The Bonn-based company, which offers mail delivery, logistics and banking services, was partially privatized through an initial public offering in November. Then the company changed its name and decided to split itself in two in March.
With more than US$30 billion of sales in 2000, the company is growing fast, buying companies in the United States, Europe and Asia in an aggressive push to build a world-class delivery company.
But before it could deliver on its profit and sales goals, Deutsche Post knew it must acquire an important strategic asset valued by any enterprise: deep understanding of its customers. In addition, with 450 delivery bases, 2 million parcels a day and 80,000 business customers in Germany alone, Deutsche Post also realized it would have a major-league logistics problem of its own if it didn't find a way to quickly and efficiently track the flood of data flowing through its vast operations.
So it called on Hari S. Chakrovertty, a mathematician with three decades' experience in computing, to create a data warehouse, and Chakrovertty called on Teradata®.
Since implementing Teradata's solution, Deutsche Post has moved beyond the limitations of disparate databases into a true data warehouse environment. With Teradata's data warehouse in place, Deutsche Post is improving its tracking of parcels, sharpening its delivery times, increasing customer retention and beefing up products and services to meet customers' needs.
Recently, Chakrovertty, Deutsche Post's project manager for data warehousing, took some time away from the implementation to talk about the pivotal role that improved customer insights play in Deutsche Post's promising future.
Q: Could you provide a quick overview of Deutsche Post's logistics and delivery business? It sounds challenging.
A: Deutsche Post has 80,000 business customers (in Germany). You have to understand the business of your own business customers: what they are doing, what they actually want. Do they need, for example, 24-hour delivery, or on-demand delivery or things like that? The other challenging part is the volume. You are talking about 649 million packages a year alone in Germany.
Q: And you somehow have to keep track of all that?
A: That's right. So the major challenge is to understand the customer and provide total supply chain management to these customers.
Q: Can I get a quick picture of the typical Deutsche Post customer?
A: The typical customer is anybody who's sending parcels. But the major business focus of Deutsche Post is mail-order companies, who make up over 60% of our volume.
Q: What were some of the specific business questions you wanted the Teradata system to answer?
A: First, to measure some key performance indicators: the delivery time, losses and
damages, things like that; then to do customer segmentation. We have the problem of customer retention. What are the indicators to see which customers are going to leave and when? What is the profitability of individual customers? What can be done to improve the profitability of an individual customer or group of customers, and (what is) the cross-selling potential?
Q: What have you learned so far about the customer?
A: We know now the affinity of the customer; not everyone is alike. Not every bookseller (customer) is the same, for example. Some booksellers sell more CDs than books. This bookseller is on the Internet, while this one doesn't do Internet business. There are lots of differences among these customers, and we can fit our services to them.
Q: What new services or products have you introduced recently based on what you're learned?
A: Deutsche Post is constantly improving service by offering new products and service levels. Now small-business customers have an easier way of sending a parcel. Instead of having to find a box and stamps, then weighing the parcel, they can simply buy a box from us, label it and send it. We also are picking up more than once a day for some customers.
Q: How does the Teradata system play into the overall business strategy of Deutsche Post, especially its expansion into different countries?
A: If all the companies or subsidiaries use (different systems, it's) going to cost you a fortune. It's much easier to shuffle (all) data into a central database. It saves a lot of money. And you have one single version of the truth, and everyone can see the customer in the same way.
Q: Can you quantify the bottom-line impact of the Teradata solution?
A: (Laughing) I can, but I'm sorry. …
Q: Would I be presumptuous if I say it's helping the bottom line?
A: It does help, and it's substantial.
Q: What key performance indicators did the company target, and how were they ultimately affected?
A: I can give statements by the company CEO. By 2005, Deutsche Post wants to be
No. 1. They want to increase their market share in certain areas, actually to double market share by 2005 in certain market segments.
Q: How did Teradata become a part of Deutsche Post's efforts?
A: In fact, the whole project (began) as a solution for quality management. Deutsche Post started with this data warehouse project three years ago. At that time, the freight division wanted to have a quality system to produce key performance indicators. My suggestion was instead of implementing a quality-measurement system it would be better to go with data warehousing because it will give them quality management plus additional capabilities.
During the evaluation process, we selected Teradata as a database because it is the widely accepted and implemented database in the decision-support environment. The single database strategy of Deutsche Post is Oracle, with all legacy systems using Oracle. Together with the Deutsche Post IT department, we started a proof-of-concept and carried out a benchmark with original data at the Teradata benchmark center in San Diego. This convinced the executives.
Q: What did the proof-of-concept phase determine?
A: We evaluated the basic feasibility of the system. For example, is it possible to load and transform event data during the night along with other data from billing systems?
Deutsche Post handles around 2 million parcels a day. We had to prove (that tracking data about these parcels) could be loaded in real time; we were looking at two or three hours. If it takes 10 hours, that's not good. But we loaded three months' worth of data into the system within two to three hours. Tracking data used to take about 20 hours to load on an Oracle database.
The next question was, how would the system respond to complex queries with large amounts of data? We currently have over a billion entries in a single table. Also, how often would it require tuning?
Q: So the Teradata system satisfied whatever challenge you put to it?
A: Yes. That's right.
Q: How many other systems were you considering?
A: We looked at Oracle, Sybase and Informix.
Q: What gave Teradata the edge?
A: Scalability (for one). We started with a very small database. The system proved it could grow very fast and change on the fly. We started with about six or seven users; we now have more than 200 using this data warehouse. We changed from a two-node machine, to four-node, to an eight-node machine last year. In terms of data volume, we started with 400 gigabytes. Today we have 1.2 terabytes of data. That kind of scalability.
