Alexa, order me a new t-shirt

I don’t know if you ever tried buying t-shirts online, but it is quite overwhelming. On alone there are currently 4,270,345 products to choose from.

Yes, over 4 million.

They all look the same, maybe slightly different, and have dozens of reviews attached to them.

In a great piece Lists are the new search Benedict Evans wrote:

Showing every SKU, of course, is exactly the Amazon approach - ‘the everything store’, and it works well for some categories, and especially when you know exactly what you want. But knowing what you want is not necessarily the starting point - that’s what needs to happen along the funnel. Amazon’s relative weakness at curation, discovery and recommendation (I’ve seen data suggesting the recommendation platform is only 1/4 of its books sales) is, I think, a big reason why, after 25 years of ruthless and relentless execution, it’s still only got to 25% of the print books market in the UK and USA. A bookshop (or any shop) is, yes, the end-point to a logistics system, but a good bookshop is primarily a discovery platform. That is, it’s more about the tables than the shelves. And the tables are lists, not inventory.

The issue for big catalogs like Amazon is how to represent them. Go the curated route and thus omit most of the selection, or go the list route and have pages after pages of products. Both of these work differently in different categories, and neither of them are ideal.

I want to be able to ask Alexa, the AI of Amazon’s Echo devices, to find me a t-shirt and bring it to me. It knows my size, it knows what I bought before, it knows what others bought, etc. It knows it all. So what I’m going to explore today is how we are going to get to the future of voice UI for retail.

Mobile is already changing our behavior

Most of the sales on Amazon are currently done via the website. However mobile e-commerce is rising fast. The difference in buying on mobile is that customers arguably spend less time researching products or their prices. It’s simply not as easy to do so as on a big desktop screen. So with less comparison shopping, and less reading of long reviews, customers buy based on first impressions more and only pay attention to the first few search results.

The limitations of a small screen and short attention span mean that customers buy differently. Especially when considering the influence social media has in this.

This meant that mobile experience has to be optimized differently too. Mobile shopping often has less information, and more focus on key points - price, title, images, etc. If it involves reviews, then it features a few best ones, because no one is going to read all of them.

For some websites like eBay this has been a major pain point. Nothing about the desktop experience works on mobile. And that’s why their mobile site/app is completely different. One of the big original features on eBay is the ability to use custom HTML and CSS to create a unique product description. Most sellers are thus using complex templates, turning the description section into a fully fledged website. And yet none of this works on mobile. Both because it doesn’t fit on a screen, and because no one has time for it. So now when you go to any product on the eBay app, the description is hidden behind an extra button click, and once clicked it reveals the same experience like browsing the web before responsive web came out. Zoomed out website filling the screen. It simply does not work.

Mobile is showing that retail is possible with less information, and less time spent before ordering. In some countries, this has also grown into text-based chat systems, which allow ordering products without browsing anything. If you compare what was the UI and the process of buying on desktop, and what it is now on mobile, you’ll find some major changes. I think those changes will continue to evolve.

Brand value

Consumer city

Some people argue that there is a decrease in reliance on brands. Sure, some categories like fashion, are all about brands, but for a lot of commodities the brand is increasingly less important. I think many overestimate how many products rely on the brand to sell them.

In the age of online retail which brought ratings and reviews, the rating score is a much better tell for just how good the product is. This is allowing new brands to launch, focusing on their own niche, as the customer focus shifts from using the brand as a quality indicator, to recommendations based on reviews.

I think in the future brands are going to lose even more importance. There is a measurable decrease in perceived brand value for most products. Instead a single product is holding its own weight, with reviews, ratings, and social networks defining its true value. That’s why I think the successful brands in the future are going to be less focused on brand marketing, but instead on communicating what each product is about.

And if the brand continues to stay important, the customers are not necessarily going to stay loyal. Brand itself will carry value, but customers will shift jump brands freely.

It is very hard for a brand to attract visitors to their own website. There is Amazon, but there is also all the small aggregators focusing on a particular niche. For example for clothing there are countless apps focused on curation. In those, the brand itself is not as important.


Amazon, more than any other e-commerce site, depends on trust.

One part of it is the trust in Amazon’s ability to source the best possible price. BloomReach estimates that “55% of consumers turn to Amazon first when searching for products online”. The reason why so many people go straight to Amazon, is because they have learned that price shopping is pointless - Amazon will have the best deal. Of course this is not always the case, but that’s the perception Amazon really wants.

Second, Amazon wants people to trust them on suggesting the best products. When someone searches for “red napkins” on Amazon, it’s up to the algorithm to surface the best deal, for the best product, which fits the closest to what the customer wanted. If Amazon is able to do this, then customers don’t need to look at other products, or compare them to each other - they can pick whatever Amazon said is the one.

These two trust points - best price and best product - are fundamental to Amazon’s future.

