Tech Insights: How to milk the big data cow?



Online advertising spending has been growing for years. In Poland it increased by 7.7 percent y/y in the first six months of 2017, reaching PLN 1.858 billion, according to a study conducted by PwC for industry association IAB Polska. “You can clearly see that online advertising spending has been growing the quickest in video, social media and classifieds. Ads sold in the Programmatic model are also recording strong growth,” Włodzimierz Schmidt, CEO of IAB Polska told PAP. The industries with the highest share in online advertising market were retail (18 percent), automotive (11 percent), telecoms (8 percent), finance (7 percent) and foodstuffs (7 percent). “The list of industries with the highest online ad budgets doesn’t change much. There are no huge moves. It’s more about constant trends, and changes are caused by the seasonal nature of marketing and sales activity,” Schmidt added. What is changing, however, is how the effectiveness of online ads will be assessed. Recently, IAB Polska introduced a new viewability standard for internet ads, where “an ad is considered as viewed if at least 50 percent of its pixels are displayed for at least one second for a graphic ad and for two seconds for video ads,” Schmidt explained.

I know what you’re doing

Given how competitive the online ad market is becoming, it comes as no surprise that 81 percent of companies already declare that they employ user data in their marketing activities, according to a recent report titled “Data-driven marketing benchmarks for success” prepared by Ascend2 and Zoominfo. Big data marketing ranges from profiling clients in online advertising, through direct marketing and sales calls, to the use of social media in marketing. All of these activities are aimed at driving up revenue, and all of them need to be used wisely. There is no doubt that companies already have tons of data at their disposal. The abundance of information is one thing, but learning how to put all the pieces of the puzzle together is what really matters. “The key to success is not only an analysis of relevant data sets, but above all, an appropriate combination of its own data with external data,” said Szymon Szmigiel, Head of Commercial Sales at TogetherData. Collating data from multiple sources allows companies to implement new marketing strategies, analyze customers’ behavior and find out what their interests, preferences and purchase intentions are. “We’ve noticed that companies are beginning to invest in their own Data Management Platforms (DMP). These platforms allow them to integrate data from multiple sources – websites, media activities, CRM, sales systems. A lot of the data from these sources can also be used in advertising,” Szmigiel said. According to IAB Europe’s “The economic value of data-driven marketing” report, 86 percent of behavioral data is already used in programmatic advertising. Call centers are another example of where behavioral data analytics can become a huge help. Who isn’t annoyed by cold calls that catch you in the middle of a meeting or when you’re trying to write an important and compelling email. But there are ways to know if your potential client has the time to take your call. You can check when they are idle or looking at funny memes websites. That’s when you call them. This simple trick managed to get the response rate up by 8 percentage points for the call center in question. If your call center agent makes, say, 200 calls a day, it could generate 16 more “leads” every day.



Scouring the web 12_web_edit

100 milliseconds is not a lot of time, yet that is how long an auction for an advertising spot takes. With Big Data analytics, advertisers no longer have to choose randomly which spots to bid for. And as many examples show, intuition can be flawed: if you want to sell men’s deodorant, don’t advertise to a sports fan but to his wife. WBJ sat down with Piotr Prajsnar, CEO of Cloud Technologies, to talk about the real life applications and the surprising effects of BigData in advertising
Interview by Beata Socha

WBJ: Your company claims to analyze over nine billion internet accounts on a regular basis. How can this huge amount of data be monetized?

Piotr Prajsnar: We got into Big Data after years of experience in data mining and internet advertising, an industry that absorbs new technologies like a sponge. AdTech (Advertising Tech) has been around for over 10 years and in 2010 AdTech firms went through a revolution, by means of so-called Programmatic Buying, an automated auctioning system for internet advertising. The way internet advertising now works is as follows: an internet user opens a website and at that moment the browser contacts an auction system, where an auction between different advertisers takes place. The winner of the auction is the one who is willing to pay the most for the banner at that moment and on that website. The entire auction takes about 100 milliseconds.

