News 10 Jun 2022
The technical validation is done - we have very high accuracy
The technical validation of Indivd's new unique technology for people counting in stores and malls is done.
- We achieve a very high level of accuracy and we have good hopes of being able to move state of the art with further testing. It is a very satisfactory result, says Leonard Johard, CTO, one of the world's leading AI researchers in Reinforcement Learning and architect behind Indivd's new technology for people counting.
The technical validation is now complete by Indivd's patented technology for collecting anonymized customer data for the creation of unique customer insights for stores and malls. The technical validation was carried out for two days in May with the help of 230 volunteers, who repeatedly went in and out of a grocery store and a café.
- Two different conditions were used for the people counter in the study, one based on older technology and one based on newer technology. This means that the technology will be easily integrated into most stores and malls without expensive technology and installation costs, says Leonard Johard, CTO, world-leading AI researcher and architect behind Indivd's new technology for visitor insights in stores and malls.
95 percent accuracy
Indivd will now, together with the company's very first customer, test and refine the technology in a pilot project in a real physical retail environment in Stockholm. The data collection is also carried out in connection with changes in the environment and product range to study in more detail what the ROI of the changes will be.
- We knew that state-of-the-art is about 95 percent, and now we know that our new people counter will be able to achieve an equal accuracy. We also know that the more people who are measured, the more accurate the statistics become because noise is then automatically sorted out. It is a very satisfactory result, says Leonard Johard.
Amongst all the technologies being used today for collecting and analyzing customer data and creating customer insights, image-based people counting is considered to be the best for understanding what is really happening in the stores. Traditional people accounts, however, have major problems in complying with GDPR and creating reliable customer insights.
However, our new type of people counter is able to gather anonymous demographic, location-based, behavioral, and interest insights without stepping over the threshold of integrity. Individual visitor counters also do not store any personal data about visitors and the anonymized insights that are collected are impossible to trace back to physically living people.
The current result from a unique study in a real restaurant and store environment is considered very promising.
- We have been able to state that the technology does not require any special equipment such as expensive 3D sensors, which reduces the total cost for stores and malls. Indivd's technology is compatible with the technology that is already available in the vast majority of stores and malls. There is a very great potential here, says Leonard Johard, says Dr. Leonard Johard and continues:
This is how the validation was conducted
Dr. Leonard Johard had defined the following conditions, in order for there to be scientifically accurate conditions.
Exactly 230 randomly selected participants were asked to visit first a restaurant and then a store. The participants first had to sign a paper that allowed the participation of the study and the collection of their personal data. All participants were offered a coupon as compensation with free crepes and coffee in the current restaurant.
Deltagarna fick vid acceptans en bonuskupong med en specifik streckkod, som kopplade personen till manuella data och data via vår besöksräknare. Därefter registrerades deltagarna registrerades både manuellt och med hjälp av besöksräknare när de först gick in i restaurangen och när de sedan detekterades när de gick in i den närliggande butiken. Detta gjorde det möjligt att i valideringstestet synkronisera data från manuell inhämtning och via vår nya typ av besöksräknare.
The participants received a bonus coupon with a specific barcode upon acceptance, which linked the person to manual data and data via our people counter. Thereafter, the participants were registered both manually and with the help of our people counters when they first entered the restaurant and when they were then detected when they entered the nearby store. This made it possible to synchronize data from manual retrieval and via our new type of people counter in the validation test.
Participants were asked to take a particular tour of both the restaurant and store. The people counters were placed at special angles and with different lights to make it extra challenging.