Today there is still a lot of noise in the marketplace, with products often changing price hourly. But the hope is that as the marketplace grows and as the number of competitors increases, there is going to come a point when the system reaches a point of balance. There is going to be so many retailers supplying products, that from the customer perspective the best possible deal is going to surface.

This is both Amazon’s big focus, and the key goal of marketplaces - make the competition work for customers. It’s challenging, and increasingly complicated because of counterfeit problems, but if we assume that it is possible, then Amazon will eventually get to the perfect trust point.

Amazon almost ruined itself

Circle of shoes

Launching new products was always hard, especially on a marketplace with hundreds of millions of products. On Amazon it is not good enough to have a great product - there are many products just as good. Instead a product needs to get trending to start ranking higher. And the only way to achieve that is to have great quality pictures and description, but also reviews confirming how good it is.

Since late 2014 sellers on the Amazon marketplace have been asking their friends and family to leave reviews on Amazon in exchange for a free product.

So by sending a product to friends and family sellers figured out they could kickstart a product by getting a few dozen positive reviews. This was often enough to make the difference and kickstart natural sales, as the product would start to rank higher because of its well-reviewed status. This was working so well as a product launch technique that as it often happens, sellers got greedy and started abusing it.

For the last couple years private-labeling has become a common practice. Find a generic product in China, for example a USB charger, come up with a brand for it, put the brand sticker on the product, and in turn have a unique product not offered by anyone else. Selling Marvel toys sounds great because of how popular they are, but thousands of retailers sell them making competition often too hard to beat. By having a product no one else has, even if all it differs by is a brand, is a way to avoid competition. All one has to do is to make it popular on Amazon - more popular than dozens of other USB chargers. Other techniques aside, getting dozens of 5 star reviews for it was the golden ticket.

By 2016 everyone knew about this technique. Thousands of guides were written about it, and tens of thousands of products were added to Amazon every day in the hope of launching them quickly with the help of these reviews. It went from being friends and family, to communities on Facebook and independent websites offering a service of getting reviews for any product. It became a well-oiled machine which worked effectively and predictably.

But there was a big problem - customers started to notice this. More and more products had overly-positive reviews, clearly biased to be positive. While technically those in-exchange reviews could be negative too, the chance of someone leaving one for a free product was slim. More often than not reviewers were positive in hopes to receive more products in the future. Amazon reviews became a stream of 5-star reviews for products, even if many of them were ok at best.

Amazon finally had enough of this, and couple months ago put a stop to incentivized reviews. It became too easy to capitalize on Amazon’s traffic to launch weak products. Recently they even went as far as to remove a lot of previously-left biased reviews.

If they hadn’t done this, this could have ruined the trust in reviews, and thus the ability to build a future of e-commerce relying on them.

To me, this change is much more important than it might appear. It revolves around the same concept of trust. Once customers do not trust the reviews and the star rating, then they have no value to anyone. But if they do, then it’s a big deal. It’s all about trust.

The future of voice UI

Alexa, order me a new t-shirt

With the recently introduced Alexa AI available in Echo devices buying is taken one step further - there is no reading of reviews or comparison shopping at all. When a customer asks “Alexa, order an iPhone replacement cable” the expectation is for Amazon to pick the best one - considering price, shipping time, reviews and description - and order it.

In this trend towards less attention being spent analyzing differences between various products, and by default going with what Amazon search suggests the importance of quality data is increasingly key. If a product can cheat its way to having the best reviews and thus get picked up by subjective algorithms, then it will get bought for months before actual customers leave enough negative reviews to offset the system.

If you think this never going to happen, the first step was the Amazon Dash Button. That physical button allows you to re-order one particular product at ease. Very useful for consumables like toilet paper, soap, baby wipes, etc. It does all the things I talked about, but is attached to a single product. Now remove that attachment, and make it accept any product.

This blind-buying relies on all the trust factors I mentioned above. To get to the point where people will use voice UI for buying, first they need to trust Amazon to buy based on its recommendations, price selection and reviews.

This model of course has limitations, as different product categories still require visual representation. Clothing for example is highly dependent on visual appeal. That’s why it is rumored, and I think inevitable, for Amazon to add a small screen to all Echo devices. It will act as a sort of visual confirmation to make sure Amazon picked what the customer really wanted.

But to me this is the inevitable feature of e-commerce. The voice part is just one of the UI methods. It is the future where we rely on algorithms to do all the work for us - we tell them what we want, and a few days or hours later it arrives. It relies on trust for those algorithms to find the best possible deal and the possible product. But it hides the complexity of having to do price shopping, reading reviews, analyzing products, even talking to sales people. It’s an optimized way of shopping.

And yet to get there, Amazon and all the other e-commerce sites have to make sure to gain trust. Like the previous mentioned reviews gaming - if an algorithm decides what products are best, but it is basing it on reviews easily influenced by a few people, then it obviously doesn’t work.

Today many websites are spending a lot of effort optimizing product descriptions, and getting as many reviews as possible, none of this will matter in the future. “Alexa, order me a t-shirt” and I get what I wanted.