Skip to content
Debbie
Back to Project Delivery
Portfolio case study

CRM Migration

A governed migration from fragmented records to a cloud CRM.

Role
Project Manager / Implementation Lead
Duration
14 weeks
Project context
Asteron Business Services — fictional case organisation

An end-to-end portfolio project developed to demonstrate project-delivery practice across planning, governance, implementation and handover. Asteron Business Services is a fictional case organisation.

At a glance

32/32
Final UAT scenarios passed
25h42
Production cutover

Within a 42-hour window

£123.6k
Final actual

Against a £128k envelope

52/58
Early weekly active users

Provisional, not mature adoption

01

The brief

Asteron Business Services—a fictional case organisation—had customer information split between a legacy CRM and controlled team spreadsheets. The project had 14 weeks to create a governed customer-data model, move essential history into a replacement cloud CRM and prepare 58 users without losing control of record identity or service continuity.

I structured the work around source inventory, profiling, controlled disposition, two migration trials, UAT, competence-based readiness, rehearsal, production reconciliation, hypercare and BAU transfer.

02

The starting point

The nominal CRM was only part of the operating system. Thirty-one users maintained parallel files, 96 custom fields had accumulated, and Customer Success held service obligations outside the original scope. The original 52% quality screen covered 20,780 core source records and is not directly comparable with the later governed loaded population.

4,860
Accounts
12,740
Contacts
3,180
Opportunities
48,600
Activities
96
legacy custom fields
31/58
users with shadow files
03

What I was responsible for

  • I established governance, the 14-week plan, tolerances, RAID and decision routes.
  • I led discovery and requirements, coordinated business, data, technical, security, reporting and change workstreams, and kept record-meaning decisions with business owners.
  • I controlled the £128,000 envelope and assessed changes across scope, schedule, cost, quality, risk and benefits.
  • I governed two trial migrations, reconciliation, UAT entry and exit, training readiness, rehearsal and production cutover.
  • I kept residual risks, quarantined records and immature benefits visible through hypercare and BAU handover.
  • Discovery
  • Data migration
  • Governance
  • UAT
  • Cutover
  • BAU handover
04

How I structured the project

  1. Weeks 0–2

    Discover

    Inventory sources, profile quality, map shadow work and validate assumptions.

    Stage 1 of 7

  2. Weeks 3–4

    Design and govern

    Baseline 39 requirements, data rules, retention, access and target processes.

    Stage 2 of 7

  3. Weeks 5–7

    Configure and Trial 1

    Build controls, test a relational cohort and stop unsafe survivor-chain mapping.

    Stage 3 of 7

  4. Weeks 8–9

    Remediate and Trial 2

    Resolve identity decisions, retest and prove a 50% representative migration.

    Stage 4 of 7

  5. Weeks 10–11

    Accept and prepare

    Run UAT, competence checks, operational readiness and a full rehearsal.

    Stage 5 of 7

  6. Week 12

    Cut over

    Freeze, migrate, reconcile, validate and release within the 42-hour window.

    Stage 6 of 7

  7. Weeks 13–14

    Stabilise and transfer

    Run hypercare, close finance and transfer benefits and residual actions to BAU.

    Stage 7 of 7
05

The hard part

The counts balanced, but the relationships were not safe.

Trial 1 reconciled to zero unexplained variance and still exposed eight defects. Twenty-three Contacts had historic Account merge chains that did not point deterministically to an approved survivor. A plausible automatic match would have made the spreadsheet look cleaner while creating an identity risk.

I stopped that route, quarantined all 23 and presented three options: accept an unsafe best match, move the milestone by three days, or fund controlled business review while independent preparation continued. The third option preserved the milestone without giving technical convenience authority over customer identity.

06

Key decisions

Decision 01

Bring service obligations into controlled scope

Context
Discovery found three registers holding 512 logical renewal and service obligations outside the CRM.
Options
Ignore them, import the overlapping rows as new entities, or model the obligation without duplicating customers.
Decision
I recommended target-standard capability plus a limited Service Obligation child record.
Why
It preserved operational commitments and customer identity without turning spreadsheet rows into duplicate Accounts.
Consequence
The Sponsor approved £6,400 from contingency after business, technical and privacy review.

