Tre Kronor har anlitat en Data Scientist inför World Cup!

Data Scientist ..? Vad är det och vad har analys av data att göra med hockey..?
Låt oss spola tillbaka bandet cirka 1,5 år. I april 2015 skrev jag det här blogginlägget om att nästa generations analys av hockeydata är här.

Den så kallade "avancerade statistiken" eller ”Hockeyanalysen 2.0” som jag benämnde den, var då bara i sin linda i Sverige. Det blev mycket surr runt om ämnet och såväl klubbar, media och förbund visade ett intresse för att börja använda statistik för att utveckla spelet och använda fakta istället för känslor för att fatta beslut om spelidéer.

Under förra Hockey-VM slängde jag, lite på chans, iväg en Tweet till Tommy Boustedt - ordförande för svenska ishockeyförbundet att de borde anlita min kollega Jon Blomqvist till nästa turnering.

Tommy Boustedt tweet

Tweeten ledde senare till ett möte med Tommy, och strax därefter tillträdde Rikard Grönborg som ny förbundskapten för Tre Kronor. Vi träffades och för Rikard var avancerad statistik inget nytt under solen efter sina år i USA och som tidigare scout. Dock hade Tre Kronor inte arbetat med visualisering av statistik och inte heller tidigare anlitat en Data Scientist för samla in, bearbeta och göra data pedagogiskt och begriplig.

Rikard och Jon fann snabbt varandra. Jon förstod snabbt att det krävdes mer än några tabeller från NHL’s hemsida och några excelark för att kunna göra en grundlig analys av alla spelare.

Rikard Grönborg_Jon Blomqvist

Så Jon samlade strukturerat in data från 1230 matcher vilket resulterade i 400.000 händelser (som t ex skottförsök, utvisningar etc ) och cirka 1.000.000 byten från hela NHL säsongen 15/16. Datat bearbetades och analyser togs fram till Rikard inför laguttagningen till World Cup Of Hockey, där Sverige spelar sin första match den 18:e september.

Är detta då en ”en svensk Moneyball”..? Nja… Liksom i filmen använde Tre Kronor visserligen statistik i sin laguttagning, med till skillnad från filmen, så fanns analysen med som ett komplement och inte ett substitut för Tre Kronor. Analysen tillför en ytterligare dimension och ger svart på vitt svar på hur spelare presterat och kompletterar vad Grönborg och hans stab ser och hur det känner spelarna.

I den här filmen berättar Rikard om hur han och Tre Kronor använder av avancerad statistik!

Här kan näringslivet lära sig mycket från sporten! Ta vara på all data som finns tillgänglig, bearbeta och analysera den, och sist men kanske inte minst - gör den begriplig för beslutsfattare genom visualisering! Läs gärna Christers Bodell’s krönika från CFO World som berättar mer om hur näringslivet kan lära sig av idrotten.

Lyck till Tre Kronor i World Cup! Det kommer att bli ännu mer data att analysera efter turneringen, då NHL meddelat att de ska börja med så kallad ”Player Tracking”

Snart dags för Hockeyanalys 3.0! Intresserad av en fortsättning? Följ med här>  @jcarlback

Read More »

Post a Comment

Analytics is a team sport at Team Sweden

  - its head coach leaves nothing to chance getting ready for World Cup of Ice Hockey

As the final push of the summer, warm weather rolled in over Göteborg, and the Swedish national hockey team "3 Kronor" stepped inside the city’s ice castle Scandinavium to practice and get ready. Saturday 17 September marks the beginning of World Cup of Hockey in Toronto. This version of 3 Kronor is not just any version. Just about everyone in the business will agree it’s the most competent group of hockey players ever to wear the Swedish jersey, extracted from the highest number of Swedish NHL players ever available. So it has not been without challenges to select the right set of players that collectively as a team has the highest chance of winning the tournament. Or, as I would put it, a team with the highest chance of delivering a performance that will yield the best possible result.

With players belonging to better or worse NHL teams, getting different amounts of game time on ice, being used differently, playing with stronger or weaker fellow players and against stronger or weaker opponent units it is difficult to compare their individual level of skills and performance. Selecting objectively *best* players is hard enough, to say the least. Selecting the *right* players for the team in this tournament is even harder.

Learning from 1,230 NHL games

To be sure to leave nothing to chance, the coaching staff headed by Rikard Grönborg brought in my SAS Institute colleague and data scientist Jon Blomqvist to support them in this process. As can be seen in this video Jon brought in all available data from 1,230 NHL games containing 400,000 events (shots, faceoffs, goals etc) along with about a million player shifts. This data has then been integrated, analyzed and visualized in order to as objectively as possible show each player’s contribution to its team. This information has then been used by Sweden’s coaching staff to evaluate each player in the selection process.

When meeting with Rikard Grönborg and some of his staff prior to engage on our joint “Analytics journey”, it was clear to me that he had an understanding, a mindset towards analytics that I would recommend any leader of an organization to have. He has also expressed this in interviews, as highlighted below.


