1. Account, users and permissions
- Domains connected and verified (email sending, tracking, CMS if used).
- Users invited with role-based permission sets, not blanket super-admin.
- Teams configured to mirror your actual org, so record ownership and reporting work.
- Two-factor authentication enforced and a named portal owner recorded.
- Sandbox or test environment agreed for anything risky.
2. Data model and properties
This is where implementations are won or lost. Decide which objects you'll use, which properties are required at each stage, and — critically — which properties you are deliberately not creating.
- Object map: contacts, companies, deals, tickets, and any custom objects with a written justification each.
- Property audit: name, type, description, who fills it, and what report it feeds. If nothing consumes it, don't create it.
- Dropdown values standardised before import (country, industry, source), never free text.
- Association labels defined for multi-company and multi-contact deals.
- A naming convention document, applied to properties, workflows, lists and forms.
3. Pipelines and lifecycle stages
Every stage needs an exit criterion a rep can verify — an event that happened, not a feeling. 'Qualified' means nothing; 'budget confirmed and decision-maker on the call' can be checked.
- 01Write lifecycle stage definitions and get sales and marketing to sign the same sheet.
- 02Map deal stages to those definitions with required properties per stage.
- 03Agree stage probabilities with finance so the forecast is credible.
- 04Define what happens to a lost deal — recycle rules, not a graveyard.
4. Migration and reconciliation
- Full export from the source system, with a record count snapshot taken and dated.
- De-duplication and normalisation before import, not after.
- Field-level mapping document reviewed by someone who uses the data daily.
- Staged import: companies, then contacts, then deals, then activities and notes.
- A reconciliation report proving record counts and key field values match the source.
5. Integrations, automation and reporting
- Integration inventory with direction of sync and system of record per field.
- Lead routing rules written down before they're built, with fallback owners.
- Workflow naming convention applied from the first workflow (our free generator enforces one).
- Core dashboards: pipeline, forecast, source attribution, activity — built before go-live, refined after.
- Data quality alerting: unassigned records, stalled deals, missing required fields.
6. Adoption — the step everyone skips
A portal nobody uses is worse than a spreadsheet, because now there are two sources of truth. Run role-based training, not a single all-hands walkthrough: reps learn the pipeline, marketers learn lists and workflows, managers learn reporting.
Then hold a hypercare window — two to four weeks where someone owns the fix queue and every 'this is annoying' comment gets triaged rather than ignored.
Common mistakes
- Building workflows before lifecycle stages are agreed in writing.
- Importing everything 'just in case' and inheriting a decade of dead data.
- Giving everyone super-admin because permissions are fiddly.
- Launching without a single dashboard, so leadership never adopts it.
- Treating training as a recording nobody watches.