The ecommerce landscape is changing with the arrival of cognitive systems. This presentation is a top level look on how to fuel new revenue channels for marketeers based on AI systems and machine learning solutions
Thanks for having me. Today I want to talk a bit more about the options behind cognitive commerce. I want to cover the basic aspects of what It actually means as well as show you some examples of retail companies utilizing a cognitive way already.
To give you a bit of an idea why we can talk about it. Productsup is an product data management system that allows non-tech personal to work directly on their own product data – without having to rely on IT everytime a feed is necessary for a new marketing partner. By that means, we know how your data is build, managed and optimze and we always think ahead to find you guys new ways to increase your revenue or options while interacting with your customers.
Productsup has clients around the globe and because we work with agencys as well as direct brands our 400+ clients represent around 10.000 Companies globally. We are build for large databases – we currently process around 80 billion + products every month.
But lets talk about cognitive. What I‘ll show you today is bascially the future of productsup as well. Its what we believe will shape the future of the interaction with humanity and ultimately from a business persepctive. So...what is cognitive?
If you ask wikipedia for this you will get the following answer.
Thats me looking at Wikipedia articles. It doesnt matter if it written in english or german – I always look like I‘ve just seen a ghost.
So lets try to break it down a bit. I would like to split it into 2 basic categories. Data you can utilise for your company and the technology that is driving that train at the moment
Currently 12% of the data available is used and 88% remain untouched. Where we store around 0.3 MB at the moment per user per second, by 2020 we will store around 1.7 MB per user per second. You can see that the data is growing exponentionally. But what is that data actually and how can we be able to use it?
If we talk about the unknow data we usually end up talking about sight and sound. Sight includes things like film and photography where sound counts for mostly hearing and speaking. We won‘t be able to touch all of this today but I want to give you a few examples on how far we have come until now.
If we talk about image recognition we can take the latest numbers Google shared as a reference. In 2014 we where able to read around 89.6% of an image, today we can already read 93.9% by image recognition software.
That doesnt seem much but you have to remember that 100% will be the human brain. Until we get to 100% we most likely also need to go 30 years – ist not about do we get there but more a when are we able to make it.
On this image....
At Productsup we are able to identify images as well. For example we will be able to identify the color of th shirt your are having in your product listings based on the image.
When you think about text to speach technology, the most funny thing to do for a long time was sending an sms to a landline number, becuase the telecomunication provider will then read the message for you. This was funny because we did this with a concatenative and parametric method. Wave net is much more advanced and as it is with image recognition we also close the gap step by step with text to speech.
There was this famous case running around in 2014 I think where, if you would have told Siri – I‘m going to jump of a bridge and die – Siri was so keen to provide the next 4 bridges closest to you. Today, and maybe also because of the bad press, Siri will more likely point you towards another person to speak to or a suicide hotline.
Although this is all the past and we have come quite some way to where we are right now, unlimited possibilites are still ahead of us. As mentioned now couple of times those two days, the Personal Assistant is going to change the way we interact with machines and how we live our day to day life. Weather it is MS Cortana, Google Allo, Siri or Echo.
Funny side note – Bixby, which will be Smasung‘s PA is develeoped by the same team, that has develeoped Siri for Apple. After they sold it they thought they can do better, did better and sold it to samsung.
But how good are those assistants currently? A recent Study has asked 5000 question to each of the systems and Thats what they have answered. First they measured how the answer was delivered. As most of the systems are Mobile Phone App where you can choose to also have the answer on screen, Google Home cleary sticks out as it only deliveres answers verbally – it is worth mentioning though that Amazons Echo didn‘t do well.
As benchmark, they also used the regular search function in Google Search App to dictate and read questions.
If you look a the amount of answers that has been correct, not making a difference if answered verbally or not, all systems are quite good except for Apple‘s Siri.
The accumulation of all tests and questions shows the difference in quality. Google is clearly leading the field here and it shouldn‘t be a surprise as they bascially have learned over the last 20 years what people search and what answer they find valuable.
In fact, I read a book recently I can recommend. It‘s call the inevitable and has been written by Kevin Kelly, the founder of wired. there he tells a story where he as met Sergejy back in 1998. They came to chat about Google and why Sergeiy wants to build a search engine. He didn‘t respond with the reason he bascially said we are not building a search engine, we build an AI that will be the personal assistant for everyone in the future.
