We’re used to seeing the reviews that other consumers leave on Google before booking a table at a restaurant or a rural getaway. But what do users say about the supermarket on the corner? Do they really say anything? Do they take the time to rate their experience in the store, even if it is a regular, even routine, event?
The answer is yes, they do. Google is completely integrated into our lives and we have a habit of sharing everything that happens to us on digital channels, even the most everyday moments of our day-to-day lives.
Google also makes this dynamic much easier: a business only has to sign up for Google MyBusiness for any consumer to share their opinion about it. Receiving first-hand feedback from users (the more positive reviews received, the better the positioning and visibility within all Google services will be), along with the possibility of interacting directly with them, are some of the most valuable contributions of the Google MyBusiness rating system.
But it’s possible to go one step further: extracting true intelligence from social data to improve business performance.
“Extracting true intelligence from social data to improve business performance”.
Valuing the customer sentiment
The DIA supermarket chain, one of Good Rebels’ long-standing clients, was aware of the importance of valuing and positioning its brand within the Google ecosystem. For this reason, in 2017 we began a project that was unprecedented in Spain within the food distribution sector: to obtain an individual assessment for each of the more than 1,500 Dia stores and supermarkets distributed throughout the country based on the opinions and comments given by consumers on Google.
The data from the Google MyBusiness API is composed of quantitative (one- to five-star scores) and qualitative (any opinions expressed in a text) ratings. However, there is not always a correlation between the scores and the sign of the opinions: four stars may be accompanied by an unfavourable opinion, or a good experience is scored with only two stars. Neither statistics nor science would be able to offer a logical explanation as to why this happens, but it does happen.
Thus, we resorted to NLP (Natural Language Processing) technology, which allowed us to extract the feeling expressed in each comment (positive or negative), regardless of the associated score. For this task, we used a combination of software created under our own development, programmed on Python, and the commercial software package BigML, in charge of the most tedious part of the analysis.
In order to unravel and optimise the information flows, we classified all the reviews according to their category (Shops, Products, Club DIA, Ecommerce, Opportunities, etc) and typology (Information, Suggestion, Complaint and Praise). We used this classification both for the preparation of the monthly KPI report and for the internal analysis carried out by Dia’s managers.
Data visualisation for better understanding
Once the themes associated with each review had been extracted, we incorporated a key element to the analysis: CARTO, a powerful platform used for the mapping and visualisation of large data packages.
When working with large data packages, it is important to translate the data analysis into an intuitive and simplified graphical representation. No matter how complex the data extraction and processing may have been, the visualisation must always be simple and easy to understand. Keep in mind that it will not be data specialists or scientists who will be confronted with the final product, but decision makers.
CARTO meets these requirements, and that is why it was the tool chosen to georeference the reviews collected in Dia establishments throughout Spain, which would allow us to identify and locate the opinions, comments, complaints or praise associated with each establishment with absolute precision.
The maps created in CARTO incorporated a series of dynamic filters to interact with the data according to different parameters. There was also a time scale showing the distribution of comments over time.
Through this analysis it was possible, for example, to observe the effect of seasonality on customer interaction: how promotions launched in the months prior to summer were valued, or how the influx of people into establishments in different areas of the country varied during holiday periods (decrease in large cities, increase in coastal areas).
In short, this service made it possible to discover and evaluate customers’ real opinions and to translate this evaluation into executive decisions aimed at improving their satisfaction with the brand.