B2B CRM Migration: A 6-Step Roadmap for RevOps
Moving systems risks broken pipeline reports, duplicate accounts, and missing consent records when teams rush cutover. This guide lays out a six-step migration roadmap, sandbox rehearsal rules, and a pre-flight scorecard to protect your sales data.
Meowket's take Row counts are not proof of a successful migration. Before celebrating go-live, check whether pipeline totals and account relationships match your old system dollar for dollar.
A B2B CRM migration succeeds when the new system shows the same customers, the same pipeline and the same consent history as the old one on day one. To get there, clean the data before it moves, map every field and consent flag, rehearse in a sandbox, and reconcile the numbers before you cut over. This guide sets out a six-step roadmap for Hong Kong commercial and RevOps leaders, with one worked example, owners and outputs for each step, and a pre-flight scorecard you can copy.
Key Takeaways
- Optifai, a CRM vendor, names skipped data cleansing and validation as a root cause of failed migrations. Rehearsals and reconciliation are the main defences.
- Load records in dependency order (users, accounts, contacts, deals, then activities and attachments) and keep each legacy record ID in a dedicated field so every record can be traced back.
- Migrate PDPO consent status, timestamps and opt-out flags with the contact records, and mute all automations during tests and cutover so real customers receive nothing.
What should you decide before a B2B CRM migration starts?
Before any data moves, decide the scope, the owner, the success measure and the stop rule. Reewamp’s guide for B2B software companies says such a migration typically takes two to eight weeks. The range depends on data volume, the number of integrations and how much of the sales process is redesigned along the way.

Reewamp also asks whether you need to move at all. If the problems are mostly bad data and missing process rules, it says a cleanup inside the current CRM may be cheaper. If you’re still choosing a platform, MarketingCat’s HubSpot vs Salesforce comparison for Hong Kong B2B covers that decision.
Cost lines to budget
The sources we reviewed give no reliable price benchmarks, so budget by line item instead. The main lines are:
- internal hours for cleansing and testing;
- partner or consultant fees for mapping and cutover;
- overlapping licences while both systems run;
- integration rework;
- lost selling time during the freeze and the first weeks after launch.
Reewamp says an in-house project works when a RevOps or business-systems person has time, admin experience on both platforms and engineering support. Its suggested middle path is a partner for design, mapping and cutover, with your internal owner running administration from day one.
One measure of success
Take a baseline from the old system before you start: open pipeline value and closed-won revenue by period. Make the primary KPI whether the new CRM matches those totals at go-live. Then judge adoption over a fixed post-launch window.
Agree in advance when to proceed, when to repeat a rehearsal and when to stop. If reconciliation fails after repeated rehearsals, pause the project rather than push the cutover date.
Illustrative example, not real client results: a Hong Kong B2B SaaS company is moving 80,000 CRM records from a legacy platform to a modern CRM across its regional sales operations. It prunes 28,000 inactive or duplicate records and migrates 52,000 clean ones. It tests with a 3,000-record sample batch, plans a 4-hour cutover freeze and runs a 4-week adoption check after launch. The steps below use this example.

Enterprise CRM Migration Roadmap
- Audit, cleanse, and prune dead-weight data: the RevOps lead produces a signed-off list of records to migrate, archive or delete.
- Document field mappings and compliance flags: RevOps and the data protection lead approve a mapping document covering every field, picklist and consent flag.
- Run sandbox rehearsals in dependency sequence: the integration engineer loads sample batches in object order and logs every rejected record.
- Reconcile pipelines, financials, and integrations: sales managers and finance sign a reconciliation report matching old and new totals.
- Execute freeze-window cutover with rollback plan: the project owner makes the go or no-go decision against written rollback criteria.
- Drive sales adoption and verify workflows: sales leadership reviews usage and workflow checks at the end of the adoption window.
If you’re still choosing between HubSpot and Salesforce, our CRM cost calculator compares their three-year cost for your team size.
Step 1: How much CRM data should you clean before migrating?
Clean before you move, and migrate only active, valuable records. Optifai’s migration checklist says a typical B2B CRM contains 30% to 40% “dead weight”, such as duplicates, inactive contacts and test data. That is a vendor’s description, not market data. Still, the 28,000 pruned records in the example (35%) sit inside that range.
Deduplication needs more than name matching. ProdRescue AI’s illustrative case describes how matching on company names created duplicate enterprise accounts. Subsidiaries, trading names and spelling variants broke the matches.
