2. Structure
• Clearing the human road block: overcoming departmental silo mentalities
o Wim Casteur ,Business Intelligence Manager, Belgacom Group Strategy Customer & Market Intelligence
o Great case study on what it takes to consolidate the customer data silos in a telco to start to use big data effectively
• Big Data – Big opportunities – Big risks?
o Dr. Richard Benjamins, Director Business Intelligence, Telefonica Digital
o Good introduction to Big Data and the issues facing Telcos on Big Data – quoted throughout the rest of the conference
• Big Data: The Next New Big Revenue Opportunity for Carriers?
o Kevin SooHoo, Sprint
o Excellent review of the opportunities and challenges in selling customer insight
• Delta Engagement management, big data for big change
o Peter Crayfourd, Qifa Solutions
o Example of the use of Big Data to review the customers complete experience to make better decisions, and importantly
treat people as individuals not segments
• Big Data and Predictive Analytics what we can and cannot achieve with analytical BI tools
o Rokas Salasevicius, Civitta
o Refreshingly frank review of the many failures of BI in delivering business results, and links nicely to the points raised
by the previous speakers on customer all the data to build a better model enabling treatments to be experiemented with
and tracked
• Moving from traditional to predictive business intelligence: Creating a consistent consumer experience
o Dejan Radosavljevik Service Intelligence, T-Mobile Netherlands.
o Excellent case study in using customer insight to better manage the network, spending the network investment where it
impacts customer satisfaction.
3. This presentation highlights Belgacom’s solution to the common problem most
telcos face in that their data is trapped in silos, based on business units’ vested
interests so telcos simply do not use all the data they have available to them.
4. BI is critical to Telecoms, however, they are generally failing to take advantage of the data available to
generate meaningful insights. A theme of the conference is not so much copying the web guys’ technology,
like Hadoop and Hive, rather realizing Telcos have failed to use the data available to them effectively.
5. The 3 internal sources of data remain partially tapped. There is regulatory and internal
political concern about using external data; with customer opt-in, education and trust-
buillding all four data sources could be used in time.
6. Belgacom had the advantage that all data sources were stored centrally, the focus was on
enabling the silos to work together and reuse (a common theme in many SDP projects.)
7.
8. This is the key development a common data model with appropriate policy control to enable divisions to share,
where appropriate, the best available data – rather than error-filed, misinterpreted local copies of data.
9. The BI group was based in strategy, so had an independent role within the organization. So Belgacom is in a
unique position compared to many telcos from an organizational and infrastructure perspective. The key
now is translating this into a performance difference to more silo’ed operators.
14. This is a good summary slide of what is Big Data
15. Again a good summary slide, we’ll discuss later the opportunities, challenges and
risks with some of these models. The first has the biggest financial impact.
16. Again a good summary slide, especially the architecture which reflects what many
telcos are doing. We’ll discuss later the opportunities, challenges and risks with
some of these models
18. A person’s social network can be determined from their call record, its often a much
more representational map of who is important to them.
19. Traditionally we think of Big Data as addressing these aspect, but it applies across
all customer data. A key point is much of the data is quiet dirty.
20. Much can be inferred to build a reasonably accurate profile of customers simply
based on network data, no third party data required.
21. Say for a casino, where are customers coming from, providing important insights on
marketing effectiveness and also how to improve the return on future spend.
22. Used to aid in planning of the next location of a chain in a region. Its not just
anonymizing the data its de-identifying it – but limits the usefulness of the data.
23. BUT its small compared to telecoms. Perhaps telcos could achieve $2B, out of a $2T telecoms market. A
question often asked in the conference is should we be selling gold ore when we should first understand how
to make gold for ourselves. Use BI internally first, before focusing on such sensitive external uses.
24. The data is not clean – tens of thousands of phone numbers for one address (business).
However, there is a significant skills gap simply on working on data internally. Never mind
being able to sell the insights into verticals. Partners will protect their turf – challenge to build
a business by working with a future competitor. Its not an easy business to build.
25.
26. Overall Kevin asked some critical questions on whether telcos have the capability to
address this opportunity. People and processes are the limiting factor, not
technology! At present the customer communications is not being well-managed on
this topic and operators need to work together to educate the market and regulators.
27. Peter has worked on these systems for Orange and Hutchison 3G for over a decade and has put together a
good framework for using customer insight across the customers’ complete experience with the operator. I
found myself as a customer strongly supporting the weaknesses in the current systems. For example, I have
received hundreds of ‘hate SMS’ from my service provider every time I land in a country that roaming data
will cost $20 per MB – that’s like $400 to read my email! Each message reaffirmed the value in local WiFi,
and further degraded my opinion of the operator – Peter is showing how we need to use more data to better
understand each customer over their lifetime experience.
28. Averages were once good enough, but as customers expectations change on what is good service, and telcos
fight to retain customers, they need to look more closely. The snapshot is inadequate. When Peter asked the
audience to keep their hands up if they had not experienced service problems in the passed hour, day, week,
month. By a month virtually everyone’s hand was down.
29. This is a key point – we need to look over the customers lifecycle with the operator –
not just averaged snapshots. As a customer, taking such an approach would have
stopped hundreds of ‘hate SMS’ being sent to me over the years
30. Digging into this temporal view in more detail across the offer, services and use all feed into the customer’s
perception of the brand – we’re a product of our conscious and subconscious mind, how we feel about a
brand is influenced by previous experiences even though we do not specifically recollect all of them as a
specific interaction point.
31. Here is a good example of why big data is important in bringing together the
customers experiences over time to better determine how to react as in some cases
not reacting may be the more profitable option.
32. With a deeper insight better decisions can be made on how and when to react to specific
customers – Big Data enables a more human interaction in recognizing people as individuals.
33. Rokas gave a great presentation on the challenges in BI, and where most efforts fail – a critical
point in much is made of the tools, without a clear focus on treatments and business results.
34.
35. Unfortunately the target customers took the bundles and spent less money. Its important to
learn from our failures. The other factor not discussed is competitive environment, as
sometimes such offers are forced on an operator by competitors.
39. Treatment is critical and taking a customer lifecycle approach as discussed by Peter Crayfourd
is critical in understanding the customer, sometimes its simply too late.
43. That is they simply moved to another SIM with a better offer – key is finding and focusing on
the 15% - don’t waste time and effort on non-churners
44.
45. Building a better view of the customer and experimenting with the treatments not just
discovering segments is as if not more important. Put simply we still have a long way to go in
using the data we have available – better business intelligence, treatments and testing.
46. This is a good case study on using big data to run the network better
53. Overall they’ve been able to significantly improve where the network investment is spent to raise
satisfaction. Hopefully next year Dejan will be able to share some quantified data on the modeling
performance. But a few points increase in satisfaction can wipe out any revenue made through selling
customer insights to third parities! This should steer the prioritization of investment.