Images create Emotions…

…and emotions lead to interest. That is one of our thoughts behind Funky Retail. This video should explain what we mean. So please, see for yourself!

Just a short note on the analytics: The data we collect is merely an example to show that it is possible to connect the physical retail space. There may be other data that is more vital. The colourful, flashing lights also have the purpose to show that actions are recorded. We’re not necessarily saying future stores have to flash like crazy in all the different colours…(although we do think that would be quite cool).

Watch the Funky Interview here!

Tiles in color, plus finalized arch poster

Now that we have all major events (except the hybris xmas party) behind us, we can finally focus on getting a few projects really finalized. Tiles  made huge progress over the last weeks and I just got the fully-colored tiles in, plus I have a finalized architecture poster that I want to share with you. Big kudos go out to Elke and DerGrueneFish, our booth building partners for this and most other projects. The tiles (21 in total, for 3 complete demo sets) are colored in 4 fresh colors for a change (no boring white!). I absolutely love the way they look.  Over one day, I was soldering the first 7 which are currently connected to one hub.

IMG_20141208_114334

For the poster, Kathi at SNK did an awesome job. I already ordered our poster which we’ll then present at the hybris summit 15 in Munich at our booth. Having a descriptive poster will greatly help us to explain the IoT setup for this prototype. Right now we expect to have cans on top of the tiles, so we made that part of the poster.

tiles-90x60

 

 

Just to recap the architecture, have a read:

  • “Tiles” are the wirelessly connected platforms. We use Punchthrough’s LightBlue Bean and remove the battery holder to make the platforms 8mm high. We still use CR2032 batteries, which gives us about 1 week battery life right now. We would get more, but I send our a MetaEvent every 10sec which is hard on the battery.
  • The “Hub” collects all data. It scans for tiles, continuously, and connects. The hub runs on the raspberry pi, uses a BLE dongle (choice is key here) and uses node.js for all programming. It sends on data to the server with CoAP – a UDP-based IoT protocol.
  • The “Server” collects all data for all hubs (yep, there can be many) and provides the necessary APIs for managing the User/Tile association, authentication and authorization (Oauth2 used here), etc.

 

One change over the last days was that we can now associate products with the tiles. That means a store manager can just scan a tile (NFC or QR) and then add this tile to his private analytics page. The UI of these web pages is currently being worked on and will feature a few cool features such as a heartbeat every 10 seconds or the color of the scanned tile, that gets pulled via some static, factory-decided data. This system is all up and running now, currently with one live hub and 7 tiles connected.

14 - 2

What’s left is the callback mechanism plus the web ui. The callback mechanism will “call out” to external systems for each event reveived. So if a LiftEvent is received and a webhook is configured, we’ll send out a HTTP Post to the configured external service. I also plan to pull in the product details from YAAS, hybris’ on demand API offering.

14 - 1

 

Tiles Update – we've added blinky blinky

Our project Tiles, little BLE-connected platforms for customer interaction tracking, is entering a project phase which allows me to blog and inform you a bit more. Since yesterday night, the Raspberry PI and Arduino in the hub uses one power source. This makes the overall design easier. We also have been working on a Raspberry PI B+ hat, using Eagle, to further optimize our design.

One visible change is also that it now blinks 🙂 The hub rotates an LED light to signal the BLE scanning process. It flashes once you liftup the product, well, the apple in this case.

IMG_20141126_204216

 

We’ve now also locked down the architecture and below is a rough sketch that should help understand it. Again, a quick summary below.

tiles technical architecture

  • “Tiles” are the wirelessly connected platforms. We use Punchthrough’s LightBlue Bean and remove the battery holder to make the platforms 8mm high. We still use CR2032 batteries, which gives us about 1 week battery life right now. We would get more, but I send our a MetaEvent every 10sec which is hard on the battery.
  • The “Hub” collects all data. It scans for tiles, continuously, and connects. The hub runs on the raspberry pi, uses a BLE dongle (choice is key here) and uses node.js for all programming. It sends on data to the server with CoAP – a UDP-based IoT protocol.
  • The “Server” collects all data for all hubs (yep, there can be many) and provides the necessary APIs for managing the User/Tile association, authentication and authorization (Oauth2 used here), etc.

One more thing – I’ve connected the server to Xively, a data logging platform. We collect mainly the battery rundown to estimate battery life and also the temperature values from the lightblue beans. At this point I just want to share some nice graphs to show you how much sense it makes to track that data. It will definitely help us to optimize the design / battery consumption further. Right now we stay optimized for demo purposes, but we can later reduce the events sent for example to get a better battery life.

Screen Shot 2014-11-27 at 9.49.19 AM

Screen Shot 2014-11-27 at 9.49.13 AM

 

 

Next up: Tiles

I started blogging bits and pieces about BLE – Bluetooth Low Energy – a few weeks ago. The research we’re doing right now will help us to connect little, battery-powered “tiles” (hence the name) to a hub that collects data from these sensors. This prototype is part of our vision for a connected retail space. We’ve now received a first CNC-milled hardware prototype for the housing, which is shown below.

