What a website move can break in the CRM and its workflows
The CRM fundamentals guide observes that a gap in what the CRMThe system that stores an organisation's contacts, companies, deals and interactions, and the processes built around it. reports almost always starts as a gap in what some entry point captured, and the website is one of those entry points. This page starts when that entry point is replaced while the CRM is live. Its records already have histories, and its workflows, merge queue, scores and alerts already read them, so every form has to keep writing the same kind of record from the new pages and everything that reads those records has to stay pointed at what the new site really does.
| Part of the site | What the CRM or a workflow takes from it | How a move can break it |
|---|---|---|
| Each form that writes records | The form's fields, the CRM property each fills and the fields required at creation | A new tool or template maps a field differently, drops one, or writes free text where the old form wrote a list value |
| Source and identifier capture | Hidden fields for campaign parameters, click identifiers, landing page, referrer and form name, filled by a script that stores them on arrival | The new form carries none, strips them, or the script does not run on the new templates |
| Matching | Each form's matching rule, where the CRM supports one, set to match on email and update the record it finds | The new form creates a record on every submission, so a returning person becomes a new one |
| Consent | The consent status and date for each channel, with the source, wording and choices behind them | A new form or banner records a choice with no wording or time, or loses a confirmation step |
| Triggers, conditions and routing | The event, form name, field value or lead type a workflow or rule starts from | The form is renamed, writes a value the rule does not list, or the record arrives by a route that never raises the trigger |
| Page-based signals | Scores and alerts that count a visit to a page such as pricing or a demo page, tied to a known contact | The page moves or merges and the signal's inputs may stop arriving, which can look like a quiet market |
| Links inside messages | Links in nurture, acknowledgement, quote, onboarding and alert messages, and the tagged links that campaign records are named after | Links reach addresses that redirect without their parameters, or that no longer exist |
| Pages the CRM's messages depend on | A preference, opt-out or confirmation page that lives on the site | The page moves while delivered messages still carry its old address |
| Reporting | The reports that group records by creating form and by source | Nothing was recorded beforehand to compare with, or a definition was changed during the launch |
Other pages own the rest of a move. Redirects and the order of launch day sit with the migrations guide for SEO and the website migration checklist, the Google Ads and Meta Ads pages cover their accounts, and the tag and the conversion events are tested through the tracking pages named in step nine, and the consent banner is tested with the checklist named in step six. The rest of this page covers only the CRM and its automation. If the CRM itself is being replaced, read the tutorial on moving a CRM; if an email platform changes in the same project, the email platform migration tutorial has its own plan.
Nothing here goes beyond what the site's guides already state about CRMs, form tools and marketing platforms, apart from the general practice named below. What a particular tool matches on, passes through, stores or fires is in that tool's help pages, and the steps point there where it matters. The register, the baseline, the freeze, the field map, the fixed order of change, changing each link in the window its page goes live, editing a workflow's trigger instead of copying the workflow, the recoverable old route, an old address kept answering for delivered messages and reading the first weekly readings against the baseline are general practice rather than tool rules. The table after the steps routes each fault that has a diagnosis page to that page.
Before you start
- Admin access to the CRM for whoever does the work, covering workflows, routing rules, scores and alerts, and the history of a record that shows who wrote a field and when.
- Access to each form tool and the tag manager, or the developer who holds them, on the old site and the new one.
- The site project's plan for the move: when it launches, which hostnames and URL patterns are replaced, which templates, form tools, booking tools, consent banners and microsites are new, who owns the redirects and what reversing the launch would take. This page supplies the CRM inputs to that plan.
- The automation inventory, if there is one, with each workflow's owner and purpose as the automation governance guide describes. Without one, the first step starts it for the website's share.
- The standard lists as they stand today: the countries, territories, lead types, job roles and stage definitions that routing, scoring and reports read. The move leaves them untouched.
- Marked test records, with invented details, a marker in the name and the email domain, and exclusions from reports, audiences and uploads in place before the first test, as the test environments and test data guide describes.
- A named person in sales operations who can say what each routing rule and alert expects, and a developer who can say what each new form sends.