Quantitative Analytics is a part of our process

Rikard and his staff know their business, they know their players. They have a grip on things that maybe can’t be fully measured yet, like what kind of atmosphere a player creates in a locker room and what that means to the team; the qualitative information. But he also knows that it’s crucial to base decisions on hard facts, that quantitative analysis needs to be applied, that they need to see things in “black and white”, without emotions, to get a full picture. He expresses it this way: “We need to combine “warm knowledge” (knowing the players) with “cold numbers” (analyzing the player’s numbers)”,"I firmly believe that the analytic department and the hockey department need to work together and create something really good"

It’s about raising the probability for a successful outcome

Ultimately, it is about being as prepared as possible, to know how to act in order to get the best possible odds in a certain situation. A single game is like a coin toss, the margin between a win and loss razor thin. This World Cup is a very short tournament, three group games to qualify for the second round, one semifinal and a final series best out of 3 means a maximum of 7 games against four or five different teams. The puck can bounce a lot of different ways with 8-12 players and four referees out there on the ice in full speed. Those familiar with stats will realize that it’s too small a sample size to have anything close to a “95% confidence interval”, but Rikard knows that his contribution is to do absolutely everything in his power to change the odds in his favor. All the way from tape, sticks, skates, via food, hotels, travel, practice schedules and coaching, and to utilize analytics to cut straight through all of that - to ultimately help improve in all those areas. Rikard knows you sometimes get the wrong outcome in a situation in spite of preparing correctly, but just like Brad Pitt playing GM Billy Beane in the movie “Moneyball”: “We’re card counters at the black jack table, and we’re changing the odds at the casino”. You don’t win every time, but you will be more successful over time by consequently acting based on the best possible odds.

Analytics is a process

Analytics is a process, a mindset, not a one-off. It’s about constantly measuring, evaluating, changing and then doing it all over again. When mankind eventually started to utilize that concept instead of just trying to figure everything out in the mind, the curve of innovation and progress took on a whole new direction. We’ve seen it in the fields of health care, medicine, transportation, energy, electronics, communication, manufacturing etc.

Ice hockey is no different than any of those fields as far as its possibility to capitalize on utilizing analytics. The difference is that analytics is still the most underrated and least developed area within the discipline of ice hockey. With Rikard Grönborg at the helm it has already taken a change in a new direction for Team Sweden. So let’s drop the puck, flip the coin and see if Team Sweden has managed to stack the odds enough to land with the bright golden side up!

If you want to read more on Why and How Sports should embrace Analytics, check out my previous blog posts and don’t hesitate to let me know what you think. It is feedback, evaluating and trying again that brings us forward so let’s move ahead together.



Post a Comment

Exploiting IoT series: Skills, roles and responsibilities

Los Angeles, California, USA early morning downtown cityscape.

Los Angeles, California, USA early morning downtown cityscape.

The scale and scope of IoT deployments in smart cities is a wonder to behold. Of course there is far more to smart cities than just IoT, from sustainable transport right through to connected citizens. But there are a number of cities that are leaping ahead in the way that they are using IoT technology to change the way that the city operates.

The city of Barcelona is using IoT technology extensively. From connected bus stops providing information about bus arrival times, to bin collections, all is grist to Barcelona’s mill. For example, a pilot has been set up to test whether using sensors on bins can improve refuse collection. The question is whether the routes of bin lorries can be optimised by sending them only to full bins. The city believes that this may save up to 10% of costs of waste disposal.

Two UK cities, Milton Keynes and Glasgow, are investing in smart lighting systems. In Milton Keynes, the city has installed a trial network of 400 LED lights linked to servers tracking light levels. This has helped to cut the city’s energy use by 40%, and the city plans to replace other, older lights with LEDs. Glasgow’s lights will turn on and off automatically when people pass, lighting streets only when someone walks down them. This is not just energy-efficient, but also good for safety: no more concerns about dark alleyways, and places where someone may be lurking, just well-lit, energy-efficient streets.

Santander is running one of the biggest sensor pilots in the world. It has over 12,000 sensors around the city, looking at a huge range of variables, including air quality and parking spaces. Although designed primarily as a research project to test the concept of a sensing network, the idea has been picked up with enthusiasm elsewhere and in the city itself. Projects based on the Santander model have been started in Guildford, Lübeck and Belgrade. In Santander, a project called Pulse of the City was launched in 2012, allowing users with GPS-enabled smartphones to check bus timings, and also report any problems such as potholes.

IoT sensors are also being used by cities to improve resilience in the event of natural disasters. Not usually associated with smart cities, this application could have very real implications for saving lives around the world. In Europe, sensors are being used in La Garrotxa, a volcanic area in the north-west of Spain. Information from sensors about river levels, CO2 levels and to prevent forest fires, are fed into a city-based dashboard, enabling immediate real-time assessment of the need for action.

Skills and roles, and application beyond smart cities

As we examine the lessons from early IoT deployments, my colleagues and I have looked at skills that teams will need to make the most of insights from these sensor-generated data. We have seen how some skills can be built, borrowed or packages bought which have pre-loaded configuration that reduces the need for specific skills.

In many cases, IoT fuels disruption. To seize the full range of opportunities, clear and visible leadership is required. Torri Martin, Director of Smart Atlanta makes illustrates the often common difference between job titles and actual roles. For example, IoT makes it necessary for partnerships to be established and managed, because useful data could be leveraged from 3rd party sensor networks. This requires both skills, and a clear role that partners can respect and consult. And most importantly, these issues are equally applicable to businesses beyond the smart city arena.

And old debate with a new twist

In some ways, this debate about roles is not new. Cos and their board-level peers have long argued the pros and cons of where enabling technologies should be controlled from - a central function or the business units. With IoT, we are seeing faster evolution of scope of responsibility. More managers now understand that we are in uncharted territories and learning quickly is the name of the game.

Which is why we are looking at IoT triggered roles, responsibilities and associated skills and experiences. Will you join us at the upcoming #saschat on Twitter on 16th September from 15hrs CEST? Our goal is to understand how emerging opportunities are being managed; specifically:

Q1: What are some good examples of external collaboration for IoT success?

Q2: How have data management and analytics architectures been managed for such multi-organisational collaboration to work?

Q3: What are the unique skills necessary to sustain and grow such relationships?

Q4: How have you seen traditional data management and analytics roles adapt to handle IoT streams?

