Weborama is an audience advertising company,
targeting consumers and identifying trends
Identifying audiences to understand consumers
The main goal of the Bigfish tool was to visualize data and allow users to manipulate parameters to understand trends.
It made use of a D3.js graphical library to display the collected information as a cloud graph. Big nodes represent the important terms relating to each other, surrounded by proportionally smaller clouds representing less relevant terms.
This graphical visualisation had the advantage of giving a first impactful view of the data. Users have a clear view of main trends and what should be prioritised for analysis.
Clusters conveniently group the data so users can see the different sections and work on them separately.
The tool was flexible enough to let the user play with the data from different angles, creating several views with multiple settings, to get the best results.
EXAMPLE of RESULTS
On the mockups above, we can see an example of clouds around the insurance sector.
The client had a view of how their world was seen by their prospects, having big nodes around "health", "car", "loans", "life" and "auto".
They also had some specific terms like "kill", "japan" or "ING" (another insurance company). These words were a good example of emerging trends that would require analysis and filtering to understand their presence in the graph: were people afraid of being attacked? did they want insurance for their next trip to japan? Was the ING company proposing a nice insurance for that?
The data collected and displayed in this tool not only raised these questions, but may help provide the answers.
Two constrains existed during the development
It could take a long time to collect enough data to have an interesting view for a new client.
The number of websites to crawl was a big factor: more websites meant more time.Generating the graph could required a lot of graphical engine power.
The code used to display the nodes and links could be very demanding for the machine.
If the data set was too large and the computer not strong enough, it would freeze the browser.
In an effort to make the tool as efficient as possible for most users, we provided them with the possibility to customise some settings before generating the graph.But in the end, a great graph required a great computer.
EXAMPLE of FEATUREs
Client feedback was gathered early in the project so the features could be developed and designed with their requirements in mind.
Here are a few of them:
Create cluster and taxonomy
One of the features we developed was to be able to create specific Taxonomy, that would restrain the data analyzed to a limited number of terms selected by the user (see mockup).
This restriction shows the user the words and connexions around the terms they listed.
Filters & blacklist
Some words like "the" or "and" were naturally very present in the cloud but most of the time were useless. So we created a "basic word banned list", that the client could activate or not, enrich or not.
We also needed to give them the ability to filter some parts and exclude words, to have a more coherent data cloud in the end. It was only by studying the graph that the client could really understand what was relevant or not.
See the source and context
The origin and the context of the data were two of the major concern for the clients: we needed to display them and let the user decide to include them or not in the final result.
On this mockup, we listed the source with their share of the results. We clearly displayed their impact with a bar graph in addition to the figures.
Based on this, users could decide to filter the sources to display a more relevant graph.
Jeremy Neveu
designer