I'm planning Q4 and have room for one cleanup project before the new year. Picking just one forced me to decide what actually has to come first, so here's where I landed.

Most teams start with duplicates because the count is easy to watch go down. But deciding which record wins a merge depends on your owner and stage rules. If you dedup first, you end up with tidier bad data.

The order I'd use, based on how much each one breaks downstream:

1. Stage definitions. If "qualified" means something different to marketing, sales and your CEO, then the forecast, the routing and every report are reading the same broken field. Next year's targets get set off that forecast, and you're the one defending the number in January.
2. Ownership. Records owned by reps who left, or by nobody. This is common after a merger. Those leads sit there and nobody gets alerted.
3. One source of truth per field. If two systems both write to lifecycle stage, they'll keep overwriting each other no matter how often you clean.
4. Duplicates. By now you have rules for which record wins.

A quick way to find your starting point: if your VP of Sales keeps a spreadsheet next to the CRM, look for the column that doesn't exist in the CRM. That's usually #1 or #3.

The team I take on will probably be one that merged or migrated in the last year or two, since all four break at once there. If that's you, comment what happened and what the side spreadsheet tracks, and I'll tell you which one I'd fix first. Same offer for anyone else reading.