Q5: What are some future roles, or even job titles, you expect to see in the next 12 months?

For inspiration and some new ideas, read our new survey Internet of Things: Visualise the Impact. The e-book gives you access to the lessons learnt by 75 executives from various industries all across Europe. They give unparalleled insights in how to integrate IoT in operations and share experiences of recent projects.

Post a Comment

Ahead of the IoT curve?

Opportunity is missed by most people, because it is dressed in overalls and looks like work. I believe this quote by Thomas A. Edison captures the last few months very well. Over the summer, the EMEA team has explored the maturity of IoT amongst major European companies. Here in this blog post, I will try to extract the essence of our findings and elaborate on why you should embrace the potential in IoT Analytics, too. Feedback is always highly appreciated.


Increasing efficiency

Many of the early use cases of IoT in hospitals relate to using resources efficiently, whether tracking particular pieces of equipment to improve utilisation, or informing family about a patient’s location. Even small improvements can greatly reduce costs, which are always under scrutiny.

Analytics will increase automation even where it has not traditionally been possible. An example of such an area is healthcare. Computers can analyse significantly more data than any healthcare professional, providing vital information and potentially making treatment more effective. Is it ethical to rely on computers? Or is it unethical not to?

IoT technology is also helping to improve demand planning and inventory management in retail. From Costa Coffee to Walmart, real-time data streaming helps identify stock shortages to reorder efficiently.

Not all IoT deployments are big, but even the most simple can make a huge difference. For example, sensor-driven replenishment of towels or emptying bins can make for a much cleaner, pleasant washroom.


Making quantum leaps: disruptive innovation

bridge-wires,low-resIoT technology is often used for incremental changes to efficiency and effectiveness. But perhaps the real value of the IoT may lie in its potential to turn the world upside-down and make disruptive innovation?

One area expecting quantum leaps is the automotive insurance ecosystem. Technology can provide advanced maintenance capabilities, service provision on the go, and education about crash risks.

The manufacturing industry has been quick to see the potential to provide a service and a solution, rather than a product (for example, selling holes, not drills). When utilising the industrial IoT, ‘think broad’ and use more data rather than less for analysis, is the recommendation.


Personalisation, self-service and experience

IoT, paired with real-time analytics, is a game-changer. By improving customer intelligence, businesses from shops to theme parks and telecoms can tailor unique customer experiences.

Predictive modelling will play a key part, with benefits ranging far beyond marketing. For example, to identify potential fraud and credit risk, employees more likely to leave, and even patients most likely to respond to treatment.

Early applications of health IoT largely focused on wearables for the ‘worried well’. But our study showed an encouraging trend: IoT technology to support self-care for the chronically ill, providing effective care management while improving the patient experience.

Interestingly, the importance of face-to-face meetings and experiences is increasing in the age of digitalisation. Events have become a key part of marketing in an IoT-linked world, and IoT helps us share them with others.


Becoming IoT-ready

The respondents in our IoT study clearly stated that it is a challenge to become IoT ready and develop the necessary skills, including data science skills. They build, borrow or buy skills, often in combination.

At the same time, we see the rise of citizen data scientists whose job role is not data science, but who analyse as part of their job. To be effective, they need good quality data. For some companies, investment are best made in data operations, to ensure ready-to-use, quality data.coloured-tents,low-res

IoT and analytics deployments also affect systems administrators, who are critical to their success and whose jobs are rendered harder by the sheer volume of data, as well as real-time streaming – and, who must support multiple functions, including the rise of citizen data scientists.

Organisations need to overcome three main challenges to become IoT-ready. They must manage data, ensure its quality, and develop use cases by supporting innovation and experimentation, while also encouraging improvement.

Data management and privacy. The IoT threatens to make previous data handling and privacy concerns look like a walk in the park. How to exploit the potential but remain on the right side of privacy concerns? The key is to put customers in control of their own data.


Learning from experience – your own and others

Deploying IoT technology necessitates learning from experience - fast: To be agile and adapt to circumstances.

While experimentation is essential, ‘learning quickly’ is better than ´failing fast´. The key is to learn from problems and overcome them. According to Gartner’s Emerging Technology Hype Cycle technology reaches a ‘peak of inflated expectations’, followed by a trough of disillusionment, before (hopefully) climbing gently up to a plateau of productivity. IoT is right at the top of the peak of inflated expectations. Does this mean that companies should stop investing? No. It merely means that they need to be prepared to be agile, and learn fast from deployments.

IoT will be useful. It is already useful, as our study has shown. It is a matter of identifying the right uses, through trying them out, and adapting them for success. IoT adoption is an ongoing process and there are already many IoT examples and areas to learn from.

For example, the development of radio-frequency identification (RFID) technology may hold lessons for IoT deployments, including the importance to let technology spread at its own pace as well as selling solutions to customers’ problems, rather than technology.

The world of sport, and particularly sports analytics, also holds plenty of lessons for IoT deployments. In sports, small changes may eventually add up to really big improvements. It is vital to focus on what really matters and not get distracted.


Design thinking and the widening skills gap

With so much data available, how do you know what to use to get the best insights? The answer may lie in design thinking, which places the user at the heart of the design process, supporting a focus on what really matters: the customer’s wants and needs.engine,lowres

Design thinking is particularly important in analytics prototyping and experimentation. While a tolerance of failure is essential for any experimentation, design thinking lets you ask more questions and perhaps avoid failure entirely.

I do not know everything about IoT, but this summer has shown me what an interesting topic it is. I encourage you to read the full study and please let me or any of the authors referred to in this blog post hear your thoughts.

