Amazon gets mind reading/shipping patent
Amazon has attained a patent for "anticipatory shopping," a system designed to cut delivery times by predicting what buyers are going to buy before they buy it.
According to the patent filed late last month, the system may box and ship product it expects customers in a specific area will want before the items are even ordered. Beyond local demographic data, the locations would be based on assorted customer data, including previous purchases, product searches, wish lists, shopping cart contents, and returns. The amount of time a user holds a cursor over a product may also be taken into account.
"It appears Amazon is taking advantage of their copious data," Sucharita Mulpuru, a Forrester Research analyst, told the Wall Street Journal. "Based on all the things they know about their customers they could predict demand based on a variety of factors."
Once in transit, the packages would theoretically wait at the shippers’ or Amazon’s warehouse hubs or on trucks until (and if) an order arrives. In some cases, partial street addresses or zip codes will be filled out with the remaining pertinent details — name, rest of address — completed once the order arrives.
The patent asserts the system would help same-day delivery for popular books or other items that customers should want the day they are released, the Journal points out. Items could also be suggested to likely buyers while in transit.
But reducing delivery time by starting the delivery process earlier was said to be the biggest benefit. Amazon has been expanding its warehouse network to support overnight and even same-day deliveries. It isn’t known whether Amazon has already tested predictive shipping.
The patent also included pages of scenarios around re-routing packages.
Wrote Natasha Lomas for TechCrunch, "At times the language of the patent sounds as if Amazon is thinking of physical item delivery in the way a utility might approach supplying water or electricity to homes — by forecasting demand spikes and lulls, and tweaking its pipeline accordingly, but above all by keeping the stuff flowing (ergo having trucks constantly filled with packages in continuous perpetual motion)."
In some cases where the demand prediction algorithm fails, the patent suggests that Amazon could provide customers discounts or present the customer with a surprise gift "to build goodwill" in order to avoid costly returns.
The system could be potentially licensed to others. TechCrunch notes that Amazon requires other e-commerce to license one-click buying, a technology it patented in 1999.
- United States Patent US 8,615,473 – United States Patent and Trademark Office
- Amazon Wants to Ship Your Package Before You Buy It – The Wall Street Journal (sub. required)
- Amazon Patents "Anticipatory" Shipping — To Start Sending Stuff Before You’ve Bought It – TechCrunch
- Amazon says it can ship items before customers order – USA Today
Discussion Questions
Do you think anticipatory shipping is a realistic notion? If accomplished, would it be a game changer for online shopping? What questions remain unanswered about such a practice?
It’s probably not realistic today – but that doesn’t mean Amazon shouldn’t be trying to build the models. If anyone is poised to do it, given the data they have, it’s Amazon. The question is whether a practical model exists for what they are trying to achieve. If it does, I’m not sure it’s a game changer, but is one more reason to go to Amazon rather than someone else.
If anyone has “BIG DATA” it would be Amazon. They already have a ongoing set of predictive analytics based on consumer purchase patterns.
If you are a regular Amazon customer, you get a host of suggestions of “recommended items” by email and other means today. Amazon does even better at “anticipating” purchases with their Prime customers.
So yes, anticipatory shipping is realistic and the base infrastructure and big data are already in place today. It will work well for consumables purchased on a regular cycle, and certain media launch items like books and movies.
There is real danger of over reach when applying this to the individual consumer. Target and others have offended consumers by predicting “life events” such as predicting pregnancy based on other purchases.
There is a very delicate boundary between the value of quicker shipping and consumers being offended that a retailer knows too much about their personal life.
There’s nothing Amazon can put out there the media won’t herald as “just around the corner” Forget the energy it would take to do this, the returns or the multiple walls to hurdle…my favorite is the line that where it fails, it would gift customers.
Which hackers wouldn’t want to mess with a predictive shipping model on their browsers?
The only way it can be implemented realistically is to have very few items that would need to be returned or rerouted. They could give the customer a “surprise” gift as “goodwill” as the patent states, but how long will that last before the economics don’t make sense?
Having said that, you have to hand it to Amazon to always be looking for the next thing that can make them even more competitive, effective, efficient, and customer centric. They will eventually get this right and everyone else will be catching up.
This is not really new stuff – it is just an extension of artificial intelligence and a great PR tool.
