David Raab Article Archive

Alterian Marketing Suite (Alterian, Inc., 312-704-1700, www.alterian.com). Here’s a blast from the past: a campaign management system built on a proprietary database. Ten years ago, such systems were by far the most common marketing database software. It seemed a new one appeared every month. But the balance quickly shifted toward systems built on standard relational databases like Oracle, DB2 and SQL Server. Today, new installations and product enhancements of proprietary marketing systems have virtually ceased, although a surprising number still chug away in the back rooms of marketing departments.

Those systems remain because proprietary databases have a fundamental advantage: they can be tuned for specific activities. Most proprietary marketing systems use an inverted data structure, meaning they group data by columns rather than rows (or, to use an older terminology, by fields rather than records). This is well suited for marketing queries, which typically look at a few items (columns) for all customers. Relational databases must load the entire row for each customer, which is much less efficient. Inverted structures also more easily support common marketing queries that require multiple passes through a relational database. Often an inverted system can return in minutes queries that would take hours on a relational system using much more powerful hardware.

But query speed isn’t everything. Probably the gravest technical issue with inverted systems is update time: adding or changing even a few records sometimes requires rebuilding the entire database, which can take hours. Although some vendors have avoided this problem, even the fastest inverted systems do not allow the real time updates that come naturally to a relational engine. More fundamentally, inverted engines are isolated because they do not natively use the SQL language shared by relational databases. When SQL interfaces are available, they are typically so slow that users lose the benefit of having an inverted database in the first place. As marketing systems become ever more tightly integrated with other corporate systems, the isolation of an inverted engine is increasingly unacceptable.

So what’s with Alterian? Is it the product of some crazy Rip Van Winklish time warp, appearing after its time has gone? Are its managers die-hard gearheads who spent so much time tinkering in the garage that they lost touch with the world outside?

Actually, no. Alterian’s technology did indeed start in the mid-1990s, when its founders developed the inverted engine for Brann Viper. But the company has remained fully engaged with industry trends since its launch in 1997. In fact, it carefully avoided positioning Alterian as a general-purpose marketing system. Instead, Alterian focused on applying its core technology where it has particular advantages.

One set of users has been marketing service providers, such as list compilers, who benefit from the low cost and high speed, refresh their data periodically rather than continuously, and need not integrate with external systems. Over half Alterian’s revenue currently comes from such firms. Another approach has been to offer the system for quick, ad hoc data analysis, taking advantage of the query flexibility, easy database design, and facility with large data sets.

The Marketing Suite represents an extension of this strategy. It provides expanded tools for data analysis, statistical modeling, Web access, and multidimensional reporting. Even the new campaign manager, scheduled for release by January, offers a circumscribed set of selection functions rather than the range of real time interactions, marketing administration, content management and other capabilities in today’s enterprise marketing systems.

The foundation is still the Alterian Engine itself. Unlike earlier inverted engines, which ran around 5 million records per minute on moderately complex queries, Alterian runs about 5 million rows per second. Simple queries could be 20 times faster. The difference is the earlier engines mostly stored raw data, while Alterian creates and stores indexes. Alterian has about 30 index types, which it applies as appropriate to different types of data. It may create multiple indexes on the same data to accommodate different query types.

This indexing has a cost: data loads at 10 to 15 gigabytes per hour and the Alterian database occupies 1.5 to 2 times the volume of input. In an era when even modest marketing databases can exceed 100 gigabytes, the load time places some constraints on Alterian applications. The system does let users add calculated values without a full rebuild, however, so analyses using such calculations can be done freely.

Alterian’s primary user interface, in a module called Data Discovery and Visualization (DDV), uses a combination of Venn diagrams (intersecting circles that specify relations among input groups) and selection rules, which can include random, Nth or ranked samples. These are combined on a graphical flow chart to produce complex queries. Cross tab and profile reports analyze results in graphic and tabular formats. DDV also provides tools to load the database and create calculated values. A Web FastCounts module provides Web access to similar reports without the database and calculation tools. The Analytic Reporting module, also due by January, lets casual users create multidimensional drill-down reports similar to Excel pivot tables.

The Campaign Manager builds its own flow chart of input segments and output cells. Segments may be taken from DDV or built within the Campaign Manager. Cells can be then split and nested indefinitely. Users can attach a schedule to a campaign–but not to individual cells–and define cell selection levels (e.g. household vs. individual), quantity limits, sampling procedures, test/control splits, and output formats. Separate flow charts define inferred responses. The system does not explicitly support multi-step or event-driven campaigns and lacks project management, financial planning and other marketing administration functions. But it is more than adequate for basic campaign selections. A separate Modeling module, using SPSS technology, provides several statistical modeling methods.

Alterian runs on Windows servers and workstations. It does provide ODBC drivers to let SQL tools read its data, plus connectors for SAS and Siebel. Pricing is based on database size, number of users and modules purchased. A 1 million customer system including the Campaign Manager would cost about $200,000 for a perpetual license. Alterian Engine has over 100 installations.

* * *

David M. Raab is a Principal at Raab Associates Inc., a consultancy specializing in marketing technology and analytics. He can be reached at draab@raabassociates.com.