SharePoint Migration Without a Knowledge Architecture
Northwind Transit Authority is six months into a SharePoint Online migration for its capital program. The PMO supports 14 active projects across three regions and 210 staff. An internal audit found that 62% of searches for an approved document fail on the first attempt; engineers routinely keep copies in personal OneDrive folders; and two regions use conflicting naming conventions for the same design deliverable. The migration team proposes recreating the old folder hierarchy exactly to reduce complaints. The PMO sponsor wants a single "source of truth" before the next funding gate in eight weeks, but project managers say taxonomy work will delay delivery. There is no dedicated knowledge manager, and metadata completion is currently about 35%. A recent incident required four days to locate the latest safety inspection report.
1. Diagnose the root causes of the search and version-control problem. Which causes are structural rather than behavioral, and why should the PMO not treat this as a training issue alone?
Read the worked answer
The visible symptoms-failed searches, personal copies, conflicting names-are downstream effects of an absent knowledge architecture. Structurally, the organization has no agreed controlled vocabulary for projects, disciplines, deliverable types, phases, or status. Folders are being used as the classification system, but folders cannot express multiple dimensions and they duplicate across regions. Metadata is optional in practice, so even correct uploads are not retrievable by faceted search. There is also no accountable content owner for each deliverable family, which means no one resolves duplicate or superseded versions. The personal OneDrive behavior is partly rational: staff store copies where retrieval is fast and version risk is invisible to them. The safety-report incident shows the cost of missing authority signals-no approved-version marker, no lifecycle state, no owner. Training alone would teach people to use a system that still lacks consistent labels, ownership, and search tuning. The PMO should treat this as a socio-technical design problem: minimum taxonomy, mandatory metadata for controlled documents, named owners, lifecycle states, and search configuration. Behavioral norms then reinforce the architecture rather than compensate for it.
2. Propose an eight-week intervention plan. State what you would deliberately defer and what trade-offs you are accepting.
Read the worked answer
Weeks 1–2: define a minimum viable taxonomy rather than a comprehensive one. Use 5–7 facets-project, discipline, deliverable type, phase, status, region, and retention class-with a short picklist and a naming standard for controlled documents. Assign a content owner for each high-risk deliverable family (safety, design, regulatory, financial). Weeks 3–4: configure search filters, promoted results, and required metadata for new controlled documents; pilot in two projects. Establish a default save location and restrict personal sync for approved records; do not attempt to eliminate every local working copy. Weeks 5–6: migrate only active and reusable high-risk content first, marking legacy material as unverified where needed. Run short role-based sessions focused on finding and publishing, not generic tool training. Weeks 7–8: measure first-attempt search success, metadata completeness, duplicate rate, and time to locate the latest safety report; fix the worst search gaps before the funding gate. Deliberately defer wholesale legacy cleanup and a full file-plan taxonomy. The trade-off is that old documents remain inconsistently described for a period. That is acceptable because the immediate risk is current controlled documents, and a big-bang taxonomy would consume the eight weeks without improving retrieval.
3. Define success measures that would convince the sponsor the intervention is working, including at least one measure that guards against gaming.
Read the worked answer
Use outcome measures tied to the original symptoms: first-attempt search success for a fixed test set of approved documents; median time to locate the latest safety inspection report; percentage of new controlled documents with complete mandatory metadata; number of duplicate approved versions per deliverable family; and percentage of active projects with a named content owner. Add a guard against gaming: pair metadata completeness with a retrieval test using independent users who did not create the documents. Otherwise teams can satisfy metadata fields with default or meaningless values. Also track exception rate-documents exempted from the standard-because a very low exception rate may mean the standard is being ignored rather than followed. Establish baselines before changes, use the same test documents each month, and report trends rather than one-off snapshots. The sponsor should see both adoption and retrieval outcomes; page views or upload counts are weak proxies and should not be primary success measures.