Post a Comment

The value of connecting in the age of IoT and Telematics

As IoT and Telematics become the cornerstone for our connected world the value of being connected humanly also increases. With the rapid technology development comes a disruption in society where collaborating around opportunities and risks becomes absolutely crucial in order to stay relevant. At SAS we are part of several such networks and in the area of IoT and Telematics we are proud members of Telematics Valley. If you are in this industry you will want to read the interview with Johan Amoruso-Wennerby, chairman of Telematics Valley, below and also to join in the discussion via these two possibilities, both free of charge:

  • a half-day seminar in Stockholm Sept 28
  • Sept 9th, Telematics and SAS experts will be on Twitter to discuss IoT. Feel free to join at 15hrs CEST using the hashtag #saschat

    Let’s connect!

Telematics Valley and IoT

Telematics Valley is 15 years old this year. Established in Goteborg in 2001, it brought together car and truck manufacturers and suppliers of components and systems that supported ‘black boxes in cars’. Over the next years, it has moved on and is now a major force in the field of telematics. On 28th September an extended lunch seminar is hosted at SAS Institute in Stockholm. While preparing for this event, we had the opportunity to speak with Johan Amoruso-Wennerby, chairman of Telematics Valley, about Telematics Valley and expectations on 2017.

What is your focus?

We are a not-for-profit networking membership organisation, open to anyone working in the field of telematics. Over the years, our focus has shifted from the technology itself to its use, and more particularly, developing profitable services using telematics. While still retaining our original focus on the transport sector, our members are interested in the Internet of Things (IoT), and in Smart City technology, because telematics is a key enabler in those areas.

What benefits do you offer your members?

We exists to ‘support and strengthen’ both business and technology, and sees ourselves as a catalyst for business development in the field of telematics and related technology and services. We are a forum for our members to exchange both ideas and more practical help and support, including partnerships and collaboration. We also act as a representative for our members to advance their interests when necessary. Our reputation has steadily grown, as it has become recognised as one of the world’s leading membership organisations in the field.

How would you describe your members?

Our ambition is to bring together expertise from multiple sectors, so you can see among our members names from both public and private sector, as well as academia. Indeed, even within the commercial sector, our members come from a wide range of sectors. Car and truck manufacturers are represented, for example, by Scania and the Volvo Car Corporation. Electric and electronic systems manufacturers on the list include Mitsubishi Electric, and mobile data and IoT experts include Ericsson and Huawei. There are also a number of smaller organisations, including some fairly recent start-ups, who provide car- or transport-specific systems, including TOMTOM, perhaps best known for its satnav systems, and Trimble Transport and Logistics, manufacturers of the CarCube.

What are your plans for geographic expansion?

We started with a strong geographical focus on Western Sweden, because this area contains a major cluster of telematics and other technology companies. Goteborg has been a leading light in mobile data communications, and this has enabled a strong regional area of expertise to develop.

Since the early days, our membership has expanded across north-western Europe. Members in other European countries include Nokia Location and Commerce, headquartered in London, Grace, a company focusing on satellite positioning technology in Nottingham in the UK, Transics in Ypres in Belgium, providing support for fleet operators and Omnitracs in the Netherlands.

The expectations for 2017 is to leverage the increased diversity among members in an even better sharing and developing of ideas associated with important trends as Internet of Things (IoT) and Smart City technology. The upcoming event on Telematics & IoT is one such initiative to support this learning and sharing among members, and should bring new perspectives to the opportunities in the future, concludes Johan Amoruso-Wennerby, chairman of Telematics Valley.

Learn more:

To participate at the event, you need to register by email with name and company to: Telematics Valley / SAS Institute Registration

An online appetizer will be available on Sept 9th, as Telematics and SAS experts will be on Twitter to discuss IoT. Feel free to join at 15hrs CEST using the hashtag #saschat.

The event is a half-day seminar, which will be held in Stockholm, at the SAS Institute in Stora Frösunda, Solna. Its title is ‘Telematics and IoT – making it happen’, and the plan is to help shape the future of telematics and IoT. With the support of the SAS Institute, there will be an opportunity to discuss the role of analytics, and how it can be used to address some of the challenges. The meeting will also hear from industry speakers and thought leaders. Registration for the event will start at 9.30am, with presentations kicking off at 10am, and the session will end at 1pm with lunch.

In true Telematics Valley style, the focus of the event will be on networking and learning from each other, with a view to getting the most from meeting up. In the same spirit, you are very welcome to bring colleagues: just add their names to the registration form. The event is free of charge, and you can register now at Telematics Valley / SAS Institute Registration.

Let´s connect and start a discussion on the tweetchat or in person September 28!

Post a Comment

Don´t forget to change, it might be the last thing you did not do

We know that strategy is a long term thing that changes rarely. For example, a company who produces an item, let say a car, has decided early on to do just this. Their strategy might be to manufacture exclusive cars, just to keep it simple.

If we then look at the “how”, the tactics, we can for example see different ways to distribute, manufacture, and advertise. This is where organisations need agility and understanding. You might ask yourself why. Because change is inevitable. Scary? Yes! However, the results of not changing is even scarier.

However, if the strategy is well defined and understood by everyone, it gets easier to adjust and develop your tactics. Simply because everyone knows where they are going, even though the way there alters. The competitive landscape is ever-evolving. Continuously adapting to others slow organisations down. To me it sounds better to drive the change, instead of waiting for others to take the lead.

If we go back to the car industry, Tesla is revolutionizing it. Electric cars are taking over. Soon they will also be self-driving. It’s a fact. How come there is only a couple of other brands who build competing electric and/or self-driving cars at the moment? Tesla can provide a complete eco-system including free charging stations to their electric vehicles who are able to cover much longer distance then the rest.

