Joe DeCosmo, head of analytical processing to the global lending agency Enova International and keynote speaker, said her team regularly reviews the course of customers on its website and in its call center to identify opportunities for improvement. The solution does not necessarily pass through the analytical treatment, but in one case, the Enova analysts found that a 50% reduction in page load time has led to an increase of 2.25% of rate conversion and an increase of $ 1.5 million in incremental revenue just for one product. The lesson to be drawn is that good models can not overcome bad processes.
2. Develop a culture of learning and testing
Amazon, CapitolOne and other leading companies have proved the value of the tests. Andy Pulkstenis, director of analytical processing to State Farm Insurance, said the willingness to test a direct correlation with innovation. He urged officials to consider the analytical processing failed tests as steps to success. He cited the example of the popular WD-40 product, which was invented only after the failed tests to the formulas 1 to 39 of water repellency. Most companies never go beyond the A / B testing or a factor both said Andy Pulkstenis. These techniques are powerful in themselves, but State Farm also uses more sophisticated techniques, including multivariate analysis with and without covariates. In an A / B test simple two direct-mail advertising touting a life insurance policy for young parents, State Farm found a letter in plain text generated a higher return of 45% to the same message surrounded by pictures related to babies.
3. Abstrayez analytical processing applications
Pressed to enter the market, Enova has integrated analytical process models directly into their transactional applications, but it’s not ideal, admits Joe DeCosmo. The withdrawal of the analytical processing of applications has simplified the continuous curve fitting and swapping models without disrupting business applications. Moreover, Enova has withdrawn analytical processing services in real time, available in the cloud (for scoring, approval of loans, etc.), the company can now offer non-competing lenders .
4. Document your work
How many teams in charge of analytical treatment they have named templates based analyst who left the company years ago? How often you do you realize you have repeated the work already done in a previous project in which you were not aware? Many knowing smiles were drawn from the audience when Joe DeCosmo asked these questions. To avoid repetition of scenarios effort, he advises teams to document their work in detail, whether analytical or other treatment. This is a step that requires time and effort, but that largely pays off, because you can reuse IP, refine previous work and avoid repeating the same mistakes.
5. Focus on analytical processing “at the time of decision”
The analytical processing continuously and in real time much interest, but keep in mind the actual time of the decision, advises Bill Franks, head of analytical processing to Teradata. For example, the US tax administration has no use of real time, as it has for weeks to detect fraud before canceling refund checks. Conversely, a bank discovered that she had to revise its batch updates made during the night when an analysis showed that three successive contacts a client about the same subject, such as bank charges indicate a high probability of client alienation. So if a customer asks a question on the web, then makes a phone call and, in the same day, went to an agency, the agency manager can not be warned that this account is at risk if the service folders customers are not updated during the day.
6. Spend an artisanal process automated
The science of analytical processing is now entering its industrial era, says Bill Franks, Teradata, with automated options for the choice of algorithms. Automated approaches based on software may not be producing a perfect model, but they are probably adequate. “The important thing is to maximize the overall impact, not optimize any decision,” said Bill Franks.
7. This is not a man against the machine, but adding them
Be open to new tools and technologies, but keep in mind that the machines need to be recalled the reality of men, warns Gina Papush, director of data and analytical processing for insurer QBE North America. “Sometimes you need to override the model based on your own knowledge of the company or of a return to corporate information,” she says. Other speakers acquiesced citing many examples of machines that produced bad results that the business analysts have identified broken immediately. It certainly lacked the contributions of the necessary data and business rules, but the machines do not know the difference.
8. Use more data to induce a deeper relationship with customers
It is always better to have more data and more data sources. Joe DeCosmo says his company has added the authorized use of customer bank records to refine the offerings and offer better terms to customers. Samih Fadli, director of intelligence for the Razorfish digital agency, urged the audience to enrich internal data to establish unique identifiers, using third-party data and tracking cookies and web devices identifiers.
9. Be close to the director of strategy
Companies are starting to use analytical processing for tactical purposes and many companies are maturing and use it throughout their organization. To go further, Joe DeCosmo encourages responsible colleagues analytic treatment “to put the analytical treatment in the heart of everything you do.” The ambitious goal of the team in charge of the analytical treatment QBE recounts Gina Papush, is not only to be part of the practice of identifying strategic opportunities, but to be a central component.
10. Start with the five key imperatives
To summarize this advice, I would say that the imperatives for leaders of analytic treatment is to know the business, understand the customer experience, using more data and multiply the tests to achieve superior results. Not least, the real challenge for managers is to promote analytical processing power data-driven analyzes within the company. “You can use the data, but you can not hide behind them,” observes Anthony Canitano, general manager in charge of advanced analytical treatment to Delta Airlines. “You have to understand the context in which the data can be applied in every area of the business.”