Migrating from manual ratesheet entry
If your team still copy-pastes ratesheets every morning, here's the migration playbook to email-in + portal automation. Two weeks. Concrete steps.
The 7am ratesheet ritual — log into vendor portals, download PDFs, copy cells into the LOS — is the most expensive recurring cost in many small and mid-size shops. It also has the worst error rate of anything in the pipeline because humans make typos and the LOS doesn't always catch them.
Migrating to automation is a two-week project on a Team or Business tier. Here's how it sequences.
Week 1, day 1: inventory
List every investor and the source you currently get their ratesheet from (email attachment, vendor portal, FTP, custom API). Note the format (Excel, PDF, image-PDF, HTML). Note the typical morning arrival time and the typical activation time.
Week 1, day 2-3: email-in
Configure email-service with an IMAP inbox dedicated to ratesheet ingestion. Forward your existing ratesheet emails there. As mail lands, email-service emits documents.raw.received and the pipeline kicks in.
Most investors will land successfully on the first sheet because we seed mapping templates for the common ones. The rest will arrive as DRAFT with low-confidence mapping warnings; QC them, resolve the mappings, and the template persists for next time.
Week 1, day 4-5: portal automation
For investors that don't email, configure agent-service with portal credentials. Playwright handles the login and the download; the rest of the pipeline is identical to email-in.
Portal automation is best-effort — it can break when vendors redesign UIs or add bot-detection. We surface failures explicitly; you fall back to manual download in those cases. In practice, 90%+ of portals stay working long-term.
Week 2, day 1-2: QC dashboard training
Train your team on the QC dashboard. The two skills are: read the confidence-score column, and resolve a low-confidence mapping by clicking on the suggested column header. Most QC items resolve in under a minute per ratesheet.
Week 2, day 3-5: parallel run
Run RateStack ingestion alongside your manual process. Compare the activated ratesheet's rows against your manually-entered rows. Differences fall into three categories: investor sent slightly different data (rare; reconcile with investor), manual entry typo (common), or mapping issue (rare after QC).
Week 2, day 5: cutover
Once the parallel run shows manual entry losing more often than RateStack ingestion (typically by day 3 of week 2), cut over. Disable the manual entry workflow. Reassign that team to QC and exception handling — the quality work, not the typing work.
What to expect after cutover
Vendor changes (column reorder, new column added) absorb automatically once mapping templates have learned them. Net QC time after one month typically drops to under 15 minutes per day across the entire investor pack.
The team that used to spend two hours every morning on data entry will have those hours back. Most shops we've worked with reinvest them in higher-value QC, scenario exploration, or just better customer conversations.