Question 1

With the very diversified technology landscape of the world, and the need for integrating all of these technologies to become data driven and truly make sense of the data we have access to, keeping social media in mind as a focus point, how do you see the importance of data integration? Why would we need to go beyond social media data?

Social media technologies, or however you want to call them, have been around for ten-plus years in various forms. There are many companies that do nothing else but show you data from within your Twitter or your Facebook account. Frankly, the business cases around those products have been suffering.

A few good companies have survived, but ultimately, what does it mean for the business that we have X followers? Or what does it mean that there is so much happening on that particular page? The value of the social data or the data from social media interactions grows exponentially once you put it in context with some of your business data.

I think that there are two main aspects to consider, both requiring integration.

One is the marketing impact of social media, which is basically what is coming to you as a business from social media, and can be seen as a very strong case in ecommerce. There is a link on the Facebook page and clicks on a call to action, which leads to the website and creates new customers.

Another aspect is the way I use social media to maintain my community of customers and whether or not their activity within the platform or within that ecosystem in any way correlates or determines their lifetime value for me as a business.

These aspects bring very clear use cases, and obviously, you need more data and to bring everything together, right?

You need to know if a certain person on Twitter or on Facebook happens to be a certain customer. You then need to know what they’re doing, what they’re reacting to, the content they’re consuming, and finally what they buy, how they interact, how they further promote you, and how they influence others. Influence on social media is a big deal as well.

On the second use case more specifically, an interest that is only becoming mainstream rather recently is to learn details of what people are actually talking about. This interest is leading us beyond simple metrics, very often called vanity metrics. How many likes do we have? How many dislikes? Where are they coming from? This shift is very good. I have come across marketers with businesses built on increasing those numbers without ever actually bothering to validate whether or not such numbers have an impact on the business whatsoever.

I think what is really important on social media is that people are sharing thoughts and people are expressing their thoughts. People are using social media as a communication channel, and whether it is one-to-one or public venting, it is important for things like brand image and perception. We see the use of social media changing a lot from what it was five years ago. Companies also follow each other on new trends. One company starts using social media in customer service, as a customer service channel, and suddenly everyone else is doing it. Close monitoring of competitors’ profiles and activities thereon is also becoming commonplace.

On the data and technology side, social media monitoring is becoming more and more important, especially with the onset of more affordable and easier to use natural language processing systems and algorithms. It is easier to run processes in customer analysis, anomaly detection, and we’re talking about anomaly detection within the text, which will show us that people are suddenly talking about particular topics. You can then be monitoring multiple channels at the same time and realize that, Hey, here’s an important topic, there’s something happening in my business. Such “chatter” can be a very strong indicator of a problem or an opportunity.

Question 1 (Interlude)

I read about a recent use case of Keboola with an airline company, where you found major problems in specific parts of the client’s processes. What I thought was most interesting was how you showed that sometimes the strongest “buzz” might not be the most important. We have to go deeper into the data, because sometimes the important element is not as “popular” in a sense. Not everyone is saying the bad things, naturally. Otherwise, a business would fail—that is, if most people hated them. But I saw that type of anomaly detection analysis you mentioned, and just wanted to reinforce how interesting it is…

Yes, that was a great use case, and I can give you an even better one. We ran a test of the same application on a completely different data set. We went into the hospitality sector. We are talking here about mostly very rich data, from channels such as TripAdvisor, Facebook, Twitter, company internal channels, their own complaint lines, and their own communications. In this particular scenario, we put about seven different data sources together, which for this particular customer represented, on a weekly basis, roughly 5,000 text entries, or 5,000 pieces of communication, 5,000 verbatims, as we call them. So nobody in their right mind reads all of those things, and even if they do, they monitor just one slice of it, just one channel. Maybe someone is responsible for the Facebook page and for responding to that, and there might be an entire team doing it, right? Some companies make their business on managing this type of communication for customers.

What happened then was that within these 5,000 texts every week, there were five pieces of text from different data sources where people complained about the behavior of a particular manager of a particular location of a restaurant chain. A very important finding, right?

The low volume, the fact that it’s not a trending topic, makes it something that you would not see on any kind of leader board. You need a sophisticated machine learning algorithm to go in there and say,

Hey, this looks important! I’m not entirely sure why because I am just a machine and I don’t know everything yet, but it is statistically significant enough to offer it as one of the things I want you, the human user, to take a look at and tell me whether or not this makes sense.

Something like that has a massive business impact. I don’t know if they fired the guy yet or not, but this is a great example of an actionable insight, if I’ve ever seen one. There is also the bigger concern there that if five people complained, there were probably one hundred more who noticed and didn’t complain, right?

