These one-off requests were challenging enough, but McCain also had to contend with continuous requests. The issue here was that often the initial request would be for a limited set of product attribute data, but over time the company would be asked to provide additional pieces of information. "You don't know the full extent when you start," says Luby, "until all of a sudden you figure out that you're spending a lot of time and effort on this and it's becoming mission-critical."
Finally, the frequency of requests for product data has been increasing. Customers and other entities are starting to want more frequent updates, according to Luby. "We are finding that these are not one-offs," he says. "They are starting to get more frequent, and so what you thought you were only doing once, now you have to do regularly." Retail customers, in particular, had an interest in having access to up-to-date information through organizations such as ECCnet, since the industry is moving to use these types of groups as centralized product data repositories — although McCain still had to deal with ad hoc requests for product data, too.
McCain divisions around the world did make moves to automate the process of collecting the information to some extent, using data extracts, Excels spreadsheets and other semi-mechanized tools. But the ad hoc systems and processes put in place were not consistent across the company, and moreover they frequently had to be patched and re-patched together to comply with new or changing requests over time. "It got to be rather 'kludgey,'" Luby says. "And people were doing this all around the world."
Eventually, McCain conceived of the idea of moving toward establishing a centralized repository for the company's product information and decided to start off in the company's Canadian division — which employs about 3,300 people, has 11 processing facilities and represents a considerable percentage of the total frozen food dollar in Canada. The idea was to set up a storehouse of product attribute data that would also incorporate transformational capabilities so that users could set up customized but reusable templates with the proper information and formatting required by individual customers, industry registries like ECCnet and other requesters. This "trusted source" of product attribute data would be fed from other systems within the company.
Luby and his team evaluated a variety of product information management solutions for use within McCain Foods before settling on an application called I-Accel from solution provider FullTilt. I-Accel is a Web-based product information management solution that offers a shared enterprise-wide repository, configurable workflow and role-based security. The solution was designed to help companies to aggregate and create product data from disparate sources and to standardize it using product domain expertise and industry standards.
McCain and the solution provider announced the selection of I-Accel in September 2003, and the deployment was set to go live in January 2004. The solution will not replace the company's enterprise resource planning (ERP) system as the master record for certain data, such as part number, but for information such as a product's official recipe, the "trusted source" will be the I-Accel database. "You might have six versions of the recipe somewhere else in the company, but if anyone in the company were to get a phone call from someone who needed the 'publishable' recipe or list of ingredients, then [I-Accel] is where you go," Luby explains.
Getting the database established and the necessary "hooks" connecting the repository to other systems has been relatively straightforward, according to Luby. "This is not rocket science," he quips, noting, however, that in order to feed external requestors the company has had to go through a largely manual task of "cleansing" the data, whether for the individual requests or for feeding the database — not an overly onerous task, given that a McCain division may have a couple hundred products, but nevertheless very important, particularly when competing data exists within the company. Currently McCain intends to populate the database with about from just under one hundred different attributes for some products to several hundred attributes for others. Luby did not expect to have all the hooks automating data feeds to the repository in place by the January launch date, but the project team planned to work in a follow-up phase to ensure that all the loading and data transfers become more fully automated than in the initial phase.