Also, our PA‘s getting funny. Ist interesting that the ones that are least google / Like Apples Siri, is telling the most jokes for questions. Even the Google search is telling jokes – Tell the Google Search app –‘make me a sandwich‘ – you get search results but if you want to have Google to read it out for you, it just goes ‚Hah – make it yourself.
What does that all mean? I think humanity will only get used to interact with machines if they get more and more sensitive to interact with. They will start to understand, they will reason with us and ultimatly they will start to learn new things. Google aims with Allo for the fact that you dont have a standart PA in your pocket like Siri – Allo will evolve and learn with your day to day usage until everyone using Allo has their own, individual Allo in their pocket.
Well, someone told me that a talk is only good when you have a luma landscape in there, so here it is .
Was heißt das für den Retail Markt? Schauen wir uns dazu ein paar konkrete Themen an.
If I tell you that supermarkets plan their stock based on the weather it‘s nothing new. Everyone knows that with good weather the barbeque meat will get stocked up and that every chain is distributing their amount of barpeque meat based on regions and the likeliness to have a barpeque.
But that goes down on so many levels. Who of you has a cat? Did you know that if there is a bad weather for couple of days, the amout of sold cat litter goes up? Of course because who whants to go out in the rain.
Or think about the last heatwave you have. If you predict the weather you can predict the amount of stock you need to have for your AC to sell. Most of the stores where sold out....
so people had to find different solutions like this....
I would probaly have done it this way.
a different kind of implementation has been done by the North Face in the US
While shopping on their website you dont search for a specific product but you will tell them, what kind of activity your are planning.
If you run through that process you will get a product recommandation by the end of it that is specificly tailored to your needs. Select the color and lets go – product recommandation made easy. You dont have to think about what works best for you, the system tells you what you should choose.
Airlines – please make use of your booking system more efficient. It‘s a case I made up...
but while you have a user running through the booking system you collect a lot of valuable data points. ... ... and then you show him a car...
But what if those data points tell you its a female, traveling from Munich to Bangkok for 3 weeks, economy, one person, two way... I can clearly see that this person most likely doesnt need a car – it might acutally need a backpack.
Looking into the future – make use of patterns and behaviour while planning advertising target students with coffee ads during exams as this is the most busy time for them.
One usecase of utulising cognitve data we build with IBM and I want to show you a quick video on how productsup is connected to the IBM system to analyse your product data to provide additional opportunities. The watson system hereby collects all internet data and tries to find the relevant keywords for your products.
You have seen we are getting to layers of information back.
Damit bin ich am Ende und ich wünsche ein Happy Cognitive Desrupting
and thats a Donkey
Cognitive technology and how it changes the ecommerce world
a new way to fuel retailers revenue
What is Productsup?
Independently create high quality,
customized product data feeds for unlimited
online shopping and marketing channels and
increase visibility, CTR and ROI
A global player
Processing 80+ billion products each month
Offices in Berlin, Munich & San Francisco,
New York, Sydney
400+ customers worldwide
„Cognitive / Cognition“
"Cognition" is the mental action or process of acquiring knowledge and understanding through
thought, experience, and the senses. It encompasses processes such as knowledge, attention,
memory and working memory, judgment and evaluation, reasoning and "computation", problem
solving and decision making, comprehension and production of language, etc.
Cognitive processes use existing knowledge and generates new knowledge.
WHAT IS COGNITIVE?
Data points collected:
• Start / Destination
• Timeframe of Journey
• Class of flight
• Amount of persons
• Price of flight
Data points delivered:
• Munich to Bangkok
• Journey of 3 weeks
• 1 person
• two-way: 1.500 EUR
Productsup uses IBM Alchemy language
to analyse your products with IBM Watson
Keywords that are directly connected to the advertising
potential of a product
e.g. Nike Sports Bra – "Bra" = 91% relevance
Keywords that are semantically connected to a product
e.g. Mont Blanc ballpoint pen – "Father‘s day gift" = 87 % relevance
First Result – Fashion Retail Case
over 1. Million concepts & over 69.000 new keywords
HAPPY COGNITIVE DISRUPTING
Volker Schmidt - CRO
Berlin, Munich, San Francisco,
New York, Sydney
+49 151 40029679