Question 2

Following on that line of thought, thinking now of a practical case where someone would knock on your door and ask how they could apply such data integration to their business cases. Can you share a practical view of the process and the possibilities of data integration?

Yes, this is precisely the kind of work we do. The use of social media data and even the text analytics example are just a few of the many applications for integration. For us, it is all about operationalizing the data. How do you make the data a part of the way you’re running business? How do you leverage data to the maximal point? We provide the tools for that, and both the aggregation of data and integration with analytics are the “Lego blocks” used to put everything together.

Typical use cases for us will start with collecting the data, which in this last example it was great because most data sources and integrations were available out of the box. When this happens, it is very easy to put things together without the companies being involved, since their primary business is to run ecommerce, or run a hospital, or whatever it is. Their core business is not to work on data integrations. They have us to do that for them.

Question 3

Regarding the investment and effort around integration, some people I speak with are often scared of data integration projects. They don’t understand the potential costs and further resources that they might need to spend on the process, such as time. This can lead to the assumption that integration can be complicated and not worth it. To help bring down this “myth,” what can you share with us in this sense?

I feel we are touching several good points with this question, so I try to look at each point at a time to better reveal the entire picture.

A lot of software and a lot of technology out there are very inexpensive or even free, but it’s free only if your own time is worthless, right? Meaning that if you can spend the time to learn it, to do it, to develop the capabilities, then you don’t actually pay for the technology.

Then you have technologies that are trying to do a lot for you, but you’re paying for it because you’re saving the time. These technologies go through a much more complex development stage because they are trying to “think in advance” for all those things that you might want to do later as a user. The complexity then grows exponentially with the variables here.

Our job as a company is then to make the technology available to our clients, that’s part of our core value. The example of our text analysis of the “chatter” within certain communication channels is something that NSA has been doing for decades, but they spent an ungodly amount of money because the technology wasn’t there at the time. What is happening now and going forward, which is something that happens with all technologies, is that it is becoming much more accessible for significantly smaller companies.

The issue then becomes the availability of skills to work with such technologies. If you are working, for example, for very large enterprises, you can find teams that will have the capabilities necessary, but you will pay them, again, a huge amount of money because they are very rare on the market. That will be harder and harder to do for smaller and smaller players. If the technology in place is requiring, or the solution is requiring, that you have all these skills, it becomes a barrier on its own.

So if I’m running a restaurant chain, chances are I don’t have a data scientist on staff. And if I do, I’m either really big and really sophisticated or really inefficient in what it is that I am doing because that’s not my core business. I don’t understand it. I don’t know how to hire those people properly. So, that’s where partnerships start kicking in. Who is out there? Whose primary business is this? And who can help me do that? And whether it can be done in a consultative basis, or on a type of OEM, which then becomes a part of the product offering, will be part of the different business decisions that come into play, and they are very specific for each customer. There is no “one size fits all.”

What we’ve done with the approach that we have at Keboola is take all those pieces that everybody needs, such as the database behind the scenes, and the API layer for the integration—like the API integration we have with quintly, which makes it easy to get data from there—and then provide this as building blocks, or a “Lego set” to clients. From there, we can either train the client or we can seriously lower the amount of skills and time required to put these solutions together.

We also do that without being bound by the limitations of one technology product. When you have one specific analytics solution, it will give you, for example, “these five data sources, and then run this type of data analytics on top of it, give you this type of dashboard,” and so forth. The flexibility we offer is great for our customers because they can choose the data sources and workaround issues with any data sources that are missing. We use this logic of the “Lego blocks” to be able to put the necessary pieces together to build what it is that the client is trying to do. The “Lego blocks” analogy is also a good one because, just as in playing with real Lego blocks, with a little bit of training “anybody” can do that. Now, keep in mind that there are big quotation marks on both sides of the word “anybody.”

The skill requirements are dropping dramatically, though. So you don’t need an architect type of person; you need people who understand what is it that you are trying to do with the data and have the basic tools at their disposal to put it together, because all that complexity is already hidden from them.

One simple example of this effect happens in voice over the Internet. You can very easily send voice over the Internet, right? But it’s actually a very complex task. You can do it because you have an app that is doing it for you. It was not built by you, it was built by someone else, but it brought the task to a level in which you can do that without having to know how to do it. The skills required were reduced to pushing a button. And that’s exactly the same principle.

Question 4

Thinking about the role of marketers facing the use of such technology and considering that most marketers do not come from a very technical development background, how do you see them dealing with such challenges, and maybe taking the role of a strategist, understanding enough from the technical side to still be able to put things to work under the demand for technical solutions?