That case suggests matching on stable identifiers instead, such as ERP customer numbers or maintained mapping tables. Ambiguous matches should go to a human review queue rather than be merged automatically. This matters for Hong Kong firms whose clients trade under English, Chinese and regional subsidiary names.
Some teams will want to use generative AI to clean or enrich records. Before anyone pastes customer data into such a tool, get internal approval and follow the PCPD’s generative AI checklist from March 2025. Use permitted tools only, and limit and anonymise personal data. Check the vendor’s terms on retention and training, and keep people reviewing the outputs. MarketingCat’s enterprise AI adoption roadmap covers the governance side.

Step 2: How do you map custom fields, workflows and consent records?
Field mapping defines which target field each source value lands in and how it is transformed on the way. ShareCRM’s field-mapping guide says each mapping should record:
- the object and the API name;
- the data type and the transformation rule;
- null handling and validation;
- sample inputs and expected outputs;
- the owner and the approval status.
ShareCRM is a CRM vendor, but its examples are practical. An ID such as 00127 needs a text field to keep its leading zeros. Currency amounts need a defined currency and unit. A date such as 04/05/2026 must have an explicit input format.
That last point applies to any team whose old and new systems read dates in a different order. The sources we reviewed don’t cover how Chinese-character names survive export and import, so include them in your sample batch.
Reewamp applies the same discipline to workflows: every rule, sequence and workflow should be rebuilt or retired, never left undecided. Preserve the first-touch source and landing-page fields too. Marketing attribution depends on them, as MarketingCat’s guide to measuring B2B SEO ROI explains.
The PDPO consent check
Consent records are the field group Hong Kong teams can least afford to lose. Under section 34(2) of the Personal Data (Privacy) Ordinance, direct marketing covers offering or advertising goods, facilities or services. It includes mail, fax, email or similar communications addressed to a named person, and phone calls to specific persons. This is set out in the PCPD’s direct marketing guidance from October 2010.
That guidance addresses service application forms that bundle service terms with consent to marketing products or services unrelated to the service sought. It says such bundled consent raises doubts about whether consent was genuine. It advises separating agreement to the service terms from consent to that marketing. As a precaution, MarketingCat suggests keeping all marketing consent separate from service terms.
Under Data Protection Principle 1(1), only necessary, adequate but not excessive data should be collected. The guidance adds that Hong Kong Identity Card numbers are not normally needed for direct marketing, so don’t carry them over into marketing fields.
A March 2013 PCPD statement refers to new “notification” and “consent” requirements: data users must notify data subjects and obtain their consent to such use. The 2010 guidance predates those changes. Moving marketing lists into a new CRM without their consent records, opt-out flags and notice logs risks unlawful direct marketing and regulatory penalties.
Map consent status, timestamp, notice version and opt-out flags as named fields. The sources we reviewed don’t cover the Unsolicited Electronic Messages Ordinance, so check its sender duties separately. How these rules apply depends on your circumstances, and this is not legal advice. MarketingCat’s B2B lead generation guide covers outreach compliance in more detail.
Step 3: How should you run sandbox rehearsals?
Rehearse in a sandbox before touching production, loading objects in dependency order. Reewamp sets that order as users, companies, contacts and deals, then activities, notes and attachments. It also advises keeping each legacy record ID in a dedicated field, so every record can be traced back to the source.
In the example, the first rehearsal uses the 3,000-record sample batch. Include edge cases in it: every picklist value, broken lookups, boundary dates, long text and empty fields.
To make rehearsals reliable, use the same source snapshot and the same filters each time. Repeat the run after every mapping change, keep a log of what was tried and what failed, and check the output against what the business knows to be true.
Before any test load, disable all automated triggers, email alerts and outbound webhooks. ShareCRM warns against letting test notifications reach customers. An unmuted sequence could email real clients from sandbox data.
Step 4: How do you reconcile pipelines, financials and integrations?
Reconciliation proves the business data survived, not just the rows. Reewamp lists the checks:
- record counts per object;
- owners and relationships;
- activity history;
- integration behaviour;
- open pipeline value and closed-won revenue by period.
Reewamp advises repeating the rehearsal until the numbers match.
ProdRescue AI’s scenario shows why row counts alone mislead. Records can exist yet drop out of reports because they lost an account link or landed in a different stage. Categorise each discrepancy as a missing record, a broken relationship or a changed report definition before you fix it.
