2. Who are we
Synthetic Data is a leading Big Data solutions
developer company from Spain
The company has had projects in Telefónica, Banco
Santander, Mercedes-Benz (Munich), Alkol Biotech
(UK), etc
Its focus is IoT + Big Data + AI, which it sees as an
indivisible unit
Its founder, Al Costa, is the author of book on Big
Data and professor at Madrid´s EOI- Escuela de
Organización Industrial
Here, we developed a spin-off called “Produtrak”
which we Will demonstrate here for technical and
investment purposes
3. What is ProduTrak?
Produtrak is a project being undertaken by Synthetic Data which uses machine
learning to recognize products at the disposal site.
The idea is to identify where they originated in order to find metrics such as the
amount of time it takes for customers to use the product, where they prefer to
purchase them, how do they use it, etc.
Thus, it is able to provide manufacturers with useful data on the complete
lifecycles of their products from the time they leave the store shelf to the time it
is disposed of.
For this, Produtrak is able to identify the product as it is being disposed using
machine learning libraries and training models.
4. How does it work?
Produtrak works by learning the shape, color, etc of a target product
To that end, the product is photographed in different positions, with different
amounts of liquid, dirty, label worn off, etc. A minimum of 300 pictures is
required for a proper identification
The system works with a videocamera attached on a support on top of the belt
carrying waste of recycling centers, identyfing products as they pass below
In these, each container is identified and thus it is able to backtrack each product
to the precise spot where it was dumped.
To that end, Productrak is speaking with leading recycling companies in Spain
such as Ecoembes in order to use our technology on their operations to offer this
service to consumer product companies
6. Applications
1 – Analise consumption patterns
Being able to know how long a product took to be completely consumed, where
it ended up in comparison to where it was purchased, how deteriorated was its
packaging when disposed, etc allows for a better understanding of consumer
patterns and thus a better product
2 – Identify sales opportunities
A supermarket which does not carry a product (for example, a shampoo) which is
ending up in a nearby dumpster is missing out on a sales opportunity
3 – Substitute use of plastic
Plastics are becoming an increasing source of marine contamination. By
identifying the amount of shelf time, it is possible to replace it with paper-based
packaging options
7. Challenges
By far the biggest challenge in this model is the proper identification of products.
This, because:
1 – Product label may be worn off or just inexistent
2 – Product may be too dirty or deformed
3 – Product may be covered by others
4 – Product may be in a position not previously trained
5 – Packaging could be transparent (PVC, glass) and thus hard to see
6 – Product could be similar to a competitor (se pics below)
8. Conclusions
1 – Produtrak completes the lifecycle analysis of a product. If we could before know
the lifecycle only until the product is sold, now we know it until it is consumed and
disposed of
2 – To this, only through optical recognition with machine learning we are able to
monitor 24x7 many dumping sites
3 – The data collected can then be matched with data from the stores which
actively sell the product
4 –The end result is knowing how is a product really consumed in order to improve
it, lower its production costs, and compete better in the marketplace