Here’s our top retail tech prediction for 2018
What do the shopping season, the 2018 Tax Bill, Artificial Intelligence and Omnichannel all have in common?
2018 is finally here, and in a big way
Whew! If 2017 felt like a wild ride, it was with good reason.
2017 was the year that a ‘phase transition’ happened for retail. A lot of the trends, events and slow changes from the past decade culminated in a landmark year for retail — a landscape shift that at least one company saw coming.
So, what’s going to happen in 2018?
Retailers will look to consolidate their 2017 gains by investing heavily in omnichannel tech
Retail in 2018
A couple of things have happened recently that offer excellent clues to where retail is heading next.
One major development that has us really excited is how good retailers had it for 2017’s holiday shopping season.
In fact, it was the best it has ever been in years.
The second big news that’s going to really affect retail in 2017 is the new tax bill passed by Congress in the twilight of 2017.
The biggest effect is the reduction of the corporate tax rate from 39% to 21% — a whopping 18 percentage points reduction!
This bill is about to have an outsized impact on retail spending — retailers now have 18% extra to spend on their operations. So what do we think they’re going to spend it on?
The answer is pretty straightforward actually — US retailers are going to fill up their war chests for the 2018 Battle for Ecommerce.
Our big prediction?
Artificial Intelligence will eventually take over much of the automation and optimization work in retail, powered by data.
For a phase shift to happen it is not enough to have technology — money is important. And this year retailers are sitting on a large pile of it.
The primary beneficiary of the largesse from tax reduction is going to be Retail Tech — all the software, data, and applications needed to quickly transition from a brick-and-mortar operation to integrated retail.
The writing is on the wall — this is Big Retail’s final chance to embrace omnichannel and ecommerce before they completely lose ground to Walmart and Amazon in the ecommerce wars.
Brick-and-mortar just isn’t cutting it anymore
It has been clear for more than a few years that having a purely brick and mortar business is not the way to long term survival.
What’s also clear now is that ‘pure play’ ecommerce is just as doomed.
Brands and retailers need to be nimble. They need to manage multiple distribution channels that are constantly evolving while ensuring that their brand image stays consistent. They have to be able to engage customers on all fronts — online, offline, and in-store.
Maintaining a singular shopping experience and product representations between all these channels is a Sisyphean task.
And the key to this task is product data. Product data can move faster than products. It is the fuel that is powering ground-up international ecommerce — like dropshipping, where a single person can set up a store and sell a product before it has even been produced in a Chinese factory.
In order to become truly part of this integrated retail universe, mainstream retailers are focussing their technology teams on 4 main areas:
Everyone within the retail ecosystem — retailers, ad-tech companies, manufactures, logistics providers — could benefit from standardized and normalized ecommerce data in order to integrate their services with each other.
Imagine an average sized electronics retailer who gets product feeds from 1000+ manufacturers. This retailer has to:
There are two ways to solve this problem.
The first method would be to turn to external sources for product data. They could compile the manufacturer feeds and query a global product database (like Semantics3, GS1 or Product Open Data) with inputs like URLs, UPCs or product names in order to get all information about a product in a standardized format.
The second method is more technologically complicated but a better fit for bigger enterprises — enriching existing product catalogs. This involves leveraging machine learning models to fill the gaps and clean-up in-house product data.
It is essentially a 2-step process:
What is NER?
Data that isn’t always available in structured form and often has to be extracted from the product information through a process called Named Entity Recognition (NER).
Named-entity recognition (NER) (also known as entity identification, entity chunking and entity extraction) is a subtask of information extraction that seeks to locate and classify named entities in text into pre-defined categories such as the names of persons, organizations, locations, expressions of times, quantities, monetary values, percentages, etc.
At Semantics3, NER is a cornerstone of our approach to AI-powered product catalog enrichment. Most research on NER systems has been structured as taking an unannotated block of text, such as this one:
HP 15-F222WM 15.6” Touch Screen Laptop Intel i5 Processor, 4GB Memory, 500GB Hard Drive, Windows 10
and turning it into this:
Feature normalization is the process by which the specific field entities that have been identified by NER are then mapped and normalized to the Semantics3 attributes taxonomy.
