Teradata: Your Next Best Action with Your Customers

by Christine Newman

The week after the terrorist attacks on September 11th, Wal-Mart knew exactly what customers bought in each of its 2,600 stores nationwide. With a latency of a mere six hours from transaction to analysis, Wal-Mart’s RetailLink‘, powered by a Teradata Warehouse, provided insight on customer buying patterns, enabling Wal-Mart to take actions to provide customers with what they needed when they needed it. An article in the Wall Street Journal on September 18, 2001 described how Wal-Mart used their Teradata Warehouse to understand their customers’ changing buying patterns the week after the attacks. On Tuesday morning of the attacks, stores were empty “ghost towns”. By that evening, customers “pantry loaded” on guns, ammunition, bottled water, gasoline containers, and TV sets. On Wednesday, sales of flags shot up dramatically, causing a shortage and triggering orders to restock the flag inventory for the next sales day. By Friday the number of customers in the stores returned to normal, but the average purchase declined. Based on analysis from Wal-Mart’s Teradata Warehouse, some stores were able to predict a sharp increase in sales by the 5th day. As a result, they increased staffing by 20% for the weekend and stocked up on items that customers had put off buying. This paid off as sales increased by 25% that weekend compared with the same weekend a year earlier.

Imagine your business being intelligent enough to recognize and predict customer buying patterns, anticipate the customer’s needs, and deliver just-in-time marketing promotions because you understand your complete relationship with the customer at the point of sale. This is no longer the stuff of a brave new world, but a reality achieved by active data warehousing with Teradata“. The business world has evolved into a fast-paced, 24-hour environment requiring tactical and operational decision-making at the point-of-touch with the customer. Thriving companies not only minimize operational costs, they also actively pursue growth in revenue and business value. It is now imperative that data warehouses do more things faster and smarter than ever before. Traditional data warehouses provide strategic decision support by using historical data to report what happened, and analyze why it happened. The traditional data warehouse primarily involves batch updates and a limited number of concurrent users who use data to make strategic decisions. These systems do not meet the needs of new business requirements for faster, individualized response and relationship management with customers. Enter active data warehousing with Teradata.

Active data warehousing is an extension of the operational, business critical component of the enterprise system, providing a single source of analytics and information and analysis across all customer touch points. This multi-channel integration is essential to enable customer relationship management (CRM). Companies need a single view of the truth across all functions throughout the enterprise, in order to take the next best action with each customer. Not only does the active data warehouse fulfill the strategic needs of the traditional data warehouse, it also allows for tactical and event-driven business decision-making in the field.

The needs of the current business environment are driving the evolution towards active data warehousing. Technically, the active data warehouse must deliver performance, scalability, availability, and data freshness. Teradata is uniquely positioned to deliver these requirements.

Performance
Today’s competitive business requires the data warehouse to perform both complex strategic decision support and tactical decision support. Tactical decisions are the operational mainstay of day-to-day business management. Pitting diverse types of work against the same data often generates conflict within the traditional data warehouse. Strategic complex queries, such as “What is the total number of red shirts sold, in the northwest region, over the last quarter as compared to sales of the same item last year in the same timeframe?” usually require more time and system resources to execute. While short tactical queries, such as “What is the total number of red shirts in stock?” must run faster (in seconds) in order to provide on-the-spot decision-making to drive the next best action with each customer.

How does Teradata handle the coexistence of both types of workload? Teradata’s Priority Scheduler ensures that the mixed workload associated with an active data warehouse is dynamically prioritized so that queries run with appropriate performance. For example, a short tactical query to determine the current credit history of a customer before an offer is made can maintain a quick turnaround when run simultaneously with a complex inventory adjustment query, just as if it was running standalone.

Teradata’s Dynamic Query Manager (DQM) is a successful, proven tool in managing complex and analytical queries before they reach Teradata. Teradata DQM supports proactive workload management by allowing users to schedule requests to run at particular times, and it supports reactive workload management by delaying requests until a heavy workload inside Teradata has subsided.

Teradata’s advanced indexing capabilities also add to the performance needed to support an active data warehouse. Join indexes (built-in query accelerators) enable a faster query response time by determining which data columns and joins are needed by the queries. Teradata’s summary tables via aggregate join index structures enable a sub-second response time when a guest’s current dollar value is assessed at check-in, or if a store manager needs to know how fast on-sale items have moved in the last hour.

Teradata’s triggers apply intelligence against the data to determine what the next step needs to be, and then executes it. These triggers can automatically generate a purchase order when inventory is below set levels, or determine whether to offer a cross-promotional item to the customer based on recent customer purchases in conjunction with current activity.

Scalability
Teradata’s scalability supports mixed workloads, hundreds to thousands of concurrent users, and ever expanding data volumes. Having linear scalability means that Teradata delivers predictable performance when adding more data, more users, and more subjects to your data warehouse. The shared nothing architecture of the database software and the MPP hardware means that if you double the number of nodes, you double the amount of work that can be done in the same amount of time. Compare the Teradata database’s MPP architecture with the limit of SMP-only databases. With SMP-only databases, you experience diminishing incremental scalability with each additional CPU because the overhead of shared resources becomes a bottleneck for the single node. This is the most important difference between shared and shared-nothing environments.

Availability
Availability and reliability are of paramount importance in the active data warehouse. Today’s global commerce runs 24 hours a day, 52 weeks a year. Companies cannot afford to miss out on opportunities because of system downtime. Teradata is built upon a highly fault-tolerant hardware platform, designed to withstand component failures with no service interruptions. A Teradata MPP system groups nodes into communities called cliques. The nodes within a clique work as a team to support each other to complete the workload. The Teradata BYNET interconnect coordinates work amongst the nodes, delivering full redundancy in case of failure. Beyond hardware, Teradata supports “fallback”, an intelligent software mirroring option that replicates a copy of the data on physically isolated disks. This capability enables data to be fully available during most software and hardware outages, and also allows for continuous backup and instantaneous restore. Teradata not only protects your business from unexpected downtime, but its architecture also minimizes planned downtime.

Data Freshness
Today’s fast-paced business environment means you need to make decisions based on recent, fresh data. For Wal-Mart, the weekend after September 11th had very different sales patterns than the weekend before. If their data warehouse was not updated continuously, Wal-Mart managers may not have made the correct decisions regarding inventory and staffing for the weekend right after the attacks. Teradata’s TPump utility provides a continuous feed of new or changed data into the data warehouse, without placing full-table locks on the tables. Data fed by TPump may come from multiple sources, updating the same table, while queries are simultaneously accessing the table, due to it’s row level locking interface. Within seconds of an account closure, a major purchase, or a default on a loan, that information can be stored in the database and available for strategic and tactical decision-making.

Active data warehousing with Teradata enables companies to take the next best action with each customer. Wal-Mart took actions based on current data and analysis obtained from their Teradata Warehouse. What will your next best action be?

PDF file of this story

Christine Newman is a freelance writer based in San Diego, California.