Sorry, this is not a Tesla-blog providing you with the perks of being a Tesla-owner. However, it’s an understandable example most people recognise. In this case, competition need to adapt or perhaps come up with another offering that is even better than what Tesla is offering. The slow ones will take a huge risk and might not be able to close the gap.

This is where agility and understanding comes into the picture. Tactics. What do we need to do now? How do we build a better offering? How do we beat the competition?

How do you decide tactic and how do you change tactics?

Most of the information is sitting there in front of you. Data. Through your data you can understand, decide and create. The list goes on. Your data gives you unlimited possibilities.

There is a term called Data Innovation (Whitepaper: From Data to Action). Take your data – play with it – find out what you can do. Change. Innovate. Be better.

Some of my colleagues have been working with the Swedish Ice-Hockey Team in their preparations towards the upcoming Canada Cup (reflective blog post will come soon). Exploiting the underlying data and deciding on tactics according to the values they found will improve the chances of a medal. Impressive? Absolutely!

Until next time, I recommend you to read this article (in Swedish) by my colleague Christer Bodell and get some insight into what is happening in Sports Analytics. He is discussing how we need to allow data to make decisions, and leave our feelings to the side. Staying rational during decision making while being angry, happy, sad etc. is very difficult. Perhaps we should let the emotions flow when we have won, not before.

I believe that sports/athletes can benefit hugely from Analytics, and it is already proven in many ways. Due to the unmatched dedication to winning, it is only a matter of time until analytics has become a cornerstone of sports.

Dare to change. It is difficult, nonetheless beautiful.

Read More »

Post a Comment

Why benchmarking is necessary for marketers on the data-driven journey

Artificial intelligence has been around for quite a while—research on it started in the 1950s, in fact. But it is only now that it is really moving beyond the pages of science fiction, and into the realms of not just reality, but usefulness. In part, this is because of recent improvements in analytics, computing capacity, and algorithm development. These together make machine learning and cognitive computing, the two main components of artificial intelligence, much more accessible and useful.

There has been a lot of excitement in particular around how artificial intelligence is likely to improve customer experience. Cognitive computing, for example, uses analytics to answer questions. It is already being used to manage fast responses to customer queries to service centres. Computers create an initial response and direct the query appropriately. It’s fast and effective. Machine learning is enabling improved planning of marketing campaigns, based on customer segmentation and preferences.


Perhaps most excitingly, the emergence of deep learning, the third component of artificial intelligence, is starting to benefit marketing, and in particular, more complex tasks. It is early days yet, but if a computer can be ‘taught’ to drive a car, there is little reason why it can’t learn to carry out marketing tasks like mapping customer journeys.



Data-driven marketing?

The idea of artificial intelligence being used in marketing, and especially to improve customer experience, is very exciting. But in reality, artificial intelligence is totally dependent on data. If we’re honest, we know that for most marketing teams, including here in the Nordics, the journey towards data-driven marketing is only just beginning.

Marketing today is both easier and more difficult than it has ever been before. We have access to huge amounts of data, but how should it be used most effectively? Marketing is not, perhaps, seen as a natural partner for mathematical modelling and analytics. But to succeed in a world that is becoming increasingly data-driven whether we like it or not, marketers need to understand and use data to answer questions about their customers’ preferences and problems. Fortunately, analytics tools are becoming increasingly user-friendly. What is required now is more a willingness to try things out, to see what value can be created, rather than a degree in mathematics or computing.

More complexity, increased reward

The Internet of Things (IoT) is one of the key drivers of the huge increase in data volume in recent years. With estimates suggesting that the number of connected items could increase exponentially over the next few years, the data volume is likely to get ever-larger, and much of it requiring real-time analysis.

But if the IoT will add complexity and data volume, it also has potential to increase the rewards. It is already, for example, being used to improve customer retail experience, through the use of improved personalisation and tailored interactions and offers. The question, as we have noted before, is not so much can value be generated, but how many of those who could benefit are even aware of the potential.

Benchmarking to assess progress

With so many aspects of marketing being influenced by digital customer experiences, how does a marketing team assess progress and seek guidance on aligning priorities? SAS believes benchmarking can help. If you're ready to take stock of your digital marketing approach, we are ready to help you develop a game plan to strengthen your marketing confidence. Take the assessment and get your score.

Post a Comment

Fraud and Fado: mournful tunes but no sentiment of resignation

In May this year a great number of telecom operators gathered in Lisbon, the city of Fado, to discuss telecom fraud. Organized by the Communications Fraud Control Association (CFCA) and the Forum for International Irregular Network Access (FIINA), operators ranging from AT&T, Vodafone, Korea Telecom to Orange and Deutsche Telekom shed light on how old and newer forms of fraud are detected and combatted.  As the Fado lyrics often deal with the life of the heart-broken and days past, telecom fraud nowadays deals with the whereabouts of criminals, increasingly organized crime. And as telecom fraud becomes more organized and sophisticated, so need the tools and methods for detecting and combatting current and future practices.

Too often, telecom organizations are working in a reactive approach where responsible fraud managers act as a plumber: they look out for anomalies in traffic, literally the ‘odd ones out’ - traffic to national and international suspect destinations and number ranges and traffic coming from new, potentially ‘risky’ applicants – before they try to repair the leaks. This is typically done by patching the firewall, maintaining static “black lists” and looking for illegal simboxes.

But looking at the individual cases, do they see the bigger picture? Aren’t there too many ‘false positives’ clogging up too many scarce resources. Are relations between fraud cases and fraudsters visible?  And what about the fraud schemes that are not solely or not at all call related? The possible connections between external fraudsters and internal employees? Supplier fraud? Subscription fraud? And the hidden cases? The false negatives, the cases that stay below the radar?