It consists basically of 2 parts:
Key Standardization — This is the process where attribute names (like “length”, “memory”, “heel length” and so on are mapped and normalized to our central taxonomy:
Value Normalization — this is the process where the values tagged to each attribute key is mapped and converted to a standard format across our entire catalog to enable easy comparison and conversion of formats:
The end result is a cleaned up, de-duplicated and standardized product catalog that makes feeding product info to integrated systems much easier.
Bad product categorization on ecommerce websites is more prevalent than you think — even Amazon has trouble getting it right:
In 2018, better categorization is going to be front-and-center of every retail CTO’s strategy.
The benefits are obvious:
Firstly, better categorization leads to a more intuitive product discovery experience. You ideally want your customer to discover products that directly address their latent interests.
Secondly — knowing what kind of products are in your inventory system is a powerful tool to have, especially when you’re trying to optimize your logistics and supply chains.
Imagine if you were an international ecommerce operator. Having better categorization helps you improve your import/export process, comply with many different countries’ HS Code and tariff regulations, and avoid fines.
Huge cost savings, right there.
Our suite of Machine-Learning Categorization modules help retailers to leverage our massive training datasets and create customized ML-powered product classifier engines targeted to any retailer taxonomy or country-specific HS code system
Here’s a quick demo in our Playground (signup and request access for testing)
Every retailer dreams of having this in their website:
Amazon makes it look so easy — but in truth achieving that level of facetting across your retailer website is incredibly difficult.
Basically the problem can be summarized as follows:
in order to be able to do this:
4. Training Datasets + Product Fingerprinting + Entity resolution == Better product matching with AI
2018 will also be the year that retailers will finally tackle the toughest problem of all — product matching for price comparison across websites.
AI-powered product matching not only allows retailers to build better price comparison engines, but it allows them to consolidate, enrich and create comprehensive product pages with all available product metadata.
Good product matching has a ton of benefits — namely it helps retailers take down Amazon, just like how Walmart did:
Specifically it allows a few things to happen:
Firstly it allows sellers to quickly onboard to your marketplace / ecommerce site
By allowing third-party sellers to input simple product identifiers like GTINs, EANs or retailer SKU IDs, retailers enable them to more rapidly onboard their SKUs to the website, and automagically create new product pages.
With good product matching, a simple, singular product identifier should be enough to summon every bit of metadata available on that product, including images, price points, normalized feature-attributes and descriptions.
This specific piece of technology is so powerful that it is helping Walmart grab market share from Amazon in an incredible show of force for the incumbent top retailer.
Secondly, it allows retailers to price their products more competitively.
There’s another reason why Walmart is trouncing Amazon in the online shopping bake-off — it is pricing goods consistently below Amazon with better price monitoring.
The results were unimaginable 2 years ago — Walmart is now consistently cheaper than Amazon across a wide range of popular categories like consumer packaged goods, electronics and home supplies.
Walmart having pricing power over Amazon is a terrifying place to be in — because Walmart has lower costs per unit sold compared to Amazon due to its superior supply chain.
It’s not something to be taken lightly.
Having good product matching (again) has a powerful effect here — it enables ecommerce sales managers to set up linkages between their products and identical offers sold on other retailers. By setting up periodic (even hourly) refreshes, they can consistently price-match or under-price Amazon at scale.
With our AI-powered price monitoring solutions, Semantics3 can seriously supercharge your price tracking infrastructure with up to hourly price refreshes across your catalog.
Our Price Monitoring API has the ability to deliver past, present and future pricing data. Check it out here.
In all transparency, a lot of the predictions we’re making here is backed by our own experience. In the past 12 months, we’ve been working closely with retailers, search engines, app developers, CTOs, ad-tech companies and other retail tech companies.
The market does not lie.
Retailers are more aware than ever that they need to capitalize on their windfall in terms of a tax break and a surprisingly successful holiday shopping season — and convert this golden opportunity into a long term moat against the ever-advancing threat from pure-play ecommerce players.
Embracing the suite of AI tools available in retail tech is crucial if they are to tackle the core problems of expanding their online product catalogs, competitively pricing their inventory and creating a compelling omnichannel experience at scale.
Artificial Intelligence is the only way to do this in a scaleable fashion while keeping costs low enough to grab market share sustainably.
Now that you made it to the end… here’s a bellyrub gif of our favorite puppy, Teddy!
Written in San Francisco and Singapore by Hari Viswanath and Anjali Krishnan
Published at: January 16, 2018