Deep Learning Revolution: How Amazon is Making Retail Obsolete

Harry Chiang is a Financial Analyst at I Know First

Deep Learning Revolution: How Amazon is Making Retail Obsolete

“I think frugality drives innovation, just like other constraints do. One of the only ways to get out of a tight box is to invent your way out.” – Jeff Bezos, Amazon CEO


  • Two and a Half Decades of Success: The Amazon Story
  • Amazon Still on the Rise: Improving Financials
  • The Retail Giant’s Involvement with Deep Learning
  • How Amazon Go Will Change Everyday Lives
  • Where Do We Go From Here


Deep Learning Revolution

Two and a Half Decades of Success: The Amazon Story

On June 2nd, investors were delighted to hear that Amazon stock had closed above $1,000. While this is already a massive increase from last year’s benchmark of around $700, there are some that think Amazon’s near future holds even more growth. Tom Forte at the Maxim Group wrote in a note to clients yesterday that Amazon he was adjusting his price target on the shares from $1,075 to $1,300. He cited 10 possible markets for that the corporate giant could further disrupt. Forte claimed that these markets will be worth around $1 trillion in the future. Forte cited the following opportunities, “Apparel, B2B E-Commerce, The Consumer Internet of Things, Gas Stations, Grocery, Pharmacy, Professional Services, Restaurants, Stores, and Travel”.

It’s not completely far-fetched to believe that Amazon has more movement in it. The corporate behemoth has quickly become a household name in the two short decades of its existence. Under Jeff Bezos’ direction, the company is now the largest Internet-based retailer in the world by total sales and market capitalization. It is also currently the fourth most valuable public company in the world. The company is the largest internet company by revenue globally. It is also the ninth largest employer in the United States.

In short, Amazon has come a long way from its beginnings in 1994 as an online bookstore. Its quick diversification and identification of lucrative opportunities has been key in its growth. Bezos’ savvy and predictive vision has led Amazon to live up to the origins of its name. Jeff Bezos conceived of Amazon as a representative of the massive river of the same name. It has certainly spread well beyond just the span of a single country, however.

Now, Bezos hopes to become even more market dominant, and continue operating several moves ahead of everyone else. Investors were dubious in the early years before Amazon had yet to turn a single profit. However, investors now wait with bated breath as Amazon continues to take bold strides in to new profitable markets, such as deep learning.

Deep Learning Revolution

Amazon Still on the Rise: Improving Financials

Before delving in to Amazon’s role in the deep learning frontier, it is important to examine the company’s recent financial situation. On February 2nd, 2017, Amazon announced financial results for its fourth quarter of 2016. Operating cash flow increased 38% to $16.4 billion for the trailing twelve months. This is compared to $11.9 billion for the trailing twelve months preceding December 2015. FRee cash flow increased to $9.7 billion for the trailing twelve months. This is up from the $7.3 billion of the financial year of 2015. Operating income increased 13% t $1.3 billion in Q4. Net Sales increased 22% to $43.7 billion in Q4. Net income was $749 million in Q4 2016. This is compared to a net income of $482 million in Q4 2015.

In Q4 2016, Amazon also announced it will create more than 100,000 new full-time full-benefit jobs in the U.S. over the next 18 months. Fulfillment by Amazon (FBA) delivered more than 2 billion units on behalf of sellers in 2016. The number of active sellers using FBA grew more than 70%. In 2016, there were over 100,000 sellers with sales of more than $100,000 selling on Amazon. It is evident that Amazon had a strong financial year in 2016 and with further expansion on the way, investors are cautiously optimistic about an even better 2017.

Amazon did not disappoint with its results for the first quarter of 2017. Operating cash flow increased 53% to $17.6 billion for the trailing twelve months, compared with the previous $11.6 billion for the trailing twelve months ending March 31, 2016. Free cash flow increased to $10.2 billion, compared with $6.7 billion for the same trailing twelve months previous. Net sales increased 23% to $35.7 billion in the first quarter. This is compared with $29.1 billion in first quarter 2016. Net income was $724 million in the first quarter 2017. This is compared with net income of $513 million in Q1 2016.

Overall, Amazon still expects to grow in the second quarter of 2017. Net sales are expected to grow between 16% and 24% compared to Q2 2016. This is a range of $35.25 billion and $37.75 billion. Furthermore, if Amazon has more success with its deep learning project, which is slated for public release this year, the company can expect a strong financial year indeed.