Fraud is big business in telecoms

The communications industry ranks in the top 5 industries most threatened by economic crime according to a recent PWC survey (Global Economic Crime Survey 2016).  It’s ahead of sectors such as insurance, manufacturing, energies, utilities and construction, and ranks straight after banking, public, media and transport/logistics.  Interestingly a great number of operators are eyeing up banking and media/entertainment companies and broaden their activities, thereby increasing their exposure to economic crime, fraud, money laundering etcetera substantially. Currently operators lose a whopping 38 billion US dollar to fraud, according to last year’s CFCA survey. Top methods are typically (IP)PBX hacking, subscription application fraud, dealer fraud and subscription identity fraud. Not only traditional telecom is affected: also cable/satellite/IP TV providers are targeted by fraudsters with unlawful card sharing and illegal streaming. Even in a small country as the Netherlands with 7.4 million households roughly EUR 14 million is lost annually due to illegal TV viewing, according to the Dutch cable association.

And as cybercrime has climbed up the ladder of fraud methods according to the PWC research, we can expect this to become a new category soon in telecoms as well. An example in case is a hack in the dealer system of a European carrier where prepaid cards were switched to postpaid, thereby enabling the fraudsters to run up high bills without paying for them. Calls were made to so-called International Revenue Share Fraud numbers, nowadays openly available and traded on the internet. Kick-back and revenue share models enable all criminals involved, home and abroad, to take their unfair share from these scams, leaving customers and often operators to foot the bill. Sometimes internal resources can be implicated: employees committing unauthorized conversions from prepaid to postpaid in the HLR or suppressing billing Call Detail Records or even call agents implicated in unduly complaints. A fresh KPMG research on fraudster profiles found that groups of fraudsters very often comprise people both inside and outside the company. Sixty-one percent of colluders are either not employees of the company, or are employees who work with people who aren’t. Some of them are former employees. This highlights the need for better third-party due diligence of such persons as vendors and customers. Another fraud form, related to vendors (and customers) are VAT fraud, or so-called Missing Trader Intra-community (MTIC) fraud which is also prevalent in telecoms, especially regarding handsets and CPE. Also malware, enabled via unsolicited messages and installed on a mobile phone, can do the trick. With Voice over LTE (VoLTE) mobiles are even more exposed to fraud because signaling is implemented in the mobile OS instead of mobile-based broadband, as for 2G/3G telephony. Many of these vulnerabilities can then actually be exploited remotely through mobile malware.

Moreover, as the number of new, connected networks grow and the number of devices explode with the advent of the ‘Internet of Things’, new forms of economic crime will emerge. We already know of hacking connected ‘things’, whereby running up phone bills is only one part of the damage done. What about criminals hacking into your smart home, as your customer forgot to change the default password settings? Reputational and brand damage can be often much bigger. Maybe operators, at least some, want to exude ‘security’ and ‘trust’ as part of their brand image and core values – just like banks do. And increasingly critical private and business customers expect their telecom provider to protect them from harm done.

Analytics coming to the rescue

Traditional, often rule-based investigations can only detect so-much of fraud. As KPMG concludes in its fraudster profiles report: “The key anti-fraud technology is data analytics, a tool that can sift through millions of transactions, looking for suspicious items. But only 3 percent used pro-active anti-fraud data analytics in detection of the fraudsters surveyed.”

Data analytics can go much deeper, look beyond the obvious. Analytics is an encompassing and multidimensional field that uses a combination of mathematics, statistics, predictive modeling and machine-learning techniques to find meaningful patterns and knowledge in recorded data. This can be internal data, but also external data sources, such as Chamber of Commerce information, social media, or (IRFS) websites. Adding powerful and cheap processing and storage methods, such as Hadoop, for analyzing increasing amounts of data and running sophisticated software algorithms – producing the fast insights needed to make fact-based decisions. By putting the science of numbers, data and analytical discovery to detect and combat fraud, we can find out if what we often assume or believe is really true. See hidden patterns. More specifically analytics can help with:

  • Increase efficiency by reducing wasted time spend on investigating false-positive alerts. A sophisticated fraud scoring engine applies risk- and value-based scoring models to prioritize events before they go to investigators. With the time saved, investigators can work many times the number of cases and focus on higher-value networks. Moreover, the outcome of the investigation is sent back to the system in order to continuously increase the detection accuracy.
  • Spot more suspicious activity by processing all data (not just a sample) through analytical models in batch or if needed in near-real time as data streams in. Also bringing in new, external data sources with structured and unstructured data (even text-analytics).
  • Detect also first and third party fraud networks and crime rings that would otherwise be missed. Also automatically identify suspicious networked behavior in the data. Identify early risky applicants and transactions.
  • Increase the collaboration between the fraud & marketing and finance/credit risk departments thanks to efficient customer screening and scoring. Without interfering with the customer journey, the analytical models can prevent losing money and high-end devices to fraudsters who should not become customers. Or use different offerings to different customers depending on their risk profile.

An encompassing fraud framework

One of the challenges for combatting telecom fraud is dealing with its many forms - TM Forum has made a classification which, ideally, is tackled from one framework, instead of using different tools for different forms.

Picture 1

Most telecom operators have a classical Fraud Management System that mainly focusses on call related frauds such as IRSF, PBX and other frauds.  When new fraud modus operandi become more important, the fraud investigators are often not able to detect these in time.

