CRM Implementation: A 90-Day Plan
A CRM implementation runs in four phases over 90 days: decisions and data model, core build, migration and integrations, then automation, reporting, and adoption. The decisions phase determines success — object model, lifecycle stages, and stage exit criteria agreed and signed before anything is configured.
- Configure nothing in the first two weeks. The decisions take longer to agree than the build takes to execute.
- Migrate less than you think. A cutoff at 18–24 months is almost always right.
- Test integrations bidirectionally, including delete behaviour. That is where the expensive surprises live.
- Adoption is a build task with its own budget, not a training session at the end.
- Name one internal owner with decision authority. Implementations without one add four weeks.
Why CRM implementations fail
Almost never for technical reasons. Modern platforms do what they claim, and competent implementers can configure them. Implementations fail because the company never decided what the system should represent, so the configuration encodes three people's differing assumptions and produces numbers nobody trusts.
The framing that prevents it: an implementation is a decision project with a configuration phase attached. The configuration is the easy part and the part everyone plans for.
| Failure | Root cause | When it surfaces |
|---|---|---|
| Nobody trusts the reports | Stage definitions were never agreed | Month 4 |
| Reps work around the system | Required fields they cannot answer | Week 3 after go-live |
| Duplicates everywhere | Contact-first model in a B2B motion | Month 6 |
| Timeline overran | Migration scope was never bounded | Week 5 |
| Shadow spreadsheets | Leadership ran the number outside the CRM | Month 2 |
Phase 1 — Decisions (weeks 1–2)
No configuration happens here. The output is a signed document.
- Object model. What accounts, contacts, opportunities, and any custom objects represent. In B2B the account is the primary object — building around contacts produces a duplicate problem that compounds for years.
- Lifecycle stages and the objective entry condition for each, plus which system sets it.
- Deal stages with exit criteria that someone who was not on the call could verify. If the criterion is a feeling, it is not a criterion.
- Deal creation trigger. If a deal is created at first conversation, your pipeline report is a conversation report.
- Source of truth per data point. For every important fact, one owning system and the direction the sync runs.
- Required fields only where a decision or routing rule depends on the value.
Phase 2 — Core build (weeks 3–5)
- 01Objects, fields, and naming
Build only what the signed model contains. Use consistent naming and fill in the description on every field — an undocumented field becomes an orphan within two staff changes.
- 02Pipelines and stages
One pipeline per genuinely distinct sales process, not per team or product. Multiple near-identical pipelines make every cross-motion report a manual reconciliation forever.
- 03Validation that enforces the criteria
Stage exit criteria enforced at the point of stage change. This is what converts your process from a document into a system, and it is the step most often deferred and never returned to.
- 04Users, teams, and sharing
Mirror how records are actually owned. Routing and reporting both inherit from this structure, so getting it wrong here propagates everywhere.
- 05Two-axis lead scoring skeleton
Fit and intent as separate scores rather than one blended number — a single score cannot tell you whether to nurture or to call.
Phase 3 — Migration and integrations (weeks 5–8)
The phase that overruns, every time, in every implementation. Three rules contain it.
- Set a cutoff and hold it. Records touched in the last 18–24 months migrate; everything older is archived outside the CRM. Importing a dead database makes it your problem again at a higher storage tier.
- Deduplicate before import, in a staging file. Merging at volume inside a live CRM is slow, lossy, and hard to reverse.
- Test every integration bidirectionally. For each connected system verify create, update, and delete behaviour in both directions. Delete behaviour is where the expensive surprises live, and nobody tests it.
A useful discipline: write down the record counts you expect after migration before you run it. Reconciling actual against expected catches silent truncation, which otherwise surfaces months later as a gap nobody can explain.
Phase 4 — Automation, reporting, adoption (weeks 8–12)
| Build | Must do | Common mistake |
|---|---|---|
| Routing | Assign by agreed rule, with a fallback and an SLA breach alert | No fallback, so unmatched records go nowhere silently |
| Lifecycle automation | Set stages from objective conditions only | Manual stage-setting, reintroducing the ambiguity you removed |
| Handoffs | Deliver as a task with a due date and context attached | An email notification nobody reads |
| Dashboards | Coverage, stage conversion, velocity, source performance | Twenty dashboards, none owned |
| Data quality jobs | Recurring dedupe, normalisation, enrichment refresh | Treated as a launch task rather than a standing job |
Adoption is a build task
Most implementation plans allocate a training session in the final week. That is not adoption work; it is a briefing. Adoption is the thing that determines whether the previous eleven weeks produced value, and it needs its own budget and its own sequence.
- 01Train by role, not by feature
An SDR and an account executive need different twenty-minute sessions. A single walkthrough of every feature teaches nobody what to do on Monday.
- 02Remove every field a rep cannot confidently answer
Do this before go-live. Each unanswerable required field is a daily reason to work around the system, and workarounds become permanent within a month.
- 03Run the first four pipeline reviews in the CRM
Leadership running the number from the system is worth more than any amount of training. If they use a spreadsheet, so will everyone else.
- 04Publish an exception report weekly for the first month
Records created outside the intended path, missing required fields, past-dated close dates. Visible, by team, with a named owner.
What to measure after go-live
- Required field completion at stage change — above 95%, or the fields or the training are wrong.
- Open deals with a close date in the past — under 5%, or the forecast is fiction.
- Records created outside the intended path — near zero, or a shadow process exists.
- Weekly active users against licences — above 90%, or adoption failed for a segment worth identifying.
Read these weekly for the first quarter. An implementation is finished when adoption holds, not when the configuration is complete — and the gap between those two moments is where most of the value is won or lost. Platform-specific detail is in HubSpot implementation and RevOps in Salesforce.
Want this diagnosed on your own numbers?
The RADAR™ Scan scores your revenue engine in 2 minutes — 12 questions, a 0–100 score, and your gate verdict. No email required.
Run your RADAR™ Scan→Questions this raises.
How long does a CRM implementation take?
What should you do first in a CRM implementation?
How much data should you migrate to a new CRM?
Why do CRM implementations fail?
How do you drive CRM adoption?
Related guides.
The sequence that makes a HubSpot build hold: decisions before configuration, lifecycle before automation, and the migration traps that cost months.
RevOps AgencyThe architecture decisions that determine whether a Salesforce org supports the revenue team in year three — or has to be rebuilt.
RevOpsThe work a RevOps agency actually delivers, what it costs in 2026, and the honest comparison against hiring in-house or going fractional.
RevOps AgencyFirst we build your pipeline. Then we build the machine that scales it.
Every engagement starts with the RADAR™ Reveal — a 2-week audit with a scored report, gate verdict, and roadmap. Yours to keep, whatever you do next.