So how do advertisers decide which websites and at what times they want to advertise?

That is where we come in. We analyze over nine billion internet accounts and we gather data that makes advertising much more effective. We are a ratings agency for clients of sorts. We can tell advertisers which user is looking for a laptop and how much a given user is willing to spend. A banner displayed for that person is worth much more to a laptop retailer than to, say, a shoe store. Advertisers would rather spend more on advertising but in a more efficient way, reaching people who are actually planning to spend money on products that the advertiser offers. By providing information about internet users, we bring order to the internet advertising ecosystem, which also makes advertising far less annoying.

How much do you know about the users you analyze?

First of all, we analyze devices, not people. Actual users are all anonymous. We analyze internet traffic on more than nine billion devices. We collect the data and then we analyze it. We create profiles, e.g. of people who visit automotive websites. Before Big Data analytics, advertisers had to target their clients by  advertising on themed portals, so a car retailer would advertise on a car portal. But knowing which car brands the user likes makes the ads way more effective. And we can provide that information. Still, we are not an intelligence agency and we do not record the data: we usually “forget” it after 30 days. Internet information is very transitory, anyway. The lifetime of cookie files, which we analyze, is about a week. After a month only a tenth of them “survive.”

Is data from social media more useful than scanning internet activity?

I wouldn’t say it’s more useful. Our data covers a much higher percentage of internet users. There are few places we don’t have access to. Social media, like Facebook and LinkedIn, protect their data. They may provide more “calorific” data, but they only cover a portion of the market. For instance, there are some 25 million internet users in Poland. Meanwhile, there are over 10 million Facebook accounts in Poland, and only some of them are active. You have no intel on the rest of the market.

Google is planning to incorporate an inbuilt adblocker into its next version of Chrome. How will that affect the advertising market?

In the long term adblockers could eradicate free content on the internet. Poland is the leader in adblock activity. Nearly 50 percent of website views is generated by people using adblockers. In Poland advertising spend per user is nearly PLN 150 a year. If advertising becomes ineffective because all ads are blocked, users will have to pay for the content themselves. The advertising market sawthe news of Google’s plans to have a built-in adblocker in its browser as a clear sign of monopolizing the market. What Facebook and Google are doing is strengthening their already monopolistic positions. The Interactive Advertising Bureau (IAB) is monitoring the situation and legal questions have already been raised.

Apart from advertising, what else are you using Big Data analytics for?

About 30 percent of our business comes from data consulting. According to Gartner, by 2020 approximately 80 percent of business processes will be automated. We are integrating with companies’ CRM systems and help companies improve their cross-sell and up-sell, as well as lower the client exit rate.

What kind of companies use Big Data in their business?

For instance banks, which can use Big Data to segment their clients better. Banks know what we buy; however, usually too late to benefit from it in any way. If you’ve already purchased a car, it’s too late; the bank already knows that the client could afford it. It would have been better to know about his purchasing plans sooner. If they had known, they might have offered him leasing options etc. Banks don’t want to call people up and offer their products randomly. Similarly, you can help call centers get through to potential clients. If a call center calls you randomly, they will most likely irritate you. But we managed to use internet data to check when a client is viewing popular pastime websites. Because when they are browsing them, it means they have nothing better to do. That’s when you should call them.

Did that have measurable effects?

Absolutely. The percentage of people who took the call increased by 8 percentage points on average.

Still, many entrepreneurs and managers wonder whether the insights they get from big data are in any way better than intuition and experience. Why should they pay extra for it?

Intuition is good, but we use data mining to reach levels that cannot be accessed by intuition, to discover entirely new things. For instance, an airline’s media planner, in one early advertising campaign aided by Big Data, wanted to increase ticket sales to Italy and they decided to target the airline’s advertising campaign at businesspeople and people who are interested in travelling. That’s what his intuition told him was the right way. However, it turned out that the people who were the No. 1 client group at that specific time period were people interested in interior design. Perhaps it was because Tuscany was a source of inspiration for them, perhaps the destination correlated with their wealth. We don’t know why the campaign worked and we don’t need to know. The media planner was good but the machine was better. There’s another example: a popular men’s deodorant brand decided to target men interested in sports. Our analysis showed that it wasn’t men but women who most frequently bought that deodorant. The guy interested in sports has a beer in one hand and the remote in the other, while his wife – in an act of despair –would buy him the deodorant.