First, back in 1985, “Robert Rosen defined an anticipatory system as follows: A system containing a predictive model of itself and/or its environment, which allows it to change state at an instant in accord with the model’s predictions pertaining to a later instant.”
Next, back in the ’90s AI was very hot, yet the noise has cooled and in a way, gone underground. So the tools are there and if you have lots of data, the AI models can do lots of things.
Last, the supply chain for all retailers is a major focus area. We saw a few years ago how many retailers flipped lots of focus to the customer areas and the loyalty space. The supply chain sat and waited for attention. Then the holidays occurred this past year and the focus is back. So, finding ways to improve the chain is a top deal with many retailers, and Amazon’s idea is another stab at a much larger problem.
Could it work? For some repetitive or simple buys – the books category is too easy – but why buy the physical book when the Kindle can deliver faster?
Some AI of mine in use.
Certainly, what Amazon proposes to do is realistic. It is also something we have a highly technical and revolutionary term for: we call it Demand Forecasting.
That is to say, the technology that Amazon is proposing to apply to its supply chain is the technology that all retailers and distributors apply (or should apply) to manage their inventory. This is to build forecast models, taking into account historic customer preferences, geography, price, seasonality and holidays. These models generate forecasts for specific regions or stores, and ship product to meet those demands. And, just like Amazon’s proposed system, when actual demand does not meet forecast, a markdown regime is instituted to deplete inventory.
I don’t want to be presumptuous, because I have not read Amazon’s patent application. But I certainly hope that the patent office does not see fit to grant a patent for a long-existing concept and (quite probably) long-existing methods. This would not be a patentable technology if it simply applies in an obvious way what already exists to a (somewhat) new domain.
Amazon is the darling of the world, and we once again are talking about their guessing what I’m having for dinner next Tuesday. Give me a break please, as this is another way to keep Amazon credible on an idea which is absurd. If I could anticipate what my customers want before they want it, I would be psychic, and never have overstocks.
They already know what customers buy, as all of us get e-mails daily, and unless they set up something like a coffee club, or wine of the month club with automatic shipping, I don’t see this happening, as customers may love Amazon, but will buy when they need something.
Hmmm… I’ve been getting more involved in anticipatory computing technology, but never thought of it for this purpose. In a nutshell, I think such a system would have limited usability and may simply be a patent grab.
There’s essentially zero cost to an anticipatory system that is all software, such as making user suggestions or playing music, etc. So errant predictions incur no cost, only impact a user. A system that requires physical intervention that incurs an expense like packaging/shipping, doesn’t impact the user, but does detract from the bottom line. So in extremely clear cases, say like when overwhelming data shows that a person who buys two parts of a movie trilogy will come back for the third, it may work. But the farther out on the algorithmic curve Amazon tries to go, the more costly wrong predictions will become.
So I’m thinking that anticipatory shipping is what retailers do all the time…to stock their brick and mortar stores. Right? Rolling the dice on inventory is the foundation (albeit wobbly) that impacts purchase and profit.
Amazon’s approach is data based and probably has a sophisticated algorithm or two associated with it. However, it’s still dependent on historical data being predictive of future behavior.
I’m curious to see a comparison of practices and outcomes…between the online shipper and brick and mortar retailers.
Based on Amazon’s ability to correctly suggest books and other products to me (and I confess to being an Amazon uber user) I’d say the answer is “doubtful.”
Ironically, the system might work better with lower frequency, more predictable customers. The whole idea of masses of inventory speeding to a staging area but not being ultimately deliverable – or deliverable at a loss – is also a tad problematic.
That said, this is Amazon and they do have a track record for making the doubtful happen so I, for one, am not voting against them quite yet.
Obviously, if it works it would be a game changer, but just for online shopping. Won’t customers expect the same kind of service/satisfaction levels from physical stores? Couldn’t Amazon, if successful, change the entire retail paradigm and make brick and mortar retailing even less relevant?
What Amazon knows is that he or she who owns the customers wins in the end.
The only real questions here are (a) will it work? And (b) what will physical retailers do to counter the strategy?
Very interesting idea and the comments so far outline many of the opportunities and challenges. Part of this strategy has to do with patenting the idea and then protecting it. Can you say “1-Click”?
For a little entertainment, here’s a YouTube video on Amazon Yesterday Shipping….