Deep Learning Revolution

The Retail Giant’s Involvement with Deep Learning 

Deep learning was mentioned several times in the previous paragraphs. This may come as surprise because many people don’t think of Amazon as a cutting-edge, high-tech company in the same vein as Apple or Google. There is a tendency to view Amazon as being simply a retail store. A retail store that is massive and limitless, but is a retail store nonetheless. However, as Amazon Web Services states, “we do a lot of AI in our company…we have thousands of people dedicated to AI in our business. Even though Amazon hasn’t always talked about its machine learning background.

Perhaps one of the most obvious displays of Amazon’s involvement in AI and deep learning is its release of Alexa. Amazon first announced Alexa alongside Echo in November 2014. Alexa is listed as an intelligent personal assistant capable of various services. These include voice interaction, music playback, making to-do lists, streaming podcasts, etc. It can also provide real-time information, such as news, weather, and traffic. Furthermore, Alexa can also control several smart devices using itself as a home automation system.

This is exciting in itself, but aside from Alexa, Amazon has had several other forays in to AI/deep-learning. In November 2016, Amazon announced two major projects through its Amazon Web Services. These were Amazon Rekognition and Amazon Polly. Users utilize the former to identify objects/people in images and the latter for speech-to-text technology.

There have been rumors that Amazon plans to combine these two with Alexa to bring automation and AI to call centers. These would replace the human workers at call centers. Using deep learning, the machines would eventually, over thousands of call hours, be able to identify and respond to problems people might have.

Amazon has continued to toy with the concept of deep learning and AI in other ventures, with perhaps its other most serious project being the release of its deep learning library. On May 10th 2017, Amazon released an open-source version of its deep learning library DSSTNE (pronounced “Destiny”). This is free for the public to engage with and serves as a means to develop deep learning applications.

Deep Learning Revolution

How Amazon Go Will Change Everyday Lives

All the aforementioned projects, however, have not yet included a serious investment in a physical application to Amazon’s deep learning technology. Amazon’s end of 2016 announcement of Amazon Go interests deep-learning experts far more. Amazon Go is a conceptual grocery store which requires no checkout on the customer’s part. Amazon claims the consumer need simply check in to the store using the free Amazon app, obtain what they want from the shelves and leave. The store and app will keep track of whatever products the customer selects, adding them to a virtual cart. It will then charge the Amazon account upon leaving and send a receipt to the user’s mobile device.

Amazon is currently marketing this under the trademark “No Lines. No Checkout. (No, Seriously)”. As of today, Amazon Go only exists as a single 1,800 square foot prototype store in Seattle, Washington. This unit opened on December 5 2016, and is in its beta testing phase, only used by Amazon employees. The advertisement claims that the store uses a combination of computer vision, deep learning technology, and sensor fusion to enable the easy shopping process. Amazon is calling this “Just Walk Technology”.

With Just Walk, Amazon aims to improve several retail aspects. Firstly, data-driven products display. The technology will be able to track consumer movement in the store and thus learn how to best modify placement and assortment of goods. Second, it hopes to improve assortment renewal. The smart grocery experience will be able to define the best shopping times, volumes, and items for replacing the goods. This will minimize costs. Finally, Amazon is hoping to drastically improve and personalize the customer experience.

Evidently there are some flaws with Amazon’s system. Many deep-learning vision experts claimed that this sort of technology was years away. While Amazon has deployed a form of this technology much earlier than expected, they are not entirely wrong. Amazon’s store can only track up to 20 customers at a time, severely limiting its utility. Furthermore, it doesn’t take much thought to realize that by huddling together and creating confusion it is possible to fool Amazon’s tracking system.

Business Insider reported in October 2016 that they had seen internal Amazon documents that described Amazon opening as many as 2000 stores, over the next 10 years. Amazon spokesmen refuted the report. Business Insider has since predicted that Amazon is likely to open only around 20 stores during the next two years. The Verge reported that Amazon’s first store is scheduled to open to the public in early 2017.

