Connecting Your Data
On this page
Search is only as good as the data you feed it
The power of Enterprise Search depends directly on the quality and breadth of connected sources. A deployment that indexes only the executive shared drive will yield poor answers. Users will conclude the tool doesn’t work, when the real issue is scope. This lesson follows the actual order of a project: inventory, connect, tune indexing, then maintain.
Start with an inventory based on four questions. Ask the teams themselves, not just IT. Which tools do people actually use daily, including shadow IT? Where does critical knowledge live—technical documentation, procedures, internal policies? What content do employees search for most often? Which data is sensitive enough to require access restrictions? The answers set your priority order. That order almost never matches technical ease: connect the sources containing the most requested, up-to-date knowledge first.
Next, verify prerequisites. They are identical for all connectors. You need admin rights on the source platform—Google Workspace Admin or SharePoint Admin, for example—a service account or OAuth app dedicated to Enterprise Search, read permissions on the spaces to index, and approval from your security team. Security approval is the most common schedule killer. Request it during the inventory phase, not the day before go-live.
Connecting major sources
Connecting Google Drive uses OAuth 2.0 with a Google Workspace service account. Follow these steps:
- Create a service account in the Google Cloud Console
- Enable the Google Drive API for that account
- Enable domain-wide delegation in Google Workspace Admin
- Configure the required read scopes
- Enter the credentials in the Enterprise Search console
The connector then indexes Google Docs, Sheets, Slides, PDFs, and other formats stored in Drive, automatically respecting sharing permissions defined in Google.
For a Microsoft 365 environment, the logic is identical but authorization flows through Azure. Register an app in Azure Active Directory, grant it Microsoft Graph permissions Sites.Read.All and Files.Read.All, obtain admin consent, then configure the connector with that app’s credentials. This indexes SharePoint sites, document libraries, and OneDrive files, faithfully inheriting SharePoint permissions.
Confluence connects faster: create an Atlassian API token with read permissions, specify the spaces to index or choose the entire site, then set the sync frequency. If your organization uses it as its primary wiki, prioritize this connection—it is almost always the richest source of technical documentation and internal procedures.
Slack requires a dedicated app in the workspace. Add the OAuth scopes channels:history and channels:read, install the app, and select channels. Resist the temptation to index everything. Focus on project, support, and documentation channels. Leave informal channels aside; they add noise without providing actionable knowledge.
Tuning indexing
The initial sync can take several hours depending on volume, and it loads your source systems the entire time. Schedule it outside peak hours. Afterward, Enterprise Search detects changes automatically. New documents are indexed within minutes of creation. Modified documents are reindexed automatically. Deleted documents disappear at the next sync. Permission changes propagate to results. You configure frequency per connector: real-time via webhooks, every fifteen minutes, hourly, or daily.
Filtering is the second lever for quality. Include or exclude folders to index only relevant spaces. Filter by file type to discard images, videos, or large files with no textual value. Filter by age, keeping only documents from the last two years, for example. Exclude naming patterns to ignore temporary files, drafts, and archives. Neglecting this setting has a precise cost: an outdated sales policy that remains indexed will eventually produce a wrong answer that looks perfectly sourced. This is the worst-case scenario because nothing alerts the reader.
Relevance also hinges on metadata. A document with a descriptive title ranks higher than a file named “v2_final.” Existing tags and labels enrich indexing. Descriptions from Confluence pages or Drive files are leveraged. Author information helps weight reliability. Finally, how users phrase questions matters as much as configuration. Encourage complete questions rather than isolated keywords. Show your teams that “What is the onboarding procedure for back-end developers?” yields a completely different result than “onboarding.” Share a few examples of effective queries internally.
Long-term monitoring
The admin console displays indexing status: number of documents indexed per source, last sync date, indexing errors and failed documents, search volume, and satisfaction rate. Check it monthly. Verify that connectors remain active. Investigate failures—usually revoked permissions or corrupted files. Adjust filters based on feedback. Connect new sources as adoption grows. A broken connector triggers no visible alert for users. Answers simply get worse, and no one makes the connection.
Key takeaways
- Start with a data source inventory and prioritize the most searched items
- Each connector requires a service account with read permissions
- Incremental indexing keeps results up to date automatically
- Filtering and exclusions refine relevance
- Continuous monitoring ensures result quality and freshness