IBM's Decision Edge Integrates Channel Allocation
But multi-step programs, which involve a sequence of messages, responses and new messages, do add one important new element: limits in channel capacity. A typical multi-step program will involve responses by call center or field staff to service requests or sales leads. Field and call center resources are usually limited. There are only so many qualified agents available, and they can make only so many calls per day. By contrast, outbound mail quantities face no real constraints other than budget, and even outbound telemarketing is usually outsourced in advance to agencies that can handle whatever volume is required.
Dealing with channel capacity involves two issues. The first is making plans that assign channel resources to the most productive use. This is the province of optimization vendors like MarketSwitch and Trajecta, which use projected responses and values to find the most profitable mix of promotions within a set of capacity and budget constraints. The second issue involves operational decisions: What do you do when actual demand exceeds available resources? Traditionally, these decisions have been made by the channels themselves, though processes like telephone call routing and sales force lead distribution. But channel managers and their systems often lack the complete data and strategic overview available to the marketers who designed the campaigns initially. So, to ensure that after-the-fact allocations support the original business strategies, these decisions are best made within the campaign manager itself. This requires a set of capabilities not needed for traditional outbound campaigns.
Decision Edge for Campaign Management (IBM, 800/772-2227, www.ibm.com) provides a rare integration of channel allocation with conventional campaign management. The system works with an existing marketing database, using an internal data dictionary to translate field names and relationships into user-friendly terms displayed on-screen. The dictionary also lets the system administrator create derived fields, such as calculated values or segment codes, that look like part of the physical database.
Although customer data can be stored in any external structure, the system uses its own layouts for internal data, including campaigns, products, messages and channels. This approach is used by most modern campaign management products. It allows the system to include capabilities that rely on specific information being available about those entities.
The most complicated data are associated with campaigns. These are assigned to groups, which are used to limit customer contacts across multiple campaigns. Users can specify how many campaigns within a group that a customer will receive per year, how many campaigns the customer can belong to simultaneously and how long to wait between campaigns. By setting these limits at the group level, the system makes it easier to enforce such policies than if users had to specify them for each individual campaign. Users do have the ability to check additional filters - such as do-not-promote flags or credit screens - at the campaign level. They can also define who is eligible to receive the campaign, using a point-and-shoot query builder.
Each campaign itself can be split into multiple cells, each having additional selection criteria. If the same record qualifies for multiple cells, an allocation mechanism will distribute the records so each cell does not exceed a specified percentage of the entire promotion. This is somewhat awkward compared to the usual approach of explicitly defining cell priorities, random samples and maximum quantities. But it does allow users to achieve most of the same things.
Each cell is assigned a branching sequence of steps, defined on a graphical flow chart. Each step is assigned a channel, message and a set of possible responses. The messages, stored in a common database, can include text and personalization variables drawn from the customer database. The responses are defined as codes that will be read from records imported into the system. This assumes the responses are actually captured in an external system, such as a call center, and then fed back into the main marketing database. This is a reasonable and typical approach, although it does preclude the system from managing interactions that require an immediate response.
Different responses, including a non-response after a specified period, direct records to different steps in the flow. A separate process looks for records that have made a purchase, as defined at the campaign level, and excludes them from all later steps in the campaign.
Capacity is managed at both the campaign and channel level. For each channel, users specify the number of records allowed per day. Then, for each campaign, they assign a priority number and specify a maximum number of records that can be selected in total and per day. The priority ensures that records from the most important campaign are selected first, while the total and daily campaign maximums ensure that no single campaign absorbs all the available capacity. When records within a campaign must be chosen for processing, they can be selected by rank, at random or on a key value. The system lets users define complex point-based scoring algorithms to help with ranking and selection.
Decision Edge for Campaign Management also includes standard promotion analysis reports and ad hoc report writers, including a cross-tab and distribution report that can compare different time periods. Users can export results as flat files or into database tables and can schedule campaigns for repeated execution.
The system runs on Microsoft Windows workstations, Windows NT or IBM AIX servers. The main database can be DB2 on System 390 and AIX, or Oracle on AIX. Pricing is set at $75,000 plus $7,000 per 100,000 customers, which should be cheaper than most competing campaign management software for small and mid-size firms. The product was originally deployed in Japan and was released for worldwide distribution this month.