How Miros AI is solving the problem of product discovery in e-commerce

A little while ago, I chatted with Estonian co-founder Heikki Haldre, chairperson of Miros AI, at the Latitude 59 conference in Estonia, who told me something that highlights the value of conference-facilitated networking investors. 

"You know, those 15-minute matchmaking meetings where you meet somebody? I did one, and at the end of the meeting, this guy said,

"I'm going to invest."

I thought to myself, "I don't believe you. We've all seen those guys with lots of money and a lot of talk. "Well, three days later, the money—$100,000—was in my bank account. And this was the second time it's happened. The first time was after a 15-minute call, albeit for a much smaller amount."

Miros is solving a problem that has plagued shoppers and retailers - searching for what you want.  

According to Miros, 20 per cent of shoppers can describe the product with simple keywords. Meanwhile, 80 per cent of shoppers are scrollers who struggle because they can't put their desires into words. This is particularly the case with shallow stock sites such as Etsy, where many items are one of a kind. 

This results in a measly two  per cent conversion rate, which means 98 per cent of customer acquisition costs are wasted. 

According to Haldre:

"Almost half of the retailers tell me that their search engine is not optimal:

'We type something in and know that this product is in stock, but it just doesn't show. We have to tag the products properly, but they're not tagged. 

Ask any retailer, "So how many products do you sell that cannot be accurately described in words? Of course, when you sell electronics, everything can be described, but when they sell design or furniture fashion, then they say, "Oh, or 80 per cent, and 90 per cent?" 

Miros has developed wordless search technology that empowers shoppers to find their perfect match in under a minute, without typing a word, and grows e-commerce sales by up to 40 per cent. 

Instead of using keywords, shoppers can find products by interacting with images. Its AI reads the contents of pictures and delivers similar products, creating a more intuitive search process.

As shoppers browse, the system provides real-time product suggestions through features like the Discovery Bar, which offers personalised recommendations based on the user's behaviour.

Haldre details: Our technology has evolved, where we can say, "We cannot read people's minds, but we can pretty well predict what people desire based on their actions without they can communicate to us through their actions rather than their words." 

Miros understands your sense of style, which enables you to find things that are better than you can imagine."

Further, Miros stands out from other personalisation technologies that rely on historic browning behaviour.

"It's unnecessary — there's no point looking at items you bought three months ago, last season, to determine your next-season purchases. 

"Why not just ask a shopper, "What are you interested in right now?" No cookies or privacy-invasive processes are needed."

This is company number six for serial entrepreneur Haldre, who has exited four companies, including fashion e-commerce tech company Fits.me which was acquired by Rakuten.

This is the second time Haldre has worked closely with academics at Europe's top 1 percent computer vision AI research institution, iCV Lab. 

Due to the specialised nature of this work and the diverse talent required, maintaining a large in-house research team would be both costly and challenging. This approach offers a more efficient and flexible solution for the "layers and layers of additional research that needs to be happening."

"This is something most startups can not even dream about, to get access to that level of AI talent. Some people were almost the grandfathers of the kind of AI we observe today." 

Haldre calls innovation when invention is made available to the public. 

"So it's invention plus commercialisation. 

However, academic scientists are just interested in invention. So our job is to work with those scientists, take their invention, and help them guide it so that it's commercially applicable. As soon as this is done, we take it over and build it in-house. 

This is engineering. 

As our company scales, we increasingly identify the need for innovative solutions. So we go back to the university and ask for help."

As a serial founder, Haldre revealed that his biggest learnings are that "you have the courage to take risks and go for bigger scale. 

"I know that the probability of failure is still super high, as this is a problem no one else has solved. But I don't have the early stage founder a constant nagging in the back of my head, "Is this the right thing to do? Am I making a mistake? 

I step into new situations without this constant anxiety, and I'm much calmer and far less emotional if I make a mistake."

Haldre also shared that he's learned the ability to listen:

In my early years, I refused to listen to people who wanted to help me. I was the smartest person in the room. Now, I'm opening my ears, and when somebody tells me something, I'm like, 'How? Teach me." And I feel that things go faster and easier as a result." 

As a founder, Haldre sees vast applications for Miros' tech:

"Every one of us is creative. We can imagine new fashion styles, sculptures, architecture, and paintings, but what happens is that human language is not rich enough, and we cannot; most of us cannot draw or paint the new sculptures."

But first, the company is committed to disrupting e-commerce.

Retail is one of the top employers in any country. So, this will make a society-level impact. And it will be here soon."

Lead image: Freepik.