IMG_20141104_175100

 

We’ll now start connecting the bits and pieces. Our overall architecture consists of 3 parts:

  • the satellites –  tiles – contain LightBlue Beans – little arduino/BLE microcontrollers that run on batteries. The tiles also contain a single pressure sensor that is used to detect if a product is on top (or lifted up). The event data contains information about the tiles ID, the event (up/down) and the battery level. The mechanism we seem to support in the end will use BLE notifications that originate from the LightBlue Bean and will be received by the hub.
  • The hub is also still under development, but some early node.js code works nicely on my Mac already. The hub scans for the tiles, which send BLE advertisements, and connects to them. It receives the events via BLE notfications and will have to manage the tiles and pass on the events. The protocol we would like to use for the first time is CoAP in this case. It is essentially a binary version of HTTP, runs on UDP and is – as the name suggests – made for constrained applications. We’ve then successfully used ZMQ, MQTT and CoAP when it comes to IoT protocols.
  • The server will receive all CoAP messages from the tiles, process and persist (or at least keep) the data and allow users for the system – customers, store managers – to manage the tiles. We intend to print a QR code or attach an NFC tag to each of them. Once you touch the tiles with the NFC tag and have passed the OAuth2-based authorization, you can add the tile to your personal analytics view. The goal is to make it really reasy for a store manager to add these ‘sensor elements’ (tiles) to his anayltics view. Once a tile is claimed, the analytics data will not be accessible to anybody else than the person that claimed it.

So… yes, it’s still quite a long way to go.  I’ll try to update you once we make some progress. But let me know what feedback you have. Just tweet me or leave a comment directly on this blog.

Funky Retail

The next hybris labs prototype is coming soon. You might have read about the idea already here on our blog, but there have been a couple of developments. Funky Retail’s the name, in-store analytics is the game. With the Smart Wine Shelf we aimed to improve  the customer experience by the use of an recommendation system based on IoT technology. But we also realised that the Internet of Things offers ways to enhance in-store analytics. This is exactly what we focused on while designing Funky Retail.

On any standard shopping website, retailers know exactly when a customer visits,  know how long he stays, which products he looks at, for how long he looks at them, can recommend upsells and sees if the customer makes a purchase. Why should this not be possible in the physical retail world? That’s what we evaluate with Funky Retail. We identify the presence of a customer in front of a Funky Box; we count the product lift-ups; we measure how long a product is being lifted; and we even combined the individual product lift-up with the playback of an engaging product video.

To spice the whole thing up a little, we collaborated with the hybris customer Mammut. Mammut equipped us with some cool products that help us to put some more life into this prototype and round the story off. We don’t want to spoil the surprise, but we’ll give you a hint: A video shoot is scheduled with the hybris media team, and those guys have got a bit of climbing experience…

Watch the Funky Interview here!

IMG_0064

Let's get funky!

Lights. Colours. Changing colours. Flashing lights. We could just as well open a dance club! But then some genius suggested we should add some analytics…  I mean, honestly…

That slightly spoilt our groovy party mood (and the dancing), but gave us a great idea. We call it “Funky Retail” and it’s coming soon! Sven is soldering day and night to get everything finished. There’ll be distance sensors, pressure sensors and LED’s. Can you guess what we’re up to? Two more clues for you: it will be an in-store prototype and involves a big screen. What would you make out of all of this? If you’ve got an idea send an email to labs@hybris.com. The best idea will receive a reward, consisting of us stealing the concept, giving it a new name, and then writing about it here in our blog. What an honour! We’ll also let you know once we’ve got a patent on it.

FUNKY RETAIL IS COMING SOON! CLICK HERE FOR AN UPDATE.

IMG_20140716_115553

Funky Retail. Analytics & Actions for the physical retail space.

We recently finished our first IoT prototype, the smart wine shelf. It was a big success and it enabled us to have many many insightful discussions with people in the IoT industry and our customers. After a bit of thinking, we realized that it makes a lot of sense to add a few more prototypes in this area. For the next one, we’ll be focusing on the sensing part, e.g. how can we detect customer’s interest in products for example. In the end, the idea is to research how we can gather analytics events in a similar fashion as on a website. On a website, using tools like Google Analytics and others, we know exactly when our customers visit us, what product they look at, what products they take a closer look – and finally when they leave the store again. In between they might have bought a product or not – and if they are soon to leave, we can think about providing incentives to buy: promotions for example.

On a website, the profile you build up yields to real-time changes to the pages you visit. Why is this not happening in the physical world? For example, why does the music in a store not change based on the demographic profile of the people visiting? Why do video ad solutions not adopt to the people in the store? You can add other senses like smells or lighting to this.

Why do we not have the same for the physical retail space? We’ll, we believe with the help of sensors and iBeacons for fine-grained location, we can do exactly that. We’ll take a look at this in our next prototype, called Funky Retail. More details soon, but let’s start discussing this now – as always on Google+.

 

funky_for_blog