The steps
List every form, page and script the CRM depends on in one register
Build one register with a row for every place that feeds the CRM or reads what the site passes it. For each row note its owner, where it lives, what it writes or expects, what it is for, what replaces it and when it last changed. Work from the CRM outward. Begin with the grouping that the fundamentals guide's missing-fields report and the duplicates diagnosis share: the records the CRM holds, grouped by the form, import or integration that created them. The grouping brings up forms that write records, including ones nobody remembers.
For each form note where it lives, its fields and the property each fills, the required fields, the hidden fields it fills and the values it sends as source, its matching setting, its consent fields and wording, and its owner. The lead routing rules template asks, under its source and form section, for the hidden fields and source values each form passes. Next list every workflow, routing rule, score and alert, marking those that start from a form, a form's name, a field a form writes or a page address, with the value each expects; the fields that routing, scoring and reports read are the ones the field standardising tutorial names. Then list the messages and the links in them, the campaign records, and any preference, opt-out or confirmation page on the site. Last, list the scripts: what reads parameters and identifiers from the address, and any tracking script from a marketing platform that ties visits to known contacts.
Build the register from the CRM and not from a crawl of the site: forms that write records can sit on campaign landing pages that a pattern-based redirect map misses, which is the exception the website migration checklist already flags for landing pages with external links. Note the reason for each row as you add it, and leave alone anything whose purpose no one can state until someone can; the Chesterton's fence framework states the rule that governs the clean-up in step ten.
Record the baseline from the CRM before anything changes
With the register showing what the CRM depends on, this step captures how leads flow while everything still works, so that a count after launch has an earlier level to be set against. Over the periods the CRM's own reports already use, and by creating form, pull the records created, the records missing required fields, the share with a source and with a click identifier, and the merge-queue pairs grouped by the entry point that made the newer record, which is the grouping the duplicates diagnosis uses. For each workflow that starts from a form, take its enrolment count over the same periods, the first reading the workflow diagnosis asks for. For routing, take the records with no owner and how long a record waits before a seller's first activity, by owner, the second of which is a report the fundamentals guide expects marketing to be able to pull. For scores and alerts, take how many leads crossed the hand-off line, the weekly reading in the lead scoring guide, and what followed each alert, which is how the alerts guide judges them.
Treat each figure as a range within the CRM's own history, not a single reading; the guide to reading small numbers covers why small counts can swing by chance alone. Date the readings and keep them with the register.
Freeze everything that is not the move
Leave as they are, from the day the first change is made until the readings settle, the parts of the CRM that the move does not need to touch: field names, list values, stage definitions, routing rules, scoring weights and thresholds, workflow criteria and campaign names. A shift in the readings then points at the move and not at a bundle of changes made together; leaving the rest alone is general practice, and the change log the governance guide asks for is what lets a later drop be traced to a change. A redesign can arrive with a changed form, new questions or a new offer, and each of those changes what the CRM receives. Take those decisions separately, date them separately and keep them off launch day.
Write the field map for each new form before it is built
For every form on the register, write one line per new field naming the CRM property it fills, and settle the values before the template is built. Where the property is a picklistA CRM field type that accepts one value from a defined list of permitted values, so that records can be reported on instead of typed in many spellings., give the form field the same list with the same values, so that the form writes what routing, scoring and reports expect. The standardising tutorial's rule is to standardise at the form, ahead of the workflow, and the workflow diagnosis follows an unassigned lead back to a rebuilt form that wrote free text where its predecessor wrote a list value, a change nobody had checked against the workflows reading the field. Mark which fields are required at creation; for a contact the fundamentals guide gives email, source and owner as the usual three. If the new design drops a field, treat that as a decision and look in the register for whatever reads it first.
Hand the developers the map and ask them to check each template against the register's workflow and routing rows as well as against the design. Where a new tool cannot write a property the way the old one did, record the mapping in the register and route unmatched values to a review queue under a fallback owner, so that no lead is stranded by a new spelling.
Rebuild the capture: hidden fields, the script and persistence
Each new form carries the same hidden fieldsA form field the visitor never sees that records values such as source, campaign and landing page, so each lead arrives in the CRM with its origin. as the others, the set that the capture table in the CRM integration guide lists. Store the form's name with every submission so that a later gap can be traced to the form that made it. On arrival a script reads the parameters and identifiers from the address, holds them in first-party storage and reads them back at submission. Where auto-tagging is on, Google Ads adds its click identifier to the landing address, as the integration guide records from the Google Ads help centre, and campaign parameters are what Google's URL builder documentation describes adding to a link.