Even if implementation does not take this idea to the end of the supply chain, just refining the inventory management would be a huge win.
To paraphrase a famous quote that I give my entrepreneurship students and I believe wholeheartedly, “Shoot for the moon. Even if you miss, you will land amongst the stars.” That is how Amazon got to where it is today.
Interesting that Amazon is making public these extravagant “futuristic” technologies. Makes me suspicious there’s some problems because these seem desperate (I’m thinking drones here).
Smart retailers will stay the course and avoid these distractions that reveal more about Amazon’s flaws than anything.
Given the recent issues shippers face with last minute orders/deliveries at Christmas, this first sounds like a great idea. Have the products customers are likely to buy closer to them to cut shipping time. The issue is, what happens when it goes wrong (and it will)?
Why do I say it will? Because humans don’t always do what we expect. An interesting question will be when it does go wrong, will we hear about it on the news or just in the financial statement of Amazon. My bet is the latter.
Personally, I would have voted Nil given the option but full marks to Amazon, yet again, for getting free opinions on an idea that they may never have any intention of implementing. And, of course, keeping everyone talking about how (potentially) very clever they are.
Depending on the quality of the data and analytics, this seems to present the opportunity to lower costs while serving customers in a more timely fashion. It seems entirely feasible when you consider blending prior interaction histories across purchases and browsing behavior should be able to inform this anticipatory kind of analysis.
There isn’t a company (or person) out there that can anticipate everything a customers needs, but they can anticipate some of those needs. It’s already being done in the medical prescription area. Just about the time the patient/customer is about to run out of their monthly prescription, the meds show up in the mail. Why not with other consumables?
If anyone can take this to the next level, it’s Amazon. Looking forward to seeing how they do it.
Anticipatory shipping makes a huge amount of sense for Amazon, since they have droves of data about the purchase habits of their heavy customers. They are not actually mind reading, rather they are “reading the tea leaves” based on past purchase behavior of specific heavy user customers. Since the best predictor of customer behavior is their past behavior, Amazon will be able to advance products to the closest depot in advance of a customer order, in order to further improve their customer service and speed of delivery.
As e-commerce moves towards same-day shipping, this approach is not a game changer but represents an incremental step forward by an e-commerce retailer dedicated to exceeding customer expectations whenever possible.
The only questions that are unanswered are logistics and financial. The key will be to balance the cost of anticipatory shipping and additional storage relative to the incremental revenue that is driven by customers spending more money with Amazon, due to their excellent customer experience.
The innovation geniuses at organizations including Amazon, Google, IBM and others continue to amaze me. This idea addresses the age-old statement that I would often share with retailers, “Retail customers don’t make ‘appointments’ like airline passengers, hotel patrons and other industries that have information to anticipate incoming demand.” Well, it looks like this obstacle is now being challenged.
I think this process has some bugs to work out, as expected, however, I do feel that there is enough information to leverage demand signals. Certainly there will be issues including those articulated already within the comments. However, I think there is great promise for this service and I also believe this will come to market sooner rather than later.
Of course this is possible. This plan is simply to leverage the newest, available data and integrate it with existing demand data, in order to take Demand Planning and Forecasting to the next level. It’s called Predictive Analytics and we’ve been doing it for years. The question is, “what are the costs and how long will it take for this process to become accurate enough to be considered profitable?”
Click stream analysis has been around for years. Integrating it with shipments and forecasts and sales data is not a new concept. We talk with companies every day about the advantages of leveraging sales data with shipments and with on-line data. Amazon is doing exactly what they should be doing.
That said, I do hope the patent office realizes that this is not a new concept. It would be a shame if Amazon gains exclusive rights to do this when every company I know has similar plans to leverage sales data from all sources with shipments, forecasts and other “big data” in order to gain better efficiencies.
Folks, I love data and I love predictive analytics but…if someone can predict when I’m going to pull the trigger, opps, hit the BUY button, then I’m running in the opposite direction.
Or even, “Gee I knew you wanted to buy this, so I just went ahead and sent it.” What??? There is clearly something wrong with that model. Just sayin’….
If shipping products to consumers gets to the point of those products showing up before they are ordered, consumers might land on the “creepy” side. Getting same day delivery for items will be seen as very desirable. For Amazon it will be interesting to see how many products are ready for shipment but never get ordered versus the number of very quick deliveries that are made.