Therefore, it is important to have a Fraud Framework that enables you to cover all types of fraud in a flexible way and works as an Enterprise Platform.  The usage of analytics is certainly a key factor in order to fire the alerts but there are also other layers within the framework that are equally important:

Picture 2

  1. Integration, quality and management of data: The telecom operators are facing a high volume of data internally (CDR, customer data, etc.) but should also leverage external data, structured (lists, demography statistics, etc.) and unstructured (social media information, web crawls, etc.). All this data should be brought together which will require a system that can access all these sources transparently and which can also handle the data cleansing in order to make sure that the data quality is optimal.
  2. Detection: Using a hybrid analytical approach (combination of business rules, anomaly detection, predictive modeling and social network analysis) in a white-boxed approach will enable not only to detect all types of fraud but will also highly decrease the false positives whilst increasing the hit rate.
  3. Alerts qualification and investigation: Alerts should be shown in an understandable manner to the fraud investigators, giving them the opportunity to visualize and prioritize the alerts following their internal processes and needs. For each of the alerts, the investigator needs to clearly understand the reasons of the alerts in order to ease and shorten the investigation time and process. After investigation, the outcome will be sent back to the system in order to increase the detection logic (feedback loop).
  4. Monitoring and ad hoc analysis: the follow-up and pilotage of the fraud information can be used by the general management to determine the future fraud prevention approach but also to fine-tune long term strategies in different domains (fraud, marketing, risk, etc.).
  5. High performance and real-time technology: In order to avoid that fraud losses explode quickly, the fraud framework has to work (near) real time. This will also prevent that the fraud detection interferes in the customer journey of new and existing customers.

As a conclusion, it seems obvious that the telecom business is still very lucrative for fraudsters. Fraud losses are still not under control and the fraudsters continuously find new ways to increase their profit.

Nevertheless, the new technologies and the usage of advanced analytics can help the operators to stop this phenomenon.  Other sectors such as the banking sector have already embraced these methodologies with success and although not obvious, there are a lot of best practices that could be copied and applied similarly to the telecom sector. But that is another story…


Matthieu Joosten is Telecom Industry Lead at SAS South West Europe

Frédéric Hennequin is Senior Solution Specialist Fraud at SAS Belux

Post a Comment

Telecom Fraud: What Operators can learn from Banks

By Matthieu Joosten & Frédéric Hennequin

Telecom operators loose annually up to a staggering $40 billion on fraud. And with the advent of the Internet of (Insecure) Things (IoT), extending connectivity far beyond smartphones and tablets, our homes, cars, clothes and lightbulbs can become little Trojan horses, vulnerable to security and fraud exploits. So how should the telecom industry tackle current and future scams? One learning is broadening the scope of fraud management, going beyond traditional blacklisting and reactive traffic anomalies. Using an encompassing, enterprise-wide Fraud Framework based on advanced and predictive analytics, detecting risk and fraud before it happens, as we argued in our previous fraud blog.

Now we concentrate on what telecoms can learn from other sectors which are hit equally hard, or even harder, by fraud and have progressed in detecting and combatting it. Banking is one such industry.

Financial industry hardest hit

Following PwC’s 2016 Global Economic Crime Survey, the Financial Services industry is most at risk regarding fraud, slightly more than Government and Retail but way ahead of Communications & Insurance. The Financial Services industry is not only at risk, it’s also been heavily affected and has faced material losses, certainly since the era of online banking (remarkably, a number of telecom operators are moving themselves into the banking industry).
Together with the facilitation of payments and increased use of credit cards, fraudsters have found new and easy ways to steal money and many banks have already suffered vast losses within the last 10 years. When you look at which banks have been targeted, the fraudsters, increasingly organized criminals, started with the big banks in big countries. After that they targeted big banks in smaller countries before targeting smaller banks everywhere. Why?  Because fraudsters are smart people (often with university degrees) and they adapt their strategy to the market.  When one builds a defense wall, they find another victim without defense instead of losing time finding a way to bypass the enhanced security.
What are the main fraud modus operandi the financial sector has been suffering the last years?

  • eFraud Transaction Fraud via mule accounts using phishing/vishing/malware/etc. techniques: online bank accounts from customers are hacked and money is transferred to mule accounts – similar to what happens in telecoms with ID fraud, malware etc.
  • Application Fraud (again, similar to telecoms): customers asking for a credit/loan (phones, set-top boxes) and disappear with the money without reimbursing the bank (or telecom provider).

How to combat banking fraud in the digital age?

Initially, banks mainly worked with simple rules where thresholds were defined above which an alert was triggered – similar to what telecom operators are doing nowadays. As fraudsters became more sophisticated, it never took long to know how to avoid these thresholds and stay under the radar. In addition, fraudsters often know that bank fraud systems rarely monitored customer behavior across multiple accounts, channels and systems. That vulnerability paved the way for cross-channel fraud,  which enables criminals to gain access to customer information in one channel, then use it to commit fraud in another channel.
Putting harm to injury, banks also had to move into the digital age. Digitally demanding, fast-paced consumers expect to be able to make easy and fast payments in real time, at any time and from any place and device. Digital channels, however, are innately more vulnerable to fraud, and, while the speed and openness of the approach made banking faster and more convenient for customers, it also made it easier for fraudsters to access money and transfer it quickly without being detected until after the crime.

Analytics coming to the rescue

Clearly banks had to take a more sophisticated approach to fraud detection and many banks moved to a state of constant readiness. Careful data monitoring and management is critical from the outset, and banks are now often, where necessary, enhancing their data quality and collating and linking a wide range of different data types ingested into an organization, including finan­cial and non-financial transactions, customer information, bank account details, computer IP addresses, and information about devices and their usage patterns.
Using these different data sources and types, many banks have started applying more advanced analytics and machine learning. And in order to be really efficient and effective, they combine multiple techniques consisting of anomaly detection, peer group analysis, text mining, (social) network analysis & predictive modeling to get answers to questions like:

  • Is the beneficiary living in a country at risk?
  • Is this beneficiary living in a country to which the customer already sent money?
  • What is the usual timing at which the customer does his transactions?
  • What is the usual device that’s used for those transactions?
  • Have other customers with similar characteristics paid off their loans in time?
  • Does this customer who is applying for a loan have contacts in his network who are already known for application fraud?