ShareCRM cites a 2023 Forrester survey to show that poor data quality is costly in general. It notes those estimates cover data quality overall, not migration errors.
For integrations, Reewamp says to inventory every system that reads or writes CRM data, including forms, enrichment tools, billing syncs and automation platforms. Accounting and billing links belong in that inventory. The sources we reviewed don’t cover how messaging-channel histories, such as WhatsApp conversations, transfer between CRMs, so test your chosen connector in the sandbox.
Step 5: How do you cut over without losing pipeline?
Cut over inside a short freeze window with a written rollback plan. Reewamp recommends making the old system read-only during the freeze. Automations that could fire on imported records should be switched off. Define in advance when and how you would return to the old system if day-one checks fail. The example plans a 4-hour freeze.
MarketingCat’s template: CRM Data Migration Pre-Flight Scorecard (a planning aid; owners are suggestions).
| Check before cutover | Suggested owner | Pass? |
|---|---|---|
| Source record IDs mapped to a dedicated custom field in the target CRM | Integration engineer | Yes / No |
| Loading order validated: users, accounts, contacts, deals, activities | Integration engineer | Yes / No |
| PDPO consent, timestamp and opt-out flags explicitly mapped | Data protection lead | Yes / No |
| Email sequences, triggers and outbound notifications muted for cutover | Marketing ops | Yes / No |
| Pipeline value and closed-won totals reconciled between systems | Finance and sales managers | Yes / No |
| Rollback criteria and escalation timeline documented for day-one failure | Project owner | Yes / No |
Any “No” means the cutover waits.
Step 6: How do you get the sales team to adopt the new CRM?
Adoption means reps work their deals in the new system, not in spreadsheets. Reewamp says to ask what support looks like after go-live and how teams’ use of the CRM will be checked. A clean import into a CRM nobody uses is still a failed project.
In the example, sales leadership reviews three things weekly during the 4-week check:
- whether reps log activities in the new system;
- whether routing and approvals fire correctly;
- whether dashboards still match finance’s figures.
Unmute automations one by one and watch each before switching on the next. If you’re rebuilding nurture flows, MarketingCat’s guide to automating a B2B sales funnel covers the design.
Limitations and common mistakes
This roadmap draws on vendor and consultancy guides, an illustrative engineering case and PCPD documents. MarketingCat has not tested these methods or tools. The PDPO guidance cited dates from October 2010, so check the current ordinance on Hong Kong e-Legislation and the PCPD’s latest guidance with your advisers.
The mistakes the sources keep returning to are:
- skipping cleansing;
- loading production without a trial run;
- matching accounts on names;
- leaving automations live during tests;
- treating a high import count as proof of success.
A CRM migration tests whether your revenue and consent data mean what you think they mean. Treat reconciliation as the finish line, not the import. This week, pull two numbers from your current CRM: open pipeline value and closed-won revenue by quarter. Also list every field that records marketing consent. Those become your baseline and your first mapping rows.
Frequently Asked Questions
Should we use a migration partner or do it in-house?
The answer depends on your internal capacity. Reewamp’s guide says in-house works when a RevOps or systems person has time, admin experience on both platforms and engineering access. It warns that migration work risks competing with daily operations and dragging on. A partner brings a tested process but needs proper documentation and handover. Reewamp suggests a middle path: a partner for design, mapping and cutover, with your internal owner administering the system from day one.
Do we need to migrate all historical activity records?
Not necessarily. Optifai, a CRM vendor, suggests migrating history selectively, for example activity logs from the last 12 to 24 months, plus win-loss records and key attachments. Decide by business need rather than habit. Keep consent and opt-out history regardless of age if it governs who you can market to. Archive the rest in a searchable export so nothing is lost, even if it doesn’t live in the new CRM.
Is a phased migration safer than a big-bang cutover?
Optifai, a CRM vendor, recommends a phased approach and estimates it takes six to 12 weeks with minimal downtime. It calls big bang risky and puts it at one to two weeks with eight to 24 hours of downtime. These are a vendor’s estimates, not independent benchmarks. Phasing by region or business unit can suit Hong Kong firms with regional teams, provided reconciliation runs at each stage.
What happens to WhatsApp or other messaging history when we switch CRMs?
The sources we reviewed don’t say. Whether chat history transfers depends on the connector or integration you use in each system, and that isn’t documented in our research. Treat it as an integration in your inventory. Ask the target vendor in writing what syncs and what doesn’t. Then test with real sample conversations in the sandbox before cutover, with outbound messaging switched off.
