NeuroApplied: A Cognitive Solution to Determine the Perception of Brands

Please note that the following article has been posted on behalf of NeuroApplied, a young Israeli startup I have been working with throughout the last couple of weeks and months.

What is NeuroApplied?

NeuroApplied, a startup in the field of neuro-marketing, provides a cognitive solution to determine the perception of brands.

The solution bridges the gap between Nobel Prize winners’ theories and typical, everyday, demands of brand managers. NeuroApplied applies machine-learning-based analytics and neural-network techniques to enable clients to tap into the associative subconscious of potential customers.

By letting people play a simple game NeuroApplied analyzes subconscious brand perceptions allowing to improve products and associated marketing efforts.
During the (mobile) game a player is usually being shown a set of (fictive) persons, an object (e.g. a product) and a simple question. At this point NeuroApplied already knows about traits usually being associated with each of the single persons being shown.
Questions are like “Do you think this person would like this product?”.
By swiping (to the left or the right) a player can answer each single question being asked with yes or no. By doing so they associate the traits underlying each person being shown with the object (i.e. product) being shown.

NeuroApplied’s SaaS-based solution provides access to an intuitive dashboard revealing insights about brands’ perceptions.

Why OpenWhisk?

As the game is usually being kicked off at a single point in time and then simultaneously played by hundreds and thousands of users, the platform had and has to be able to handle high peaks of usage. In order to meet the demand of high availability while also lowering server and resources costs, NeuroApplied decided to use IBM’s OpenWhisk platform. OpenWhisk’s serverless platform allowed to implement and operate the solution in a flexible and cost-effective way.

How Does It Work?

OpenWhisk provides a Function-as-a-Service (FaaS) platform to execute application logic in response to events. Thus, companies such as NeuroApplied, are enabled to create and run event-driven apps which scale on demand and which are able to execute code in a highly scalable serverless environment.

Phase I: A beginning of the game

A new game starts once an invited potential player clicks a dedicated URL. This click represents an API call triggering the execution of an OpenWhisk action.

The action selects n (usually about 20) out of all available (fictive) persons (stored in a database) to be presented to the particular player. The persons (and hence the traits) to be presented are selected in a way (using some smart algorithms) allowing to learn most about the shown objects’ (i.e. products) perception.

The action responds with a JSON object containing the n persons to be shown as well as the previously mentioned objects (i.e. products) and questions.

Phase II: Playing the game

As said, the player has to swipe to the left or to the right to respond with yes or no to each single question being asked. Each swipe of each user for each shown person represents another API call triggering the execution of another OpenWhisk action.

The action updates the database storing all players’ responses to particular questions.

Phase III: The end of the game

Finally, once a user has finished the game (which causes yet another OpenWhisk action to be invoked), some validation is being performed before the data resulting from the particular player is applied during the overall brand perception determination.

The Bottom Line

Although the first request’s execution time may be a bit longer — about 500 milliseconds — subsequent execution times are about only 70 milliseconds or even less. Moreover, OpenWhisk enabled NeuroApplied to cut operational costs and, additionally, increased the user experience to be fast and flexible.

Therefore, based upon the implemented architecture, NeuroApplied could improve (game) response times by about 23%. In addition, operational costs could be lowered by about 38%.