The integration guide names the two capture faults behind most failures. One is persistence, where someone lands on a page that carries the identifier and converts on another page that does not. The other is a form that discards hidden fields, and the same guide gives the remedy and the test that confirms it. A new template or form tool can stop the script running or the fields passing, and whether a given tool keeps unknown fields is a question for that tool's own help pages. Check every new form and landing template on the staging copyA private copy of a website or tool where changes are tested before release, kept behind access control and out of search and production analytics. for the fields and the script. As general practice, addresses on staging differ from live ones, so the identifier through a real tracked link is proved after launch, in step nine, and a tagged link is sent through any redirect as well, since one that fails to carry the parameters removes what the capture reads.
Set matching and consent on each form
Where the CRM supports matching rules, every form that writes into it, embedded and lead forms included, should match on email and then update the record it finds, with creating a new record the deliberate exception, so a returning contact's record is updated and not duplicated. The fundamentals guide sets that rule, and it is the fix for the first cause in the duplicates diagnosis, which adds that a form rebuilt or moved to a new tool should have its mapping checked, not assumed. For companies, the matching guide explains why personal addresses and shared domains wait for review instead of being guessed, and the new forms should feed the same rule. DeduplicationRemoving double counts when one conversion is reported by more than one source. starts at the entry point.
Then settle consent for each form. The CRM holds a consent status and date for each channel, and the consent recordThe stored evidence of a marketing permission: the channel, the wording the person saw, the date and the source of each opt-in, kept with the contact record. behind them, as the diagnosis of subscribers who say they never signed up describes it, keeps the source, the time, the wording shown, the choices made and any confirmation. A rebuilt form that changes the wording or the confirmation step changes that record. Keep the wording as it was unless changing it is a deliberate decision, and record any new wording from the day it goes live; that diagnosis recommends the same after a change to a form or platform. The GDPR, as that diagnosis summarises it, asks that consent be demonstrable, and the ICO publishes guidance on direct marketing in the UK; this page is not legal advice, and whether a given wording suffices is a question for the person who owns data protection. Where a marketing platform holds the same people, consent crosses the sync as the syncing guide describes, so add the new consent field to the field ownership table instead of writing around it. The banner is tested separately with the consent banner QA checklist.
Re-point the workflows, routing rules, scores and alerts that start from the site
Go through each workflow row on the register. The workflow diagnosis explains that a workflow that starts from a form submission listens for that event, so ask whether the new form still raises it. Where the old trigger named a form, that diagnosis points to a change that every path makes instead, for instance the standardised field being set, so decide what should happen to records that arrive by import or sync and so raise no form event at all. Where the tool allows it, edit the existing workflow's trigger rather than copying the workflow for the new form; the governance guide calls copies the commonest source of orphans. The standardising tutorial has routingRules that assign each new lead to an owner, queue or next step based on territory, segment, score or source. read the clean field, so confirm that every value the new form writes sits in both the standard list and the rule's own list, since the wrong salesperson diagnosis counts a value spelled differently from the rule's own list among its causes, and check the fallback owner for what no rule matches.
Then the signals that count visits. The scoring guide asks for a monthly check that web tracking still fires on the pricing page and says scoring fails quietly when its inputs do. The sales alerts guide builds alerts on events such as a return to the pricing page; those rely on identifiers tied to a known contact, and the identifiers are held back until the visitor permits them where consent applies, which the privacy and consent guide traces to Google's consent mode documentation. For each page-based signal, write down the page it names and that page's address after the move, and, where the tool allows it, change the signal in the window the page goes live. How a given platform matches a page is a question for its own help pages.
Update the addresses inside messages and records, in the window the page goes live
Nurture, acknowledgement, quote follow-up, onboarding and alert messages link to pages on the site, and the campaign records carry names that must match the tagged links which brought people to them. Change each link in the window when the page it points to goes live: a link changed before that page is live leads nowhere, and one changed after it keeps travelling through a redirect. The redirect map covers messages already delivered and links already published; messages the business controls should carry the new addresses themselves. Whether a tool lets a sent message be edited is not something these guides say, so the plan should not rely on it.