Regardless of the current rumors, it is undeniable that Amazon is pushing the frontier of consumer technology by involving such advanced deep learning. Experts compare it to self-driving car technology. If Amazon Go becomes widespread and fully functional, it stands to change the way we shop entirely. Investors should not underestimate the level of convenience that it provides compared to the traditional model of grocery shopping. Integrate this with Amazon’s online dominant presence and recent advances in online grocery shopping and this will change our everyday routines.


Deep Learning Revolution

(Image Source: Amazon, accessed 5:06pm 06/07/17).

Where Do We Go From Here?

It is tempting to pass off Amazon Go as a gimmick. However, that would be an erroneous presumption. Consider the full scale of what Amazon Go is attempting to achieve. Yes, it is essentially making shopping easier for the average person, but it is also simultaneously Amazon’s spearhead attempt in cracking a $800 billion global food business.

The Food Marketing Institute and Nielsen released preliminary findings in January 2017 indicating that by 2025, online grocery shopping could reach $100 billion annually. This is 20% of the market share. Not only is Amazon making strong plays in the field of the online food business, it is now attempting to crack the brick-and-mortar business.

Amazon’s foray in to the food business began in 2007, with its grocery delivery program called AmazonFresh. Despite a few upsets over the last couple of years, Amazon is not slowing down. Amazon is reportedly hiring executives from major food brands to redesign their packaging and operations to prepare items for shipping directly to online shoppers rather than to sit on supermarket shelves. Forrester analyst Brendan Witcher stated, “All retailers should see this as a shot across the bow…Anyone who is paying attention can see that Amazon is building a ‘grocery-to-home’ ecosystem not unlike what Apple did for the music industry”.

If Amazon Go can successfully take off, it stands to reason that it will likely eat up the traditional business models of grocery shopping. It plays on Amazon;s main strength against other traditional businesses: understanding how consumer needs meet convenience. Although this may seem like a trivial factor, it has proved itself to be paramount over the last two decades to a business which is now the fourth most valuable public company in the world.

In terms of the online front of business, Amazon recently announced that it will cut its Prime membership prices by almost 50% for low-income shoppers on federal welfare. This is a direct challenge to Walmart. Walmart relies heavily on low-income shoppers and receives 1 of every 5 dollars of its revenue through food stamps each year. Walmart and Amazon combined are worth more than $600 billion in revenue in 2016, almost the size of the Defense Department. However, Amazon is moving faster towards Walmart’s consumer market than Walmart can respond. Walmart is going toe-to-toe with an absolute behemoth, and it is looking increasingly likely that Amazon will be attempting to edge Walmart and the traditional consumer model it represents out of the market.

For perspective, there are between 60 and 80 million Prime subscribers in the country. The typical Prime household spends more than twice as much on Amazon as a non-Prime family. This is a growing market which is massive and continues to expand. With Amazon Go, Amazon is poised to dominate both online and brick-and-mortar markets.

Furthermore, Amazon has its moved planned well in advance of just domestic borders. Last year, Amazon announced plans to invest $3 billion in India operations in addition to the $2 billion announced in 2014. The company has already invested more than $1 billion in India. Currently, 20 fulfillment centers are operational across 10 Indian states. As the second largest retailer of its kind in the Indian market, Amazon is well placed to take advantage of the predicted $75 billion worth of the Indian market in 2020.


Overall, it might not be so far-fetched to believe that Forte’s prediction of a $1,300 high might be reachable. It will depend upon Amazon’s execution of its latest driving technology: deep learning. If Amazon can truly implement this increasingly important technology, then despite a predicted unfavorable $720 million impact in 2017 due to foreign exchange rates, Amazon is well on its way to even stronger growth.

I Know First’s deep learning algorithm continues to predict bullish trends for AMZN stock prices. As seen below, in all three time horizons Amazon is looking to trend upwards. There is a signal strength of 44.30 predictability of 0.46 for the one year forecast.

I Know First

The sign of signal tells the user which direction the asset price is expected to go. Positive means an upward direction, a long position. Negative means a drop, a short position. The signal strength is related to the magnitude of the expected return. It is used for ranking purposes regards investment opportunities.

Predictability is the actual closeness of fit function being optimized daily. In other words, it is a correlation-based quality measure of the signal. This is a unique I Know First algorithm indicator. It allows the user to focus on the most predictable assets according to the algorithm. It ranges from -1 to 1.


To subscribe today and receive exclusive AI-based algorithmic predictions, click here.