1–3: the data mistakes
- 01No agreed data model. Properties get created ad hoc by whoever needs them that week. Two years later you have four fields that all mean 'industry' and no report that works. Fix: one property register with an owner and a consumer per field.
- 02Migrating dirty data. Duplicates, dead records and free-text country fields all survive the import and multiply. Fix: clean at source, de-duplicate before import, normalise dropdowns first.
- 03No reconciliation. The import 'looked fine', and nobody proved record counts and key fields matched. Fix: a written reconciliation report signed off before the source system is switched off.
4–6: the process mistakes
- 01Lifecycle stages nobody can verify. If 'Qualified' is a feeling, your funnel report is fiction. Fix: every stage gets an observable exit criterion.
- 02Workflows with no naming convention. Fifty workflows called 'Copy of Copy of Lead Flow' is how portals become unmaintainable. Fix: apply a convention from workflow one — our generator produces them.
- 03Automation built before testing a single real record journey. Fix: walk one record end-to-end manually first, then automate exactly that path.
7–9: the people mistakes
- 01Blanket super-admin access. Someone deletes a property that fed the board deck. Fix: role-based permission sets and a named portal owner.
- 02No adoption plan. Training was one all-hands call, and reps quietly kept their spreadsheet. Fix: role-based training plus a hypercare window with an owned fix queue.
- 03No documentation handover. The agency leaves and the knowledge leaves with them. Fix: contract the deliverables — data model diagram, workflow inventory, naming conventions, lifecycle definitions.
How to check whether your portal has these
You can find most of them in ninety minutes without any tooling: count properties with no data in the last year, count workflows with no description, check whether two dashboards disagree on the same number, and ask three reps to describe what 'MQL' means. Divergent answers are the diagnosis.
We wrote the full walkthrough as a standalone method you can run yourself.
Common mistakes
- Assuming a clean-looking portal is a healthy one — the damage is usually in the data model.
- Fixing symptoms (a broken report) instead of causes (an undefined lifecycle stage).
- Rebuilding everything at once instead of sequencing fixes by risk.
- Switching off legacy workflows without tracing what depends on them.
- Not documenting the fix, so the same drift restarts immediately.