The Big Data market is growing at a much faster pace than the entire IT industry. Is this growth supported by actual increases in productivity of companies using the technology?

We are indeed growing much faster than the average for the market. That’s because the saturation of Big Data solutions in the market is still very low – it’s still a very fresh market. Even if we can improve efficiency by a few percentage points, it brings a lot of extra profit and savings. In internet advertising, using Big Data can increase a campaign’s effectiveness two or three-fold.

Are Polish companies open to Big Data?

It depends. Innovation is not always embraced easily and many firms still consider it risky. The Polish economy has many growth drivers. In more developed economies being more competitive is much more important. Our economy still has a way to go before it reaches that stage. It’s like with Maslow’s hierarchy of needs: Big Data is right there at the top, and we are still somewhere in the middle. Highly advanced technological solutions come later.

Is that why the majority of your revenue comes from abroad?

Yes, some 80 percent of our revenue comes from markets outside Poland. Our clients are media houses, advertising agencies and affiliation networks. In our data-consulting segment we work with large B2C businesses: telecoms, banks, insurers and multimedia companies.


Przechwytywanie Przechwytywanie 2

Likes, clicks, and ambassadors

Social media undoubtedly is a cornucopia of information for marketers. Take Facebook for instance, where every minute users post 510,000 comments, update their statuses 293,000 times, and upload 136,000 photos (source: The Social Skinny). It comes as no surprise that Facebook users alone generate some 100 TB of data each day. Twitter users send 6,000 tweets every minute (data for 2016), which is still less than a half of the traffic recorded in the service’s peak month of August 2014. According to different studies, 42-50 percent of marketers already considered social media critical or very important to their business. This means that social media is definitely crowded – you can bet that most of your competitors are already there. Marketers are therefore being taught not to overload their social media with content but to use them wisely. For instance, Facebook sees the most traffic mid-week between 1 and 3 pm, according to blog, but posts published later in the day, at around 7 pm, usually garner more clicks. It’s a trade-off between reaching a wider audience and getting through to fewer people but keeping them more engaged. This information alone can be useful for marketers but making your posts stand out from the crowd of other content is becoming a real challenge. And companies do want to get their money’s worth. Imagine you are about to publish a post about your latest winter collection. Based on your previous track record you know that posts published around 6 pm have been the most successful at getting clicks. You also know that the most successful posts were those that contained between four and six short sentences. There, you have a very general recipe. But if you want to tailor your social media marketing even more, you can go one step further and use predictive analytics to foresee how successful your social media communication will be before it is even posted. “You can predict which words make your posts more visible, for instance. We also check for sentence length and time when the post is published,” said Łukasz Trębicki, head of Social Media Analytics at Valkea, which recently launched its new online analytics tool called Viewell. Big data algorithms can also help you identify trendsetters among your target group: people who are followed the most by your target client base and whose likes generate the most traffic. “You could approach such individuals and make them ambassadors of your brand, for instance,” said Trębicki. Whether you sell shoes, cars or financial services, if you know your client’s daily routine, his or her favorite pastimes and favorite websites, you can definitely reach them more effectively and increase your visibility. The pressure to increase efficiency of your marketing efforts, be it online ads, sales calls or social media activity, will almost definitely continue to grow in the future.

shutterstock_big data for web

How much data is big data?

One of the most commonly used definitions of big data encompasses anything that is calculated in petabytes. According to McKinsey, big data refers to datasets that exceed the capacity of standard data storage, analysis and management. In addition to that, IBM also emphasizes that big data requires that the data come from diverse sources and be generated in real-time (or close to real-time).

Pin It