By combining all these analytical approaches, banks are increasingly capable to:

  • Avoid future fraud losses for existing and new modus operandi
  • Maintain the manual workload of the fraud investigators at a manageable level
  • Increase customer satisfaction and avoid customer annoyance during the customer journey

Moreover, banks, like telecom operators, often took (and some still take) a siloed approach to addressing fraud – check fraud is handled by one group, credit card fraud by another. So over the past years many banks have taken a single, holistic platform approach independent of product, channel or geography to address all financial crime, including money laundering. Cost savings and efficiency improvements are also drivers: reducing data storage costs, enabling reuse across departments and increased flexibility to add new products, services and channels to the enterprise platform at a far lower incremental cost than installing another customized fraud detection system.

But what about telecom operators?

All telecom operators are suffering fraud losses in various modes (IRSF, application fraud for high-end devices, etc.). Most have static and reactive detection systems in place where leaks are plumbed as they are detected (see our previous blog post on telecom fraud).

So it’s time for telecoms to “get inspired” by the advanced fraud detection models that exist within the banking industry (certainly as some of them are becoming banks!) and start getting answers to questions like:

  • Is the destination number located in a country at risk?
  • Is the destination number in a country to which the customer has already called in the past?
  • What’s the usual device the customer uses and what’s the usual time at which the customer performs his calls/messages/data usage?
  • Have other customers with similar profiles asked for a similar high-end device?
  • Is a customer, applicant, dealer, distributor, call-agent or even supplier connected to fraud, and if so, how and does that constitute a risk?
  • How can I decrease the number of ‘false positives’ and increase efficiency?

If telecom operators invest time and effort in more advanced fraud detection techniques they will avoid losing money to customers and companies they shouldn’t have accepted in the first place and will be able to reward customers who should get their full attention.
Want to read more? Read about the SAS Fraud Framework or download our whitepapers: ‘Using Analytics to Proactively Detect Insider Threats’ or ‘Addressing Fraudulent Payment Activity
Another good read is: ‘Protecting the Enterprise’
For any questions, don’t hesitate to reach out to us.

Matthieu Joosten is Telecom Industry Lead at SAS South West Europe & Frédéric Hennequin is Senior Solution Specialist Fraud at SAS Belux

Post a Comment

How a simple blog performance dashboard motivates authors

As SAS, we are fortunate to work with some of the brightest analytics consultants in the world. Most of them are happy to share their ideas with the extended community, to inspire and encourage development of new use cases and improve analytics performance. Presentations at SAS conferences, ideas in SAS insights and blog articles are just some examples of how these ideas are being shared.

As digital marketers, my colleagues and I are able to leverage these ideas in our campaigns. Like so many companies these days, we want to be a ‘digital organization’, and also be data-driven. To become a fully-fledged data-driven digital marketing organization, we need to be active online and the entire organization needs to be involved. To be truly data-driven, we also need to be able to make sense of our activity and its outcomes.

One way to do this is via a dashboard, which is why we are in the process of developing one now.

 Blogs dashboard article header image

Impact and motivation

One particular section of our new dashboard will show the impact of our bloggers. Bloggers are spread through the organization: across countries, geos, and departments. Everyone here is skilled and knowledgeable, so anyone can blog, and share their knowledge about their area of expertise. We are always on the lookout for more company bloggers.

We run programs to encourage staff to become more social and digital, and develop into online thought leaders, sharing their ideas and vision. But only by tracking and measuring, and being data-driven, can we assess the impact of this activity. It is vital that we know what works so we can improve our outreach, build on good practice, and get closer to becoming a truly digital marketing organization.

A dashboard will also encourage competition among bloggers, and motivate them to improve. The dashboard will show bloggers their own performance compared to others, with a list of the top 10 posts and authors. We hope that the challenge of working out how to feature in the top 10 will encourage bloggers to ask digital/social leads in the company what they can do to perform better and to write better – in other words, to improve their impact.

Once we add the number of clicks, engagement levels, and leads generated from each blog post, the competition will really heat up. Being able to show the business impact of our digital activities makes their importance very obvious, and encourages bloggers to make even more effort to improve.

Top 10 Authors



With SAS being the acknowledged leader in analytics, we have used SAS Visual Analytics for the dashboard. The great thing is that there is a lot of blog data available and we do not have to wait for IT to prepare data sets. Instead, we can create our own reports, so that business intelligence is fully integrated into the business.

What the digital marketers and bloggers see is an interactive dashboard that is easy to use. They can set filters to see subsets of the results and enable them to drill down further. In the future, we hope to enable the spearheads to get involved in data exploration, so that they can analyze their own activity and increase their impact. They may even want to explore data from other bloggers and do some benchmarking to improve their own performance.

Views per month


Connect with customers and prospects

Our blogs help us to connect with customers and prospects, and anyone else interested in analytics in the broadest sense of the word. Their responses to our blogs provide even more, and smarter, data. By adding this to the dashboard, digital marketers can analyze which topics are resonating in the market, which blog posts get the most engagement, and much more. This in turn helps us to design more relevant campaigns, and generate more engagement, and closer relationships with customers. This, in turn, of course, results in still more data.

Truly digital and data-driven

The new dashboard will enable us to make better decisions, faster, and become more relevant for our customers and prospects. Using data to improve our (online) marketing campaigns moves us closer to becoming not just a digital, but a truly data-driven organization.

Do you have a similar dashboard in place or are you setting one up? I would love to hear from you – just send me a tweet or direct message!

Post a Comment