FEATURE SINGLE CUSTOMER VIEW
ááá “It’s usually about call centres or websites
where clients want to switch channels. Those opting for full real-time updating often have a strong brand and need excellent service,” says David Barker, Head of CDI at Acxiom Europe, noting that one prospective client is planning to implement real-time data cleansing and updating across 13 countries.
“It tends to be an inbound issue to support dialogue management,” agrees Ben Salmon, marketing solutions strategist at Rapp. “It originally manifested itself in call centres. It’s about understanding from the SCV what the customer has done on the web.” With all the extraction, transformation, cleansing and loading involved, updating large and complex customer databases is extremely processor-intensive, and tends to be a batch mode activity. Moving this process into real time causes quite a few problems. If batch updating takes the target system offline then that’s not usually an issue with an overnight load. With real-time, the target system will be live some, if not all of the time. With heavy loads combined with lots of immediate record updates, as you might find in a call centre on a weekday lunchtime, this means that updating can grind to a halt as the system prioritises operational users over updates. If the source data goes stale, it will start to go out of sync with the data derived from it, such as aggregated fields. And keeping all the data consistent is just one of the barriers to updating quickly.
“What if a system fails?” says Barker, who lists obstacles such as hard-to-integrate legacy systems that can’t work easily with modern service-oriented architectures, resolving update conflicts quickly enough and dealing with matching records in multiple languages. “If it’s mission-critical, real time can be very expensive because you need redundant systems. It’s not really that hard from an engineering point of view, it’s getting value that’s difficult. There are workable solutions and models but there can be a significant investment in hardware.” So a full real-time SCV is possible, but it’s not going to be easy to build and it’s certainly not going to be cheap. Unsurprisingly, there are alternatives to going the whole hog. This might mean increasing the load frequency to twice daily
32 April 2010
instead of overnight, or only updating the absolutely essential data items in real time. What this means is that most data is updated via a normal weekly or overnight cycle while the small proportion needed to support an application is continuously refreshed.
“It is absolutely possible to update the customer data that is needed quickly, while leaving less urgent data until later,” says Gareth Mitchell-Jones, Director of Analytics at Experian Integrated Marketing. “This is what would make a real-time customer view solution work brilliantly, once what is real, what is near, what is batch and what is offline has been assessed. Building these preferences into your strategy and long-term development will create the real-time customer view your business requires, without over-burdening your system with unnecessary activity.”
This is the approach that Rapp has taken in an ongoing project for a large gaming client. The operational delivery of offers is online-only but data from all channels has to come together to help build the offers in the first place. With all their systems online, this is no great problem for pureplay operators like Amazon but most established retailers have work harder.
“It is absolutely possible to update the customer data that is needed quickly, while leaving less urgent data until later.”
Gareth Mitchell-Jones, Director of Analytics, Experian Integrated Marketing
“The web uses two kinds of data,” says
Salmon, using work done for this large gaming client as an example. “One is static and doesn’t change very much day to day, like name, address, interests and demographics. The other is click data and that is changing every few seconds. What we do for our client is cache the static data and load the click data on top in real time.”
What this means is Rapp rebuilds the customer view completely every night, refreshing model scores, segmentation codes and the like, and then generates a list of possible offers based on this mix of transactional and demographic information. The top 10 most relevant offers are then picked for each individual and loaded into the call centre inbound marketing system where click data is used to pick which one to finally serve to the customer. “Static data is where the heavy aggregation occurs,” says Salmon. “It takes less than five
“The real challenge is maintaining the quality of data, while getting close to the expectations
customers have for interactions with real people.”
Nigel Magson, MD, Tangible Data
seconds from a click by a customer to that data being available for analysis, and it’s actually a fairly simple decision to make. What product or service are they looking at and what’s the most relevant offer for them on the list?” This move towards partially or fully-automated dialogues with the customer is one trend driving the need for faster updating. Nigel Magson, Managing Director of Tangible Data, thinks marketing systems should need to pass some sort of Turing Test so that “customers could believe that they were interacting with a real salesperson or service rep”. Famous mathematician, Enigma decoder and Bletchley Park alumnus Alan Turing created this test, whereby a computer would be deemed sentient if it responded to questions in a way that would be indistinguishable from true human response. “The real challenge is maintaining the quality and integrity of customer data, while meeting the increasing demands from the market for interaction with systems that at least gets close to the expectations customers have for interactions with real people,” says Magson. “In this scenario, an ‘I’ll get back to you in 15 minutes’ response may well be acceptable. This time will decrease as technology around data exchange improves.” Portrait’s work at The Nationwide is probably the best known example of
real-time marketing in action, and partial real- time updating of its SCV supports this. Customer
prompts based on customer contact history, attributes and propensity scores were originally introduced in its call centre to assist agents with customer dialogues, and have since been expanded for use across the branch network and Internet bank. Prompts can be channel- and role- specific, and updating is fast enough to respond to information received within the current interaction.
www.dmarket.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40