Big Brother is watching… From the fridge?

Data Architecture 18 May 2017

Smart object: what are the current trends?

Research company Gartner estimates that some 8.4 billion objects around the world will be “smart” by the end of 2017, and that this number will increase to 20.4 billion by 2020. This is a movement undergoing extraordinary growth, and at the heart of today’s trends. This was evident at the 2017 Consumer Electronics Show (CES) in Las Vegas, which was full of presentations for new products in the general public electronics industry. This trade show, at which some 4,300 companies were present, saw smart objects taking centre stage. In his annual recap of the CES (available here, in French), Olivier Ezratti, former Marketing and Communication Director at Microsoft France noted that the most promising areas in the Internet of Things (IoT) industry are sports, household, and chatbots. CES attendees this year admired smart insoles from ATO-Gear, which provide analysis after a jog and tips to avoid injury, the SensorWake alarm which releases a scent of choice (chocolate, seaside, or grass, to name a few!) to gently wake the user up; or even Ubtech’s new Lynx robot, which uses Amazon’s Alexa personal assistant to help with everyday tasks.

Does every smart object have a clear purpose?

It is usually pretty clear what everyday objects are for – or else it is easy enough to figure out. However, the “smart” dimension sometimes makes it difficult to understand or imagine what the added value of a given smart object might be. For example, take the new K’Track Glucose watch from PK Vitality, the biotech subsidiary of the start-up PK Paris. This new smart watch gives real-time updates on a user’s glucose levels without painful finger-pricking: a clear added value for those suffering from diabetes. However, something like LoveBox, a smart wooden box that sends secret messages between partners at great distances, falls more into the “Cute/Gadgety” category, as sending private messages is already possible by phone or e-mail.

Finally, what is a smart object?

They’re the words on everyone’s lips, but it is fairly difficult to give a precise definition of a “smart object.” These objects are all a little bit out there, with a purpose that is more or less useful, but they all have one thing in common: smart objects capture information, and send it to a “smart” interface, whether this be human or technological. The user can thus analyse data gathered by the object to understand its impact on daily life, and make improvements.

How can a regular object be turned into a smart object?

At fifty-five, gathering data on websites and mobile apps is part of our daily grind, which is why we were immediately keen on the idea of gathering information from the “real” world using a regular object and transmitting it to an interface. So our teams undertook a study of all the possibilities for tracking, identifying different KPIs for the objects in our offices.

We weren’t able to track the use of meeting rooms (too ambitious for a test), nor the office badge-reading machines (not anonymous, and personal data is not allowed in Google Analytics). In the end, we decided to “track” use of one of the fridges in our break areas: a simple and inexpensive implementation test, with an object used frequently every day.

Before beginning, we came up with a list of questions that we wanted the test to answer:

  • How many times per day is the fridge opened?
  • Does outdoor temperature influence frequency of use?
  • At what time of day is traffic highest?

To respond to these questions, we decided to set up a Raspberry Pi 3 with internet connection along with a sensor on the fridge door. We set it up so that each time the door was opened or closed, Google Analytics recorded a page view. Thanks to the OpenWeatherMap API, which gave us extra information such as outdoor temperature, weather conditions, atmospheric pressure, and air humidity, we were able to gain enriched data with additional dimensions.


Some analyses following the test

During the week from January 30th to February 3rd, 2017, we saw, for example, that the fridge was used most frequently on Mondays and Thursdays — in other words, at the beginning and end of the week (on Fridays, employees generally finish their work day with a drink together at the local watering hole).


On average, the door is opened and closed 120 times a day. We can thus assume that 60 people use the fridge each day. We are, of course, aware that this data could represent 30 people that use the fridge twice a day, or 20 people that come three times, etc.

Now let’s look at what times of day the fridge is most used:


We weren’t surprised to see that many fifty-fivers come to the fridge around 9:00 AM every day — either to take a drink or to drop off their lunch (which they come back for around 1:00 PM) — according to our hypotheses and observations. We also note peaks in use in the afternoon, between 3:00 PM and 6:00 PM: time for a mid-afternoon snack!

At the outset, we hypothesised that minor changes in outdoor temperature wouldn’t have a big influence on fridge use. The table below shows interactions by hour (8:00 AM – 8:00 PM) and temperature (C°).


In the winter, if at 9:00 AM it is 9°C, and if at 1:00 PM and 5:00 PM it is 12°C, we can assert that 55ers may use the fridge more frequently. Contrary to our initial hypotheses, it would seem that outdoor temperature does indeed influence use. To refine this conclusion, we would need to run the experiment for longer to gather significant statistics. We would also need additional and more granular-level information, such as room temperature. We could also imagine placing sensors on other fridges in other offices, to perform differentiated analyses based on fridge location. This way, we could confirm or reject our hypotheses, and determine if the results show a simple correlation or, rather, a causal connection.

We wanted to explore the possibilities, and gather information from different sources. Some information proved useful for our experiment, such as outdoor temperature, while others such as pressure, humidity, or wind speed were not useful. We made the mistake of wanting to collect data without a specific goal, thinking we could treat the data later. The logical next step would be to think about the information that we would use again in a next iteration, and any other information we would add. In the future, we could take things further by putting other sensors in the fridge, perhaps to measure fridge inventory for example. Using this kind of information, fifty-five could plan to re-stock the fridge when inventory is running low, perhaps with an SMS alert to the person who fills the fridge (thanks, Wizard!) when the time comes. Imagine the possibilities on an even bigger scale – like for large supermarkets!

This first phase of this experiment helped us to get familiar with the tools being used. We are encouraged to continue with a sprint, following some principles of AGILE methodology. Repeating iterations over the coming weeks will help us to confirm or reject our hypotheses, and to avoid collecting too much irrelevant data. Now that we know that making a smart object is similar to tracking a website, there’s only one thing left to say: watch out, fruit basket and water fountain!

Would you like another cup of tea?