UX & UI Design

Big Data Dashboard


Big Data Management with one of the hotel trade industry biggest players. Analyzing travelers flux and hotel performance, then rendering it as visual as possible in order to display easy to absorb information and insights for further decision-making.

Project context & problem(s) at handWho? What? Where? When? Why? How?

The project was part of my 4th year of studies at the _Web School Factory+. The objective was to help an important hotel trade company tackle big data+ and implement it in its management decision-making processes in Brazil. Three main axis stood out.

Firstly: grasp the market. Dealing with Brazil — a dynamic but unstable market — it was crucial to understand both the target and its ecosystem, and how its travel and tourism economy evolved and fared nowadays.

Brazil is a wide and dynamic but unstable market.
With over 205 million inhabitants, Brazil is a wide and dynamic but unstable market.

Secondly, it was necessary to work out how to use the data we could access and how to enrich it. We had at our disposal a 10 million entries database for us to use and were asked to cross-reference it with outside sources of our own choice (APIs, public databases, open data+, etc.). We were relatively free for that matter, as long as sources were viable and accessible.

Finally, we had to provide recommendations as well as render the information generated easily readable and usable by the management teams so as to facilitate their surveillance and decision-making tasks. In order to accomplish that, the only guideline was to (preferably) go for a dashboard solution.

While researching about Brazil, we obtained several important insights from both literature and Brazilians residents, especially regarding the latter's uses in domestic travel and tourism. Contrarily to our first assumption Brazil's domestic tourism represents 95% of the whole industry. This major find lead us to focus our attention on this specific target: the locals. Further research in that direction brought us very interesting studies and data, such as the FIPE report on domestic tourism+.

This extremely complete study, part of the many efforts made by Brazil to improve open data and transparency, has an incredible amount of information regarding domestic travelers' habits and behaviour. Our team quickly decided to make it our main outside source for cross-referencing with our client's database. Seeing how much data on users' we had, we decided to go for the following model :

How to enhance hotel evaluation by putting domestic user [traveler] behaviour at its heart?

SolutionGreat minds don't play tricks but can still do magic.

The answer our team came up with revolved around three elements. A model using data from our database and the FIPE report to express specific and relevant recommandations. Then, a tool (dashboard) allowing for easy comprehension of new datasets. Finally, establishing good practices in analyzing said data to grasp the right insights.

The model

The main data used relate to the travelers flows within Brazil. By identifying the composition* of tourism in each State and hotel, and creating a « picture » of it, we can compare it with the data from our client's database to see if the hotel tourism matches its ecosystem. From there, cross-referencing it to classic KPIs (i.e. occupancy rate) allows us to have a double reading of a State or hotel performance : is it doing well or poorly, and how wide is the difference with the tourism composition of its environment.

* By “composition” we mean the percentage of travelers, depending on their origin State. For instance, State-1 tourism is made up by 10% travelers from State-2, 15% from State-3, 20% from State-4, and so on. Meanwhile our data on Hotel-1 could have 3% from State-2, 18% from State-3, 7% from State-4, etc. thus highlighting a discrepancy. It can be good or bad. So far it's just a discrepancy.

We now had to make that information as visual and understandable as possible. So we started working on the dashboard.

Dashboard v.1

The first version was based on two main features : an interactive map combined with an infographics panel.

big data dashboard v.1
An overview of the country with an alerts panel to focus on important elements.
big data dashboard v.1
The user could access a State and its hotels' informations on various levels of depth.

While the interactive map allowed for visual analysis and offered a first level of information, the side panel displayed either alerts or more in-depth data about the region or the hotels within. However, problems arose with this choice of representation. Brazil being incredibly large, its population density is also highly uneven : some small but highly populated area were much less noticeable, compared to huge but empty areas.

Dashboard v.2

In order to solve these issues, the new version left the interactive cartography behind to the advantage of a more impacting datavisualization. The aim was to put our model (calculating the difference in tourism composition) at the center of the design. In the same way, we got other secondary variables such as the size of a State out of the way.

big data dashboard v.1
Not as important as before, we kept an alert panel for monitoring.

The dashboard now has 2 levels of focus with a top-bottom approch. On a larger scale we display all of the Brazilian States, each item including several layers of information. The datavisualization is the most visible and delivers the core information. Then comes a layer of number data such as the hotel types present, or tourism volume.

Then we can move down to a smaller scale with the hotels present in a region. Same as before, the datavisualization is the main element in each item, completed by additional information.

big data dashboard v.1
Overview of the States of Brazil.
big data dashboard v.1
Overview of the hotels within a selected Brazilian State

And each level of focus possesses both an overview state and a more detailed one allowing to analyze either an item or its environment. For a State, we go though the clientele per hotel or users habits. While the hotel page focuses on breaking down its composition to better understand it.

big data dashboard v.1
In-depth analysis of a Brazilian State
big data dashboard v.1
In-depth analysis of a hotel

Implementing good practices in analysis

A hotel can be doing poorly in term of classic KPIs (occupancy rate, gross profit, etc.) while having a positive gap in its tourism composition. And the opposite situation can occur as well. As we explained earlier, our model focuses on the flows of travelers, their origin and destination, in order to picture the tourism environment. Cross-referencing this outside data with our in-house information gives us a typology of situations that can then be used to formulate recommandations faster while taking into account the context, the environment and not only pure financial results

Hotel KPIs in the green / Negative gap in tourism composition Hotel KPIs in the green / Neutral gap in tourism composition Hotel KPIs in the green / Positive gap in tourism composition
Hotel KPIs in the red / Negative gap in tourism composition Hotel KPIs in the red / Neutral gap in tourism composition Hotel KPIs in the red / Positive gap in tourism composition

Thinking with one more dimension creates a wider panel of possibilities, however creating a typology ought to facilitate reading and processing information.

ResultsQuod Erat Demonstrandum

We aim to offer a double reading for managers, to help them tackle situations and hotel performance observation from a different angle, and fasten decision-making. We received a lot of positive and encouraging feedback for the model, and while the dashboard seemed a bit overwhelming to the client, they quickly got the grasp on how to use it efficiently and provided some additional feedback to make it clearer (focus only on the 5 or 10 first items, adapt an item's size [which was already in our boxes, just not yea on the actual prototype], etc.).

The Dream TeamThis was made (possible) with...

Baptiste Trouillet

Baptiste Trouillet

Laurie Tazaro

Laurie Tazaro

Alain Seng

Alain Seng

Manuel Lemaire

Manuel Lemaire

Antoine Beauvillain

Good ol' me