The second thing: We were looking for a system to process a three-year amount of data, or about 3 terabytes each year. We were looking for (a system to allow us) to grow infinitely.
Also, the operating cost must be very low. All other databases, including Oracle, required administrators as the database grows and reorganizes. In Teradata, you don't need any (additional) administrators at all, because you don't have to reorganize, defragment, etc. With Teradata, it's very easy to handle a large amount of data with very few people.
Q: Can you quickly go through each phase of the implementation process and explain the goal of each?
A: First there was Phase 0, where we did proof-of-concept and evaluated all the database, OLAP, and back-end and extraction tools. That was from August until December of 1998.
The first phase, which started in January 1999 and ended in July 1999, was to develop 40 reports and key performance indicators for the quality department. After that, we did cleansing. After collecting data from many systems like SAP for billing information and loading it in the data warehouse every night, we found that the legacy system was not always producing the correct data. There were a lot of problems also in identifying a package because we didn't have a unique identifier for a parcel.
Q: Was that a serious hurdle?
A: No, it was just trial and error. It took some time, up to the end of 1999, for us to finish cleansing the data and have valid data loaded in the system every day, including our billing and loss-and-damage information.
We started intermediate Phase 2 in January 2000, and it ended October 2000. We produced a lot of analytical matrices for quality and for sales.
The current phase we're in is called database marketing. We are trying to understand the customer better, and produce products and services according to the demands of the customer.
Q: What have you learned through implementing the Teradata data warehouse, and how did the implementation process go?
A: We were very fast in implementing the data warehouse and very fast in generating the first 40 reports. We did it in seven months, and it was OK.
We were very satisfied and very excited from the beginning that if we put the data together and taught people how to use certain OLAP (online analytical processing) and reporting tools, they would be able to do the analysis and we wouldn't need IT anymore. This is only partially true. We really didn't succeed in making everyone understand data and how to analyze it. We have very powerful tools for OLAP and reporting. It took a long time for us to understand why people couldn't use these tools and to teach them. It took longer than we believed it would.
Q: How long did the training take and what was your expectation?
A: It took more than a year. I thought it would take a month or so. We started training people in July 1999 until middle of 2000. In this time frame, we created a number of so-called power users in almost every department who are the focal points and who can generate reports.
Q: How well are employees accepting the system?
A: Acceptance of the system is pretty high because it's simple to use.
Q: So is the implementation finished now?
A: Yes. We have 200 users daily, and we are producing roughly 1,200 reports every day, just from the quality department.
Q: How many reports did the division produce before Teradata?
A: I do not know exact figures (for the entire company), but the quality department produced one report monthly, consisting of about 60 pages.
Q: Do you have plans to expand the amount of reports?
A: Yes, we are going to integrate the salespeople, about 500 users, by Q4 of this year.
Q: And the salespeople have specific questions and needs?
A: They want to see a total picture of the customer, the profitability of each customer, what service level is given to this customer.
Q: How many other departments does customer information affect, and what are their specific needs?
A: Marketing has just started. They are going to start the first direct-mailing campaign in Q4 based on the customer segmentation created by the data warehouse.
Q: And operations?
A: Operations has been using the system for two years now because they were the first to want analysis of delays and how to avoid them.
Q: How many people are on the development team?
A: We've got about 15 from Teradata on the data warehouse project and about 10 internal people working on implementation. There are about five (people) generating reports externally—mostly Computer Associates.
Q: What are some of the functions you're planning for the data warehouse project? I know the company is planning to integrate foreign subsidiaries under the data warehouse umbrella.
A: That's the next thing we'll do ...
Q: What else?
A: We do not have predictive models to help us predict what customers might need or if (they) may want to leave. We are using data mining to create such predictive models. We still do not know what customers really want. We do not have data from all touchpoints like call center, kiosk and Web. So we have to get closer to the customer.
Q: You've been in this business a while. What have you learned from prior jobs that's helping you here?
A: I worked for IBM for more than 17 years and have been in the business 30 years. At IBM, I was head of a department, then a lab manager at San Jose and Tokyo. I left in 1981, as lab manager in Germany, with 1,200 reporting to me. My background is very broad-based. I have been in hardware development and software development. I've been a freelancer, in the consultancy business, for 15 years now. During this consultancy, I have worked for Unilever (and) Daimler-Benz. This is the first time I have taken on a project with a government entity. That's why I took this job, because it was interesting to see how to move this dinosaur.
Q: You have described Deutsche Post as an innovative company and yourself as a motivator. Can you please explain?
A: Deutsche Post was a government entity and is a 500-year-old company. So you can understand what it means going from being a government entity and not facing any competition in some businesses to all of a sudden trying to get a foot into the free-market area. They have totally changed their view of operations and the marketplace. They renamed themselves Deutsche Post World Net to signal that they are a global player. They have bought a lot of companies like DHL International to establish their own network around the world. A lot of changes have been made, and the company is growing very, very fast as far as revenue is concerned. It's the largest postal service in Europe, with its own network in nearly every European country.
We are working very hard to get the company to become customer-centric. Moving from a product-centric company is also a big challenge. It's not as easy as it sounds, but we are making excellent progress. T
C.C. Williams is a New York-based writer who covers business, finance and technology issues.
Photo of Hari Chakrovertty by Wolfgang Steche
Teradata Magazine - Q3 2001
Quick Links
