopping provides another way to build on this knowledge, and it does it while delighting your customer.

Is there anything more gratifying than that?

Actually, there may be. If you read this far, first, thank you. Writing is not the easiest thing for me, so I appreciate you appreciating my work. Second, as a token (no pun intended) of my appreciation, I want to give you a free Skafos NFT. Hit us up at hello@skafos.ai and we will make it happen. Cool, right?!

Happy conversions!