Enterprise search – just some of the challenges
It seems to me that many enterprise search vendors are selling an ‘instant AI solution’ and their marketing suggests that they have little understanding of the requirements and challenges of enterprise search, especially in the case with multi-national multi-language organizations. Unless an enterprise search vendor has experienced enterprise search managers on their staff and a range of established clients, they will inevitably have an inadequate understanding of why enterprise search is an outlier to other ‘search’ applications. Research by Paul Cleverely indicates that technology, poor quality content and a lack of training and mentoring are equal factors in search dissatisfaction, and the list below is categorized by these factors. This list is based on my experience with around 50 large-scale enterprise search projects. Sam Marshall, CEO Clearbox Consulting, published a very good schematic for diagnosing enterprise search failures based on his project experience which is very much in line with this list.
Technology
- Enterprise search is federated across multiple applications and multiple business entities.
- There are multiple points of failure at collection, index, query and ranking levels.
- Security mapping at a document level and an employee level will be required.
- AI/ML/NLP personalization and recommendation features, and security policies, play havoc with the interpretation of search analytics.
Content
- Massive amounts of structured and unstructured content (perhaps 500 million files), including text, data, image and video file formats.
- Rarely is this content curated for quality because there is no incentive to make it findable and no corporate information management policies and governance.
- Content is in multiple, and often mixed, languages with extensive use of corporate jargon and abbreviations, together with a wide range of entities such as product codes and short names for office locations.
- In many multinational companies the majority of employees may not have English as their primary language, which impacts query formulation and results scanning.
- Changes in business direction (mergers, acquisitions, product expansion) can take place with no notice and yet cause major instantaneous changes in query topics.
- The requirement to identify employees by name, role, experience and responsibility can account for 50-60% of all searches.
Training
- Employees may not be searching for their personal use (e.g. searching on behalf of team members) so past search history, recent documents and notional current role will be poor guides to search intent.
- Users have multiple roles and therefore multiple personas and information requirements.
- Around 10% of employees will leave each year and perhaps another 10% will take on new responsibilities so there is a significant on-going training requirement
- No senior level owner of ‘search’ and usually inadequate investment in a search support team with a combination of a knowledge of the business and experience with a range of search applications
- There has to be an employee-managed balance between precision and recall
The Bottom Line
Failure to find information that the organization has invested in compiling is not an option as it puts at risk the organization’s reputation and performance and the career prospects of employees.
Martin White Principal Analyst