Reissue tagged links with the UTM link request procedure, so values stay in step with the campaign records, and use the page URL change SOP for one address, since it lists email templates and automated journeys among the places to update. If a preference, opt-out or confirmation page lives on the site, then as general practice ask the owner of the redirect mapThe table of every old URL and its new destination, one row per address, from which a site migration or URL change is built and then checked. to keep its old address answering for as long as delivered messages carry it.
Test each form through to the CRM record, before launch and after
Run the tests with invented, marked records and the exclusions already in place. For every form on the register, note what the record shows, not what the screen says.
- A tagged link, a second page, then the form. The record holds the source, medium, campaign, landing page, form name and the identifier from the first page.
- Each consent choice. The record holds the choice, the wording and the time, and the confirmation step runs where one is used.
- Each workflow that should start. It enrols the record once, and no copy of it enrols the record a second time.
- A visit to each page a score or alert names, with consent accepted, which shows whether the signal reaches the contact.
The rest of the capture, routing and follow-up tests are the ones in the Capture, Routing and Follow-up groups of the lead routing QA checklist; run them for each form with the same marked records.
On staging, check first where each form writes. The test environments guide treats a staging site as no test environment when its tags reach live analytics or live ad accounts, and as general practice a form that writes into the live CRM is the same case, so marked records and exclusions apply there too. After launch, run each form on the live site with consent accepted and again with it refused. This step stops at the CRM: the tag, the conversion events and a count of one per submission belong to the tracking pages. The tracking verification checklist covers the tags, conversion paths after a release walks each path on the devices visitors use, and the form tracking tutorial sets the success event that should fire once the server accepts a submission. When the passes are done, remove or mark every test record and confirm that none reached a report, an audience or an upload to an ad platform.
Read the first weekly readings against the baseline, then retire the old route and write it down
At the interval the baseline used, set each form's own count of submissions beside the records the CRM created for it, the comparison of attempts against records received that the diagnosis of submissions not arriving makes, then set the baseline's reports beside the new ones: records created, with a source and with an identifier by creating form, merge-queue pairs by entry point, records with no owner, workflow enrolment counts, the leads crossing the hand-off line, what followed each alert, and the other readings taken in step two. Read steady levels as matching the baseline, a sudden fall on one form as a possible break on that form, and a count that runs ahead of the form's submissions as a prompt to look for a second record or a second enrolment. Keep the new pages' conversion rate out of this reading: it shows how many visitors act on the new templates, not how the CRM handles them, and the last row of the table after the steps sends that question to its own diagnosis.
Keep the old forms, scripts and routes recoverable until the tests have passed and the weekly readings are clean. The framework one-way and two-way doors frames the question as how cheaply a decision can be reversed, and where the tool allows it a form that is switched off can be switched on again, while a merge is irreversible in most systems, as the fundamentals guide notes. Then retire what the move left behind: old form integrations, scripts, triggers and any copied workflow, touching each item only after its purpose is understood, as the fence rule in the first step asks. Record the move in the change logA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced., in the way that keeping a measurement change log describes, so that a later fault can be placed against it. File the register beside the integration register that the documenting marketing integrations tutorial builds: the register is then the CRM's standing record of what depends on the site, and a later move can begin from it.
What goes wrong after launch, and which page diagnoses it
Several faults can appear together after a move. The table routes each symptom it lists to its diagnosis, and running the marked tests from step nine again separates a form fault from a rule fault.
