The use of analytic tools to generate real business advantages requires considerable knowledge:
- Advanced training in statistics or econometrics
- High level of comfort with stateof- the-art data processing technologies
- Familiarity with the typically “messy” data sets involved, and experience converting them into usable form. (Data gleaned from Internet browsing does not arrive at the desktop organized in a way that is conducive to a balanced, scientific analysis.)
- Understanding of the underlying business, including approach to marketing, current product capabilities, and industry trends
5. Make sure you have a firm grasp of fiscal and personnel requirements. There is no one answer to the question of cost. At the low end, the implementation of a packaged targeting model might cost between $100,000 and $200,000. The system might be adequate to identify prospects for a direct marketing campaign, but is unlikely to take into consideration the specifics of the business. At the high end (in the millions of dollars), a retained firm may get involved in the reengineering of a complex business unit for about a year. Typical factors that drive the complexity of the situation is the type (demographic, credit, purchase, etc.) and amount (transaction level as in credit card, retail, or telecom) of data, as well as whether the data is stored in multiple databases.
Whatever your investment, getting the best return on project dollars depends on careful planning, the team’s expertise, and management’s willingness to adopt an unbiased perspective. You must be clear about how you define success and keep an open mind about possible solutions. Implement a well-designed pilot to anticipate costs, identify potential stumbling blocks, and project results that will yield the desired bottom-line impact.
Advances in analytics are breathing new life into companies’ efforts to create new businesses and line extensions, improve pricing, and cut costs. Successful enterprises are harnessing the power of data for better strategic decision making. The rewards are rapid implementation of new ideas, products, and services, which result in greater profits and shareholder value. This is why Harrah’s is betting on analytics to sustain its winning streak.
Sandeep Tyagi is managing principal and founder of Inductis, a firm providing management consulting, analytics, and outsourcing services.
| |
DESIGNING AN
EFFECTIVE PILOT
In an attempt to understand why certain products weren't selling, a catalog
retailer essentially rounded up the usual suspects: It lowered prices, offered
free shipping and other incentives, and took a tougher stance on costs with
suppliers. Despite these efforts, however, revenues did not recover.
Intent on finding the answer, the company then designed an analytics
pilot that looked at customer behavior over three years, sifting through tens
of gigabytes of data. It decided to test five levels of prices and several product
features, including shipping and handling charges, as separate segments.
The company also decided to vary the way products were presented
in the catalogs (for example, size and placement of pictures).
In all, the company tested six catalogs containing more than two dozen
distinct product offerings, using an analytics technique called fractional
factorial design. The analytics demonstrated, surprisingly, that increasing
prices could be the solution to the revenue problem. Through the testing,
the team was able to ascertain which combination of factors-such as shipping
charges, size and type of pictures, and payment terms-would support
higher pricing. In the end, the company increased prices by 15%, and
enjoyed an equivalent increase in revenue.
The pilot analysis also helped the company create a more efficient and
profitable approach to customer targeting. The end result of most pilots is
to select the offer that got the best response as the "winner." However, this
does not address two important dimensions: (1) All responses are not of
equal value; (2) all customer groups do not share the same offer preference.
By matching offer characteristics to specific customer profiles-for example,
size of credit line by likely need for credit-the result is usually a suite
of alternate offers, each of which can then be rolled out to the optimal
group of customers. Even before pilot offers go out, it is important to simulate
the likely dollar contribution of particular offer/customer combinations
along multiple dimensions such as demographics, credit behavior, prior
buying history, and the history of offers made to them.
In this case, outreach to a refined customer population segmented on
the basis of total profitability was instrumental in reducing costs by about
22%, while increasing revenue by 5%. |
|
|
|
|