Decision 02

Quarantine ambiguous survivor chains

Context
Trial 1 found 23 Contacts whose historic Account merge chains were not deterministic.
Options
Auto-match, delay three days, or fund controlled owner review alongside independent work.
Decision
I stopped automatic mapping and recommended controlled review capacity.
Why
Business owners—not a matching rule—held authority over customer identity.
Consequence
Fourteen mapped to approved survivors, three remained separate/new relationships and six stayed quarantined.

Decision 03

Defer predictive renewal scoring

Context
A late UAT request added five days of work and new governance questions.
Options
Add it before cutover, reduce assurance elsewhere, or move it into BAU discovery.
Decision
I recommended deferral to a 90-day BAU decision.
Why
It was valuable but not required for safe migration or the agreed release.
Consequence
UAT and cutover stayed on baseline with £0 project commitment for the request.

Decision 04

Continue after a contained cutover incident

Context
The BI service identity lacked read scope for one recreated view during production validation.
Options
Roll back, continue under fallback, or proceed without fixing the control.
Decision
I paused the checkpoint and recommended CONTINUE only after scope correction, retest and business review.
Why
Data remained reconciled, access was not excessive, a fallback report existed and no abort trigger had fired.
Consequence
The 37-minute incident variance was retained; cutover completed in 25h42 with all checks intact.
07

Project control

  • The RAID position closed with 16 risks, 11 issues, seven assumptions and eight dependencies controlled or transferred.
  • Six disposition routes—migrate, cleanse then migrate, merge, archive, exclude and decision required—gave every source record an authorised outcome.
  • The 96 legacy custom fields were rationalised into 30 retained fields, six consolidation targets and 20 standard replacements; 13 were archived and 19 retired.
  • Budget closed at £123,600: £116,000 base delivery plus £7,600 approved contingency actual, leaving £4,400 unused inside the £128,000 envelope.
08

Testing and assurance

Assurance separated accounting completeness from quality. Two trials, relationship checks, access tests, frozen UAT data, 32 unique business scenarios, 12 focused retests, role-based competence and a full rehearsal all had distinct gates.

  • Trial 1: zero unexplained variance, eight defects, gate passed with conditions.
  • Trial 2: 50% of each core entity, all 512 obligations and 100% relational integrity for loaded records.
  • UAT Cycle 1: 27 pass, four fail and one blocked; final position 32/32 after 12/12 focused retests.
  • Week 11 competence: 56/58, with two users access-restricted until evidence closed.
09

A rehearsed cutover with the evidence sequence intact

The full rehearsal took 26h10 and showed that reconciliation reporting was too slow. The team improved indexing and generated evidence in controlled parallel, bringing the production forecast to 25h05 without removing a check.

Production ran for 25h42 inside the 42-hour window. Accounts, Contacts, Opportunities, Activities and Service Obligations all reconciled across load, merge absorption, archive, exclusion and quarantine, with zero technical failures and zero unexplained variance.

10

What happened

Simulated actual · 42h approved window

25h42

Production cutover

Simulated final unique scenarios

32/32

Final UAT

Simulated actual · £128,000 envelope

£123,600

Final spend

Early indicator · not mature adoption

52/58 · 89.66%

Early weekly active users

Cutover result · different denominator from 52% source baseline

17,489/18,366 · 95.22%

Governed loaded-population quality

Simulated actual · on baseline

End Week 14

Project completion

11 · Evidence gallery

These are readable web-native representations of the source artefacts—not fake software screenshots. Open any item for a larger view and its source reference.

12

What I learned

Migration leadership is not about moving the largest possible volume. It is about making every source record explainable and preserving the authority to decide what a record means.

Reconciliation can pass while quality fails. Trials need relationship, access, behaviour and business checks as well as counts, and early adoption evidence needs its denominator and observation period attached.

Previous case AI Essay MarkerBack to DeliveryNext caseParentLink