| What you see after launch | Check first | Page that diagnoses it |
|---|---|---|
| Records from the new form arrive with no source or identifier | A test from a tagged link, through a second page, to the record | CRM lead source blank or wrong |
| A returning person has a second record, created by one form | A test submission from an address the CRM already holds | Duplicate contacts in the CRM |
| Records from the new form never enter a workflow, or fall to the fallback owner | The value a missed record actually holds, set beside the one the rule expects | CRM workflow not running and leads assigned to the wrong salesperson |
| The form shows success but nothing arrives | Submission attempts against the records received, then the notification and the responder's view | Website form submissions not arriving |
| Records arrive with no confirmation time or wording, or people say they never signed up | The consent record of one new record, and where the form posts | Subscribers saying they never signed up |
| Forms and records read correctly but fewer visitors act on the new pages | Each path, form and page on the new site, not the CRM | Conversion rate dropped after a redesign |
A grain equipment maker moves its site (illustrative)
Bragorn Grain Systems changes platform, form tool and consent banner
Bragorn Grain Systems, an invented maker of grain drying and storage equipment for arable farms and farm contractors, replaces its website with a new platform, new addresses, a new form tool and a new consent banner. The web team believes four forms feed the CRM: a quote request, a brochure download, a dealer enquiry and a service call-out. The sales operations lead builds the register in the first week, starting from the records the CRM holds grouped by creating form, and finds six forms, because a spring open-day registration page and a partner page's brochure form were added by earlier campaigns. The register ends with 23 rows, not counting its capture script and campaign records: 6 forms, 5 workflows and rules, 2 page-based signals, 7 message templates and 3 pages that messages point at. In the weekly readings before the move the CRM received between 12 and 16 quote requests, between 28 and 34 brochure downloads, between 7 and 11 service call-outs and between 2 and 4 dealer enquiries a week, nearly all with a source, with 1 to 3 pairs a week in the merge queue and none waiting for an owner.
| Where | What the register found | What the team changed |
|---|---|---|
| Quote request form | The new tool writes the county as typed text, while the routing rule reads a county list with the sales regions behind it | Made county the same list on the form, with the same values, and sent unmatched values to the review queue with the fallback owner |
| Brochure download form | The new form carries none of the hidden fields, and the nurture starts from the old form's name | Added the hidden fields and the script that fills them, and re-pointed the nurture's trigger from the form's name to the field the form writes |
| Every form | The new tool creates a record on every submission | Set the forms to look the person up by email first and update the record found |
| Open-day registration page | Known only from the creating-form grouping, and dropped from the new site's plan | Kept the page's address answering and its form posting until the event follow-up was done and the first weekly readings were clean, then retired both |
| Consent wording | The new banner and the brochure checkbox use shorter wording | Kept the old wording, and wrote the wording version and its date beside each form's consent field |
| Pricing page signal | The score and the return-to-pricing alert counted the old pricing path | Pointed both at the new path in the launch window and noted the date, so that fewer alerts in the first days read as a counting gap |
| Messages and pages | Seven templates linked to old addresses, and the dealer confirmation page moved | Changed every link in the launch window, reissued the tagged links and kept the old confirmation address answering |
On staging the fields and the script pass. After launch the marked tests find one more fault: a marked quote request from an address the CRM already holds creates a second record, because one form still has the tool's default, which the team changes before the first real requests arrive. In the first two weekly readings after launch the CRM receives 13 and 15 quote requests, 29 and 33 brochure downloads and 9 and 8 service call-outs, inside the baseline ranges, with 2 and 3 pairs in the merge queue. The records with no source, grouped by creating form, show all 7 dealer enquiries since launch with none: the dealer form sits on a subdomain where the capture script never ran, and its marked test had been run from a main-site page. The team adds the script, tests the form again and records the move in the change log.
The CRM for marketers hub lists the CRM guides this tutorial builds on.
Common mistakes
- Listing forms from a crawl of the site. Forms that write records can sit on pages that a crawl misses, so list them from the CRM's records grouped by creating form.
- Testing the form and not the record. A success message proves the form; only the record shows the source, the match, the consent and the owner.
- Testing only a new person. A form can create a record correctly for a stranger and duplicate a returning contact, so test an address the CRM already holds.
- Leaving a copy of a workflow behind. Copies are the commonest source of orphans, as the governance guide says, so a copy made for the new form is retired once the move is proved.
- Reading a quiet score or alert as a quiet market. A signal whose inputs stop arriving fails quietly, as the scoring guide says, and the fall can look like a change in demand.
- Letting the consent wording change with the template. Wording that changes with the redesign and is not logged can leave records that cannot say what the person saw.
- Changing the form and the offer on the same day. A changed form or new questions change what the CRM receives, so they cannot be read as the move.
- Retiring the old route before the new one is proved. Where the tool allows it a form can be switched on again; a merge is irreversible in most systems.
- Stopping work on launch day. The first weekly readings by creating form can be where a missing script is found, and they need an owner until the reading is clean.