Don’t you take that role of strategist for any other aspect of your business that you’re not personally specialized in? You will figure out, I need this skill set to achieve this goal, I don’t know it myself, so I’ll hire someone. Then you can think later on, Do I have enough of that skill in my team to hire someone internally? Or is it too specific?

We can take a very simple example and look at the most basic infrastructure of business, which is accounting. You obviously need it to run your business. You, as a businessperson and not being an accountant, have a rough idea of what you need from it to run your business. You then think in details. Do I need an in-house bookkeeper? Am I going to hire an accounting company? Do I need a CFO? There are multiple ways of solving the problem depending on what type of organization you are and how you will satisfy this particular need.

And frankly, data is becoming an asset that flows through the company in a very similar way to money, so you need to be managing it in the same way. You must understand enough details to be able to judge if the opportunity can be satisfied by an external organization, group, team, or advisor helping you, or if you need to internalize it. The internal elements will then branch out into multiple aspects, which are mostly different for every single company.

We call this the soft data stack, which is the people side of this process within the organization. You have the soft system and the hard system, terminologies from 20 years ago. I guess now I dated myself. But you have the same thing in data. You have the technology, but then you have the people. The people have different roles.

Some of the roles you might want to bring into the company. Some of them you might want to outsource. Some of them you might want to automate. It’s exactly the same principle of what you’re doing with financial management in the organization. Do I use an automatic billing system? Or do I have two people whose only job is to write invoices?

As a marketer, I would then be looking at it from the perspective of the usage of data. How to make sense of the data? How to draw the conclusions? Or how to draw the lines between what the data is telling me and what the business is? How do I leverage the data to make a better business decision? These are not technical questions. It doesn’t require coding. That’s not what you’re doing. This is the human component of interpretation of the data.

Then there will be a lot of things happening behind the scenes, which will be allowing you to do what you need to do, which will be serving you the data and giving you ways to make use of it, and so forth. An intriguing point is that the technical knowledge required to be on the marketer side is decreasing, while such knowledge for the development and technical side is increasing.

So, the answer for someone coming from the marketing space and being asked to make data-driven decisions or make data-driven recommendations is that they have to understand the interpretation of the data and the impact of it to the business. They do not need to become data engineers, which is a role somewhat further down the stack. The data engineer will be making sure that the infrastructure is doing what is expected from it and that the updated data shows up the next morning, which is a totally different task from the marketer. The engineer will have a limited reach regarding an understanding of the ultimate use case of the application and if it is being used in marketing, ecommerce, or hospitality analytics. For the engineer, it is, in a sense, “all the same.”

Question 5

As a last question, thinking about the diversified technology landscape we have and current need for integration, how would you see integration companies such as Keboola and the integration process itself in the future? Is it going to become more popular and less scary for people that don’t work with it yet? How do you see the future in this sense?

The understanding that the data is an essential element, and that working with it properly is as crucial as managing cash within any organization, becomes a business reality. If your competition knows more about what your customer thinks about your product than you do, you are obviously at a huge disadvantage, going back into our example of text analysis. That means that the demand for this service is going to grow, which is going to drive the cost down, because the technologies are going to become more and more efficient.

A few years ago, we worked with a company that used humans to tag every single message coming through. They then had other solutions built on top of that enriched data. That is obviously expensive. Their customers still happily paid for it because they understood it was giving them the competitive advantage over smaller competitors who were not able to afford it. But guess what? Now, and going forward, that type of technology, together with machine learning, will make such tasks repeatable and affordable using the out-of-the-box solutions that are suddenly becoming available.

In conclusion, I believe we are going to see two things: the technologies making themselves more accessible and more data products—meaning that in situations where we are talking about something the company should build, there will be someone who will be selling it to them out of the box.

We will also see an expansion on availability of data. As an example, it can be difficult to get certain data from Yelp at this point. Why is it difficult? Well, because Yelp understands the value of their data. Sooner or later, however, they are going to come to the market with the idea that, Hey, this is what people are trying to do with our data. Voilà! Let’s sell it to them. When they do that, suddenly anyone who is willing to spend a little bit of money to get some of that data toward their insights will do it. Data providers will also come into play, facilitating the process of data collection. So it makes sense for companies like Yelp to do this, because it adds more and more value to their business.

Add to that integration capabilities of the various technologies. A comment made by someone on public channel can directly trigger a process within the organization, customer’s favorite conversation topics can be provided via the CRM system to sales staff, a chat can be triggered post-purchase to collect customer feedback via their preferred channel.

As a last thought, something that will not change is that no data lives alone, once you put it into context with other data sources, with other information, you will start to make sense of it, find insights, and see the broader picture.