The checklist
- One CRM, named the system of record out loud
- Objects and lifecycle stages match how buyers actually move
- Fields nobody fills are deleted, not hidden
- Enrichment writes automatically from the intelligence layer
- Every outreach touch and reply logs itself
- Dedupe runs on domain, continuously
- Routing gets a new lead to an owner in minutes
- Dashboards answer the weekly verdict, not decorate slides
- Sales trusts the data enough to work from it daily
- Any future hire can learn the model from one document
Why CRMs rot
Only about a third of sales professionals say their organisation fully uses its CRM. The rest live with a graveyard: duplicate accounts, fields added for a campaign in 2023, lifecycle stages nobody can define. The rot is not a tooling problem. It is what happens when a system of record has many writers and no architecture.
In our stack the CRM has one job: hold the truth. Everything else, enrichment, outreach, scoring, happens in layers built for those jobs, and reports their results into the record. The moment the CRM is also the workbench, it stops being the truth.
Model the pipeline you actually run
Start from the buyer journey you can observe, not the one in a template. Each lifecycle stage needs an entry condition a machine can evaluate. If two people would argue about whether a record is an MQL, the stage is an opinion, not a stage.
Then delete. Every unused field is friction that teaches the team the CRM does not matter. A lean model that is 100 percent filled beats a rich one that is 40 percent guessed.
Wire the inputs so humans never copy paste
Enrichment writes from Clay on a schedule and on demand. Sequencers log sends, opens and replies as activities. The orchestration layer routes new leads to owners and stamps the source. When no data enters manually, data quality stops being a discipline problem and becomes a property of the system.
Give reps one convenience: an enrich now button on the record. It builds trust in the machine faster than any training session.
Hygiene runs as automation, not as a quarterly cleanup
Dedupe matches on domain, merges automatically, and flags conflicts for a human. Bounced addresses get marked the day they bounce. Ownerless records escalate. A rot dashboard shows fill rates and duplicate counts, and its numbers belong in the weekly review next to pipeline, because bad data is pipeline you cannot see.
Dashboards that end in verdicts
Every dashboard answers one question: what did the experiments produce and what do we kill, iterate or scale. Pipeline by segment and by signal source, replies per sequence, conversion between stages. If a chart cannot change a decision on the weekly review, it does not earn a spot.
Five CRM failure modes
- Two systems of record
The CRM and a spreadsheet that is secretly the real one. Pick a single writer of truth and make the spreadsheet read only.
- Stages as opinions
Lifecycle stages without machine checkable entry conditions produce reports everyone interprets differently.
- Manual data entry as policy
Whatever depends on a rep typing after calls will be missing by Friday. Automate the write, keep the human for judgment.
- Field hoarding
Every field someone might need someday costs trust today. Deleting fields is a feature release.
- Dashboards as wallpaper
Reports nobody acts on train the team that numbers are decoration. Fewer charts, each attached to a decision.
What good looks like
| Metric | Target |
|---|---|
| Required field fill rate | above 95 percent |
| Duplicate accounts | under 2 percent |
| New lead to owner | under 15 minutes |
| Manual data entry in the funnel | zero steps |
| Reports used in the weekly verdict | every single one |