Branda

Do you wonder what they are talking about, when you are offline? And can you predict their reaction on your next post? Branda has solution. Branda knows and she can tell you.

Comment

Our Inspiration

As students we are almost every day in contact with the business environment, we are having long discussions about innovation, we do like technologies and when we received this assignment, we put our heads together and the brainstorm began. We looked at it from many angles and wondered what would make life easier for marketers and companies. In a world where informational flow is getting wider and deeper, we understand that it is almost impossible to find relevant data and fight with disinformation at the same time. As a team, we discussed possible ways to create clearer and smarter paths for companies and we figured out that technologies such as AI, Web Scraping, and Machine Learning could be very helpful. We got inspired by this and we created a proposal for Branda.

What it does

Branda is an autonomous web application which is monitoring social media and various website activities and predicting solutions and possible threats to the company. Branda will report immediate threats and inadequate responses to the company's posts. As a threat, we consider for example negative reviews in comments with more than 100 interactions. With Branda’s unique features any company on social media can consider it as another shield against cyber-attacks. Those attacks could be for example Troll farms whose task is to spread disinformation and cause damage to the company. Branda has the ability to collect data from various social media through hashtags, comments, and reactions.

And how does it work?
User X will post about company Y that the company’s product is unhealthy, the post will have 500 interactions in total. Branda will catch it because User X used the company’s #name or keywords in the post, Branda will start with an evaluation and analyze how many reactions and interactions the post has. Based on this search, Branda will consider if it can be a threat for the company and create a report which will be sent to the company. Now it is up to the company to decide how to react. The company’s employee can see all the data, generate a report which he/she can present at the next brainstorming meeting, think about their following steps or strategy, and save the brand.

Branda has of course the ability to monitor and analyze the positive results as well; for example, the marketers can find out what people are talking about and create a great campaign, software engineers can let Branda browsers in forums and find out what customers are missing and based on that create additive features. Due to a large amount of informational flow, the employees of the company have the possibility to filter their notifications (for example see negative posts with more than 200 reactions or positive posts with more than 1000 interactions).

Another great thing about Branda is that it knows the most trending words on the internet, it can improve copywriting and make the company’s SEO unbreakable. The employee can plug Branda as an extension to the browser and always while writing the next post or comment there will be a pop-up window with suggestions. The suggestion could be for example word lasting, this word is not that popular, however, its synonym sustainable is very popular, highly searched and Branda’s innovative predictions shows it will be relevant also in the future! So, the marketer can listen to Branda and change the wording and improve the SEO.

And that is not all! Branda is still learning and educating itself by ratings and feedback on search results from its users.

How we built it

Based on data from Twitter, we have created a database in Cosmos DB. Submitted the information received to the Power BI and then began to work on eliminating duplicates and correcting the information. After that, we used the services of Microsoft to analyze the data obtained. Text Analytics was used to find out what people think of our brand, and we can get results such as “negative”, “positive” or “neutral”. Also, with Text Analytics we find out key phrases which quickly identify the concepts in the text.

We used the key phrases to identify important elements of the review, those elements also can help to determine that this review is positive. The voice recognition feature also came in handy for creating our application. The company will be able to download call records with her customer service and analyze this information through positive or negative keys. This function will help the company understand in what state the customer has terminated the call. We also needed a translator of the text in the conditions of Danish culture, we have the property of many nationalities, which means that comments can be in different languages. We specialize only in Danish and English, which are predominant in Denmark.

For the first representation of how we see this web application, we used Figma. We also start with programming web apps and getting data from social media. Users can log in with social media accounts and thanks to access tokens we are able to collect additional data from Facebook, Instagram, YouTube and display them.

Challenges we ran into

Our biggest challenge was how best to understand the assignment and create the most suitable solution, it helped us a lot when we took on the role of an employee who solves the issue of branding every day.
We've come up with a lot of ideas, but can they all be implemented? Another problem was to find our way around and learn more about the possibilities of Azure AI services in a relatively short time.

Not all sources of information have the same rules, which is why we ran into a challenge if we wanted to obtain the same data, for example from Twitter and Facebook. Facebook has changed its rules to what extent developers can retrieve their data, but after this finding, we always came up with a replacement variation that complied with community policy.

We came across a moment in the Power BI, when we tried to remove duplicates, then all duplicates were removed except for one, respectively, in the analysis, there may be an error, although not great.

Developing a complex project with the implementation of great ideas within 2 days is really challenging and therefore we had to identify and prioritize the most important things even though we really wanted to implement more and more.

Accomplishments that we're proud of

When we are looking back to what we did in the past days we have to say that we are very proud of our idea and the whole process behind it. We found a way to create an innovative product that will improve most of the processes in the company related to building relationships with customers. From creating products based on customer needs to monitor marketing campaigns, customer retention, and ensure security.

Hand in hand with teamwork and the help of mentors, we consider improving our skills in Azure, Python, and machine learning as our biggest achievement.

We are also proud to have completed prototypes and an MVP solution, while working on this MVP, we accomplished the whole process from collecting to displaying data to the user. That is the most valuable thing for us.

We are also proud of Branda's logo and the whole video, which is used for its presentation.

What we learned

As part of the research, we learned new information about Big data and how we can use it for company branding. We also learned and tried different forms of collecting this data (Python script, Logic app, Web scraping). We understand that not all data on the Internet about a company is valid and therefore it is important to validate them correctly (delete duplicates).

During the solution of storing the collected data, we learned about the possibilities offered by different types of databases, and we tried things out for the first time in Cosmos DB.

In order to be able to use AI, we learned to use some Microsoft Azure AI services. In practice, we use Text analytics, but we gained theoretical knowledge from more of them. We also learned more about machine learning and about how it is important to properly train a machine learning model.

Once we tried this whole process of collecting data, we realized that thanks to relatively simple scripts, we can get a lot of information about users from the Internet and social networks, and after this finding, we started looking at the whole thing from the point of view of security.

We definitely learned to work in a group and in a relatively short time use all the resources we have to create a great project. In addition to project management, we also learned more about branding and how important it is for the success of every company.

We practiced our graphic skills by creating prototypes in Figma and we learned to create an animated video using PowerPoint.

What's next for Branda

While working on Branda, we discovered that there are many possibilities for how AI can process the collected data, so we will give just a few examples of what other possibilities Branda can theoretically use in the future.

  • With the help of image recognition on the Internet, it will be possible to find reviews about the company, if the hashtag has not been written.
  • With the help of a voice translator, it will be possible to download recordings of conversations with clients in different languages and then process them into statistics.
Share this project:

Updates