Step 1 (days 1–3): connect with scoped permissions and record the baseline
Connect the CRM to the SDR Digital FTE with a dedicated token that can read records and propose changes, and cannot delete, merge, or change a stage without approval. Record the baseline: share of opportunities with a current next step, opportunities with no activity past your staleness threshold, duplicate contacts found by a simple match, required fields empty, and the difference between last quarter's forecast and its outcome.
Write the protected list into the remit: accounts whose records the FTE may flag but never propose changes to without the account owner's approval.
Step 2 (days 4–10): surface the backlog as proposals with evidence
The FTE reviews the CRM and raises proposals grouped by rep: stale opportunities with the last activity date; stage corrections with the logged activity that suggests them, such as a reply thread agreeing a demo; duplicate contacts with the matching fields side by side and the fields that would be kept; empty fields with a proposed value from a public or connected source. Every proposal shows its evidence.
Expect the first week's list to be long. That is the backlog, and it is the reason the cleanup is a month rather than a morning.
- Stale: last activity date, proposed next step for the rep
- Stage: current, proposed, and the activity behind it
- Duplicate: both records, matching fields highlighted
- Field fill: field, value, source
Step 3 (days 8–20): reps confirm in minutes a day
Each rep opens their group each morning and confirms or corrects, which takes seconds per item because the evidence is on the screen. A correction with a reason teaches the FTE: a rep who knows an account is not stale corrects the flag, and repeated corrections of one kind become a remit rule at the weekly read. The sales leader reads the send-back reasons weekly and adjusts thresholds and criteria.
Reps who dislike CRM admin tend to like this, because the admin now arrives as yes-or-no questions with the evidence attached.
Step 4 (days 15–25): promote the low-risk class, keep the rest gated
Once field fills from public sources show a clean run of confirmations, promote that one class to auto-apply within bounds, such as company size and industry only, never on protected accounts. Keep stage changes and merges gated, because a wrong stage affects the forecast and a wrong merge is hard to undo. Reminders to reps about overdue next steps are the next candidate, after the cleanup.
Log the promotion, and expect to step it back on the first correction.
Step 5 (days 26–30): measure against the baseline, in the forecast
At day 30, compare with the baseline: share of opportunities with a current next step, stale opportunities open, duplicates open, required fields empty. Then look where the cleanup actually shows: the pipeline review that no longer starts with cleaning the data, and stages that match logged activity, which is what makes next quarter's forecast closer to its outcome. Set the steady state: hygiene proposals continue daily and reps confirm them in a minute.
Run the pipeline diagnostic mission once now, because the data is clean enough to answer why pipeline moved with ranked hypotheses.
What does done look like?
Done is a CRM whose opportunities have current next steps, whose stages match activity, whose duplicates are merged with a record of what was kept, and whose required fields are filled from stated sources; reps who spend a minute a morning on confirmations instead of an afternoon a quarter on cleanup; and a sales leader whose forecast is built on stages that mean something. Nobody's notes were touched and nothing was deleted.
The steady state is what the cleanup was for; the month was the backlog.
What should you refuse to do?
Refuse to let the FTE change stages or merge records without approval during the cleanup, however clean the proposals look. Refuse to touch protected accounts without the owner's approval. Refuse to measure by proposals confirmed; measure by next-step coverage and forecast accuracy. And refuse to skip the reps' confirmation to go faster, because their corrections are what make the remit right.
The refusals are what keep a cleanup from becoming a data incident.