Theoretically, Amazon has the data to determine what items were being searched for randomly versus seriously to order, what items were ordered for relatives versus personal use, and what impact this has on inventory, costs, and consumer loyalty.
I see potential in using predictive modeling to move certain higher-demand items out into the supply chain where they can be more rapidly delivered over the last mile. The absolute impact on business results or shopper experience seems a little tenuous, though.
From this announcement and other recent rhetoric, I’d say Amazon is trying to keep the pressure on its competitors by focusing the conversation on delivery speed. Even if the scenarios have limited practicality, the announced patent may be a distraction to others, with a chilling effect on certain business plans.
Speed may be a metric Amazon can win, but I don’t see it as a game-changer for shoppers.
The PR geniuses at Amazon just keep cooking up great new stories for us all.
Can I just ask…did anyone understand the graphic?
The part in here that makes this more feasible is the notion that this is not about any single customer, rather it’s aggregating probable demand in a shipping zone of many customers. In some ways that’s not unlike staging most frequently ordered product towards the front of the DC, with less frequent stuff buried far in the back. If the population base of the delivery route is big enough, then the accuracy could be quite okay.
Amazon does seem to be making a few amazing announcements lately, between drones and now this. Anyone want to predict when the company will turn a profit?
Think of the delivery chain as an extension of the warehouse, and it makes a great deal of sense. You wouldn’t even have to be very sophisticated to recognize that if xyz truck always carries 5 of item abc, maybe we should put 7 on there in case 1 or 2 more orders arrive AFTER the truck has left the warehouse. The truck then is an extension of the warehouse for those 2 items – and possibly many more.
I realize that is not exactly what Amazon is doing, but it is close enough. Without reading the patent, I would suggest that ANYONE could do this without infringing – IF they control the delivery truck directly.
This is absolutely a step in the right direction for Amazon and while the capabilities are not yet currently in place, this is slowly becoming the new paradigm. Big data, predictive analytics and superior customer experiences will continue to set Amazon apart from the rest of the retail world.
Considering that Amazon still sends me promotional emails about diapers and wipes even though our youngest child was out of diapers more than 5 years ago; my assessment is that Amazon still has a ways to go before they can rely so heavily on their predictive capabilities.
Feels more like PR at this point.
All the comments on the potential challenges are valid. However, I would not rule out the ability of Amazon to utilize their data to make this a reality, at least for more commodity-type categories.
This doesn’t appear to be any different than what Target, Walmart or any other large retailer currently does. Basically, anticipate sales of current and/or new items based on customer and sales data in order to supply the distribution centers with merchandise needed to stock their stores with the right merchandise.
It is somewhat realistic, but in a limited sense. I have spoken with companies that claim they can predict with a high level of confidence what consumers will buy next. They then use that information to create a targeted offer for the very item they expect the consumer to buy. At best, these might have a 20-40% response rate. Even if Amazon could improve upon those percentages, there will always be a loss component when they get their predictions wrong. As such, one can create mathematical models calculating the benefit realized when the prediction is correct, and subtracting the cost of the items and shipping when the prediction is wrong (with some factor for the potential of goodwill that may be generated if the incorrectly predicted item is gifted to the recipient).
As others have stated, what is most noteworthy here is that Amazon is continually thinking about the end-to-end purchasing process, going as far back as possible into the consumer’s thought processes, and seeking to gratify the consumer’s wants as quickly as they can. Credit to them for pushing the pace, and for finding ways to continually obtain media attention and lay claim to a leadership position with respect to customer service.
Extending this thinking forward a bit, drones are so yesterday. I predict that Amazon will soon be applying for patents or releasing stories about the Star Trek-like transporters or replicators for getting products into consumers’ hands in near real-time. The next step will be to plug those into the consumers’ brains like in the Matrix, or even better, create a wireless version that reads consumers’ minds and automatically replicates it before the consumer even realized at a conscious level that they wanted it. 🙂
Not knowing the actual methodology being used, I believe it’s not a perfect science and at this point any company that installs such a program will end up having a lot of returns and do not accepts. Because it becomes a budgetary issue for consumers, just like it has been for people since Amway and other MLM companies started to automatically send out products on a scheduled basis.
If done right with the right type of consumer, I see great value. I’m interested in this.