A completely new player has entered the local, fast-growing Big Data and web analytics market. Recently, Valkea Media introduced Viewell – innovative software for optimizing and reporting content on social media. WBJ asked its experts about merging Big Data with marketing and new services for clients. Here is what Jacek Krawczak, chief data scientist at Valkea Media, Łukasz Trębicki, head of Social Media Analytics, and Franciszek Rakowski, data science consultant at Valkea Media from the Interdisciplinary Center for Mathematical and Computer Modeling at the University of Warsaw had to say
WBJ: How long have you been developing the new big data tools?
Jacek Krawczak: Our work on the entire tool set for social media analytics began nearly 18 months ago. We started with Facebook, because it is the most frequently used media channel by our customers and we have just finished this stage of our project. Perhaps not completely finished, because we are still fine-tuning the software, but we have reached a stage where we have achieved the planned functionality and implemented the tools for production use in our company.
What is included in the package?
JK: Viewell includes tools that, on the one hand, help our copywriters to create a post that, when published, gets a lot of engagement. And on the other hand, they will help identify the influencers – people who have the greatest impact on how posts spread through the social network.
The main component of our solution is a predictive model that predicts whether a post intended for publication will engage users. The applied machine learning algorithms analyze both the content of the post and its other features, but also consider the planned date and time of publication.
You are not the first to offer software with similar functionality. There are programs that show users the dates and times when it is worth publishing, what punctuation and emoticons to use to get the most traction, etc. How is your product different?
Łukasz Trębicki: Yes, fortunately today it is already quite an obvious solution for any marketer who even minimally supports his brand’s communication with web analytics tools, but we have not only expanded the features you’ve mentioned, but we have also enriched them with unique market-wide and previously unavailable solutions. The functionality we are most proud of is the module that predicts if the content you just wrote will effectively engage your audience before it is even published!
This is an absolute game changer that brings the existing rules and models of copywriting work and social media analysts to a whole new level. It sets them on a whole new course of development, based on artificial intelligence and advanced prediction models. It allows marketers to support human creativity with the power of modern calculating machines.
You don’t often use terms like “machine learning” and “predictive model” when talking about copywriting, which in essence is a very creative and human-centered concept. What is machine learning and what does it look like from the user’s point of view?
JK: Machine learning is about giving computers the ability to learn without being explicitly programmed. It applies in cases where there are a large number of variables and it is difficult or even impossible to prepare a set of rules on which we could base the algorithm of a “traditional” computer program.
On the other hand, from the user’s point of view, it is quite straightforward. Just fill out the table with the calendar of posts – enter the planned text, the publication date and time, and some additional information, then run the program.
Now the machine takes over. It takes a moment to calculate the results: the probability that the analyzed draft/post proposal is likely to generate good engagement result for fans/subscribers/followers.
What is the effectiveness of this solution?
JK: The results are really very good. Otherwise we wouldn’t be using it in our daily work. For the past four months we have been monitoring the predictive model for all of the Facebook profiles we hold for our clients. For most of them, the accuracy of prediction was in the range of 70-90 percent, and in some months the model flawlessly predicted, on several occasions, which posts would have the greatest interactivity.
What about the other solutions you offer and their performance?
ŁT: They are equally useful for a professional marketer. They allow us to find out at what time our post will gather the most engagement, and when it could have the largest range. It tells us what the optimal number of words per post is, which words we should use, and which of the words that we have used so far could potentially lower the appeal of our message. We can identify potential influencers. We have packed so many functions into the program that I can only name a handful of the most important ones. It’s better to just take the program out for a test run.
Are the solutions currently offered on the market, including your program, user-friendly for a “nonanalytical” person? How can they be tested?
ŁT: Generally they are. Developers of data analytics solutions realize that on the other end, you reach users accustomed to simple graphic interfaces. Google knows it and that’s how their Analytics functionality is designed. So does Facebook with its Insights panel. Finally, developers themselves know that too. Semi-professional users should reach the desired information in a just few clicks. Viewell is no different; it is one of the most intuitive analytical tools on the market.
What exactly does that mean?
ŁT: All data is returned in graphical form and in interactive graphs, which is still not standard on the market. Almost all visualized data is clickable, configurable and updated in real time. The interactivity of the tool and the feedback we received from our corporate clients made us decide to add an extra module for reporting brand presence in social media.
What does it do?
ŁT: Our analysis of the market shows that the current analytical tools available on the market cause many problems in the relationship between the client and the marketing agency. The biggest ones are:
- results are reported with insufficient frequency and with significant delays;
- data is presented in such a way that makes it difficult to understand and digest, e.g. static PowerPoint presentations, PDFs or spreadsheets;
- the systems don’t offer enough customization and reports are often overly generalized.
Analytics has to evolve, and we believe that thanks to Viewell we have accelerated this process. Our solution eliminates all of these problems. A complete report summarizing the entire previous month is already made available to the client on the first calendar day of the new month. Over the next few days, we only send written recommendations for changes resulting from manual data analysis.
What are your plans for analytics?
JK: Further development and improvement of created tools. Social media platforms are still very young: Facebook is 13 years old, YouTube is 12 years old, Twitter is 11, and Snapchat is only six. The social media landscape is constantly changing and new platforms are being born. The tools to analyze them also need to change in order to maintain their usefulness in the daily work of the marketer.
What are the benefits for a commercial company from collaboration with scientists?
Franciszek Rakowski: Scientific work allows us to get an in-depth understanding of a wide range of analytical methods. Data science methods have been used successfully for years in important research areas such as genetics, proteomics, medicine, and also in an area that is close to my heart: cognitive neuroscience. It turns out that you can successfully transfer the knowledge gained from your research to the field of commercial activity.
How is scientific work different from business practices in this matter?
FR: First of all, what counts here is reliability, simplicity and efficiency. Researchers may develop their methods and algorithms in a direction that is interesting and valuable to them, but that may not be worth pursuing in business practice. Often, sophisticated solutions have to give way to the more effective ones as well as those that clearly translate into customer satisfaction and company profitability. These are, as the mathematicians say, the additional constraints imposed on the model. They can be practical, business and social in nature, which makes the work even more interesting.