Some bullets on search business case
In thinking through the question I posted, I came up with a few bullet points. Perhaps people can add to these:
- Building the search case is like any other business case – its about solving specific problems for specific audiences
- Align with goals that are strategically important to the organization. Understand the drivers to achieving those goals.
- Business justification is about detailed use cases. Build these use cases before trying to develop a solution or choose a technology
- Enable specific processes that are part of business imperatives (for example customer acquisition if sales growth is the imperative) and determine the findability bottlenecks in that process
- Who are the primary audiences, what are their tasks and how do they locate the answers they need?
- Pinpoint “information leverage points” or “high velocity decisions” where answers are needed quickly. What is the cost to not getting the answers or getting the wrong answers? Do those costs multiply or cascade through the organization? What is the long term cost? Decreased competitive advantage? Poor customer service? Lost revenue? Lost customers? Erosion of the brand? Impact on future capabilities?
- Determine what is currently being measured about the process or task and enroll experts from finance and operations when building the justification. If the accountants believe the numbers it will be easier to sell.
- Recognize that data is heterogeneous – it will vary in terms of structure, application and metadata. Search mechanisms need to take this into consideration. No one size fits all
- Different search tools are appropriate for different levels of structure and metadata in content. Untagged content can be well formed and thus amenable to entity extraction and autocategorization. Poorly formed content is less appropriate for these approaches
- Test use cases as a baseline to measure success against. People are afraid of baselines because they can also reveal failure – but know what does not work is as important as what does
- Help people understand that a search appliance is not the answer by testing simple (and ambiguous queries) against complex information. Show by way of example when full text search returns results with low levels of precision.
- Think of search as a platform to integrate information sources
- Each class of search technology is a tool in a toolkit, not a monolithic application.
- Search applications need to be tuned, configured, integrated. Different algorithms are appropriate for different content sources and structures. Test the usefulness of tools against derived use cases
- Create scenarios to evaluate recall and precision against test data. This way you can achieve meaningful comparisons between approaches
EARLEY & ASSOCIATES, Inc.