Enterprise Search Challenges: Why Staff Can’t Find What They Need

A staff member types a question into the firm’s search and gets nothing, a wrong file or an old one. The reasons sit in how the firm keeps, words and shares its documents, more than in the search box itself. This explainer sets out five enterprise search challenges a manager can check without a tool. It also says who in the firm can fix each one, and it follows an invented case that shows them together. Each cause is a question of knowledge management, the way a firm keeps, labels and shares what it knows, as much as a question of search. It sits under the Signisys use case on finding answers across business documents.

Enterprise Search Challenges: The Five Causes in Brief

In plain words, a person asks a question in the firm’s search. The reply is no answer, a wrong one or an old one. Enterprise search, here, means one place from which staff can look across the firm’s own documents and systems. The Wikipedia entry on enterprise search names the kinds of sources such a system draws from, among them file stores, intranets, document management tools, email and databases.

Five causes sit behind a missed answer. Each one shows up in the documents, or in the people around them, before it shows up in the search box.

The sections below take the causes in pairs, then the fifth on its own.

Where Content Lives and Who Owns It

Cause 1 is content spread across systems. The answer exists, yet it sits in a shared drive, an email thread, a ticket system or a chat tool. Each of those places may keep its own separate index. One question can then take several searches, and the person asking may stop after the first. Some content is also locked inside scanned forms or images, where a search over text can find little. Pulling text out of scans and forms is a separate job for those files.

Two questions help here. Where does your team keep the answer to its last missed question? Would a new joiner know to look there?

Cause 2 is files with no owner or date. A policy may exist in two versions, one on an old drive and one sent round by email. Neither shows a date or a name. The person searching finds both and has to guess. In that case, the search did its part, but the documents left the person with no way to tell which one counts.

Two more questions apply. Who owns this file? When did someone last read it and confirm that it still holds?

Neither cause is about the search box. Both sit in where the firm keeps its documents and who answers for them.

How People Ask and Who May See What

Cause 3 is question wording. Staff ask in plain words, while documents carry the firm’s internal titles and codes. For example, someone types “notice period for leave”, and the answer sits in a file called “HR-Policy-v7”. Search relevance is how well the order of results fits what the person meant. In a search that matches words, a result can rank high on its words matching the question, which is not the same as answering it. A file with the right words and the wrong content can sit above the file the person needs.

Cause 4 is access rules. The same Wikipedia entry calls security a central requirement for enterprise search and names access control lists, set per document, as one way access is controlled. A rule set too tight hides a document from the person who needs it. A rule set too loose shows it to a person with no need of it. That is a different problem, though it starts in the same place. OWASP’s guidance on securing AI search systems describes the risk from the other side: a person’s query returning a passage from a restricted document, when the document’s rules were not carried over to its passages.

Ask who sets access for the folder that holds the answer, and whether that person knows who needs it. Rules for who may see what matter even more once AI tools read the same documents. That is where setting rules for AI use across the firm comes in. Which route suits those documents is weighed in a comparison of search tools and answer-writing assistants.

Nobody Reads the Searches That Found Nothing

Cause 5 is the one nobody sees. A failed search is a search that returned nothing useful. A list of them shows what people looked for and did not find. When no one looks at that list, the same miss repeats.

For example, a staff member types “laptop replacement” and gets no result; the IT standard is titled “end-user device refresh”. Nobody sees that search, and the wording gap stays open. The next person who asks then meets the same blank page, although the answer existed the whole time.

In that sense, cause 5 sits on top of the other four. It decides whether anyone in the firm finds out about them at all.

Enterprise Search Challenges: Who Fixes What

In Signisys’ working description, each cause has a question to ask and a role in the firm that can act on the answer.

Start with the cause you can already see in a recent missed answer; that one comes with a person and a file attached. The order is a way to think, not a rule, and a single miss can involve more than one cause. Each fix sits with a role rather than a tool, which gives the knowledge management side of the work named people.

Enterprise Search Challenges: One Missed Answer Followed Through

The firm in this example is invented to show the causes, and it is not a Signisys project. For example, a staff member at a services firm wants the travel claim limit before she books a trip. She types “travel claim limit” and finds two files, one from an old shared drive and one with no date at all.

Followed through, her miss touches all five causes. The current policy sits in a different system from the two files she found, which is cause 1. Neither of those two names an owner or a date, which is cause 2. She has no way to tell which one holds. Meanwhile, the current file calls the figure a “reimbursement ceiling”, which her words did not match; that is cause 3. It also sits in a finance drive she may not open, which is cause 4. Finally, nobody had read the searches that found nothing, and earlier misses of the same kind went unheard; that is cause 5.

A single fix may not close this miss. The finance owner of the policy, the person who sets access and the IT lead each hold one piece of it. That is why the table names roles instead of tools.

Questions Only Your Own Documents Can Answer

This explainer does not settle which cause sits behind your own firm’s misses. Nor does it show whether your documents are in a state to be searched, or what a given fix would involve in time and people. Those answers sit in your own files, folders and access settings. For the order of work, see a guide to the six steps of AI search over company documents. For a record-based check of the documents themselves, see 15 scored checks, each needing a dated record to pass.

Ask for an outside view when staff keep missing the same answer and nobody can say whether the document exists.

Signisys provides data and AI services through its enterprise AI solution and its data and AI service page. The box below describes a Signisys service, not neutral advice, and it makes no promise. Bring one question your staff were unable to answer last month as a place to begin.

Frequently Asked Questions on Enterprise Search Challenges