It was in the third week of July when a startup in the online retail technology space, SmarterHQ, blipped briefly across the industry's radar. Simon Venture Group, the venture capital arm of the shopping center owner Simon Property Group, had led a financing round for the company, helping it secure $8 million.
There was a flurry of headlines about the deal and then it was on to the next big thing.
Which, quite ironically as you will see in a moment, was about Amazon's quarterly earnings and how the e-commerce giant turned a profit in large part because of its Amazon Web Services division. Then, the spotlight moved on to the fact that Amazon had a $265 billion valuation, topping by $30 billion Walmart's valuation.
Meanwhile, SmarterHQ with its paltry $8 million and all but forgotten three days after the funding announcement remained hard at work building technology that could push e-commerce into a new era of personalization.
Of course, SmarterHQ is not the only firm looking to disrupt e-commerce. Last month we also saw the launch of Jet.com, a site that many believe has the chops to threaten Amazon with its new pricing model.
Briefly, Jet.com charges a membership fee, but in exchange it gives customers much cheaper pricing because of the way it has stripped out every last bit of supply chain inefficiency.
SmarterHQ isn’t looking to join the online pricing wars with a new product or business model -- but it is working on another aspect of e-commerce that has been stagnant for some time.
A Retail Site That Morphs In Real Time
The SmarterHQ platform is able to offer products to customers that they are most likely to want as they navigate onto and throughout the retailer's site -- even if that person has no history with the retailer. It does this with a mix of contextualized data, descriptive model scores and predictive modeling.
A word about the latter: Conceptually, predictive models do nothing that humans couldn’t do themselves if they had enough time to do it. But the average retailer – or giant conglomerate for that matter – simplly doesn’t have the time considering the huge number of possible purchase scenarios based against an inventory that can count into the thousands of SKUs. This is where machine-learning algorithms take over, shifting through these combinations to find the most likely models for each product.
This is why when you navigate onto, for example, a retail site that sells among other products exercise clothing and equipment and start browsing, say, for yoga clothes, you are also shown yoga-related equipment and gloves.
This is one area where SmarterHQ excels: it can build tens of thousands of models for each retail customer, using automated data preparation and model building. It can build a lot of models, in other words, and very quickly too -- fast enough so that they can be updated when new data comes in.
So one day when you navigate onto that site that sells yoga clothes you aren't shown the nonslip fingerless gloves and the cork brick. Instead you are also shown some walking flats because, like a lot of other people who have bought yoga clothes on the site, you aren't necessarily wearing the clothes just in the studio. The site figured that out at some point and added that model into the mix.
But this isn’t really SmarterHQ's secret sauce, or at least not all of it.
Where it really excels is in understanding what that yoga-pants wearing consumer wants in real time while she is on the site. Not the next visit or next month.
All retailers have segment definitions for their customers, one of SmarterHQ's co-founders, Dean Abbott, explained to me. A retailer will know that a certain demographic practices yoga and will build its personalization rules -- or what items are displayed to that customer -- around what it knows. And if one customer starts to exhibit other characteristics, then the retailer moves her into a different rule category. No, she is not really into yoga as a practice. She just likes the clothes.
The SmarterHQ technology makes that process happen in real time, Abbott explained. Back to the yoga customer: now shopping on a site fueled by SmarterHQ, she clicks on the yoga pants -- but then goes against type and meanders into other clothing categories. Instead of showing her the yoga equipment under the "rule" created by the retailer, the site automatically shifts directions and shows her shoes that work well with the new pants.
On the back end it works something like this: Java script captures what the shopper is clicking on and in real time the site matches the clicks with segment definition after segmentation definition depending on where she clicks. All the while, messages are being sent to the server that say 'bring in these set of images.'
The Latency Problem
Not a lot of sites can do this, Abbott says. "It is a lot harder than it looks to reconfigure a Web site on the fly like that."
The biggest obstacle is rendering the message quickly enough to appear instantaneous to the shopper. Abbott calls it a dance between the offline segmentation by the retailer and the online deployment of those segments by the tech platform.
Having it happen in real time "took a while for us to crack. Not many people have."
Amazon hasn’t, he believes.
Specifically this process is called onsite messaging to the customer during a visit based on a segment that is predefined.
To be clear Amazon does likely use onsite messaging, Abbott said. You can see that in the changing page layouts each visit, based on recent behavior.
But he believes this is a static process with updates made, say, every hour instead of in real time.
I asked Amazon precisely that question -- was its onsite messaging in real time or not -- and did not receive an answer.
There are other elements to Amazon's online retail experience that are slightly off, at least to Abbott's eye. Amazon’s personalization and recommendations are based on single-SKU interactions and therefore inadvertent product views or 'quick-look' views of products are reflected in the recommendations, he says. Also, "these recommendations do not take into account the level of engagement the shopper has with the product or product category. Moreover, purchase events are not taken into account. Post-purchase recommendations may include purchased products that are no longer in the visitors consideration set," he says.
Amazon's Innovation
The irony is Amazon is innovating -- just not, noticeably at least, in online retail, the area where it got its start.
Its Amazon Web Services -- a high-margin business as it turns out now that the company is breaking out this unit in its earnings -- recently launched Amazon Machine Learning, a cloud-based analytics service.
It's also innovating in, or rather make that disrupting, the online retail delivery model. For example, the company is reportedly developing a drive-up store concept in California where shoppers would order grocery items online and then schedule a pickup, according to the Silicon Valley Business Journal. This unconfirmed development is being watched with great interest as it could propel the industry that much closer to the vaunted zero delivery time benchmark.
This is what the Washington Post said about Amazon's latest project:
The idea of a drive-through grocery store is in the same vein as some of Amazon's other inspired retail concepts, such as using drones to drop off packages on doorsteps across America. That idea for Amazon Prime Air sounded a bit daft when it first appeared at the end of 2013, but now the Federal Aviation Administration is increasingly on board with drone deliveries, and a whole lobbying movement is forming around commercial drones. The NASA Ames Research Center recently held a three-day event to discuss just exactly how drone deliveries might happen in the future.
Without a doubt, Amazon is creating online retail history with these projects. So, too, however, are other companies.
Also on Forbes:
