Category identification refers to selecting the items that should be included in a particular category as well as selecting what categories that should be prioritized.
A category encompasses a group of similar items that are required for specific business activities of the firm. A category can also be defined as ―a group of coherent products or services, bought from the supply market that are used in our company to satisfy internal or external customer demands‖. Examples of categories are casting, bottles, parts, sheet metal, and tires. The most important characteristic of a category is that it must mirror how the individual marketplaces are organized.
The two definitions provided are not necessarily contrasting views, but the focus is slightly different. It can be complicated to find categories that truly mirror the individual market places, however. For instance, a company might be tempted to choose ‗travel‘ as a category based on that there are travel agencies that are offering travel solutions in all forms. But traveling is not a real market, and agencies are rather intermediaries who facilitate communication between its customers and multiple real markets: ‗air travel‘, ‗hotels‘, ‗car rental‘ etc.
A thorough spend analysis is the starting point for a good category identification. However, the complexity of many decentralized purchasing organizations can make it a complex and long-lasting project to get good quality data. For instance, a major petroleum company spent some six months just to get a picture of what they bought at different business units. Common hurdles are low data quality and different IT and coding systems across business units. It is not unusual even for large companies to have to ask their major suppliers what they are buying in order to get even a sufficiently good picture of what they are buying. In fact, few large corporations have fully integrated information systems where they can access data on group level. Growth through acquisitions will elevate the complexity in spend integration further. Good quality data that can be aggregated over all business units is not enough however, but the data must be coded so that it can be sorted by means of a spend cube: per category or item, per supplier, and per cost center.
When spend data is available, companies can use the Pareto principle to sort the data and focus their categorization on the 80 percent of the spend that are with 20 percent of the suppliers. Furthermore, companies have to choose how much of total spend they will aim to categorize. Total spend can be categorised in the three groups: ‗categories‘, ‗non-addressable spend‘, ‗and rest of spend‘.
Non-addressable spend is spend that is almost impossible to influence, like tax or government-set license fees. Rest of spend is the part of total spend that is not categorized because it is not economically viable. MRO goods, commodities, and indirect materials are the most preferred item groups when companies start a pooling initiative. In addition, leverage and routine categories are most likely
to be prioritized from a portfolio perspective, using the Kraljic model. Although portfolio models are frequently and successfully used in many purchasing activities, the approach results in categories substantially different from what is normally used in industry. Both current and future spend should be included in an analysis in order to create a comprehensive picture of the spend on a particular category. This is particularly important for categories with a high variation in spend over time.