First, group the records with no source by the form, import or system that created them
Start with the report the CRM fundamentals guide asks every marketing team to own: records missing required fields, grouped by the form, import or system that created them. Run it for original source over the period in question and read the groups rather than the total. A blank source is almost never spread evenly; it clusters under one form added without the hidden fields the others carry, one or two imports, or the integration user that created records from a sync, and the group names the cause before the table is opened.
Separate three symptoms while reading, because they have different owners. A blank value means nothing was captured at creation. A generic value such as import, website or sales means a value was written that identifies no campaign, which is a decision rather than a capture fault. A wrong value, such as a campaign that never ran in the lead's market or a most recent touch sitting in the original source field, means something captured a real value and something else changed it or read the wrong field. The fundamentals guide's design is the reference: original source records the channel and campaign of the first known touch, written once at creation and never overwritten, and most recent source records the latest touch, overwritten on each visit or submission. A report that reads the second as the first is a reporting fault, not a capture one.
The blank matters more than its share suggests. Built for this question, a KPI tree starts from won revenue and splits it into deals won, qualified leads, leads created and the spend that produced them, with the arithmetic between levels written beside each split, as the framework asks, so that when the top moves the tree says which branch to look in. Source is the dimension that lets every level be cut by campaign. Remove it and spend sits on one side of the tree with no leads to divide it by, the acquisition cost that the fundamentals guide says should be calculated on won deals rather than on form fills (CACCost to acquire one new customer.) cannot be cut by source, and first-touch and last-touch attributionAssigning conversion credit to touchpoints., which the two source fields support directly, has nothing to credit. The sourceless records are not missing from the tree; they are a branch nobody can read.
Two neighbouring diagnoses handle what this page does not. Sessions that analytics cannot assign to a channel, because their source and medium pairs match none of the channel rules, are the subject of the diagnosis for unassigned traffic; this page starts in the CRM, where a record exists with nothing to join to. And when the UTMURL tags identifying traffic source. values themselves drift between people, teams, platforms and agencies, the diagnosis for inconsistent campaign naming counts the variants and fixes the ownership behind them; this page meets that drift only where it stops a source being grouped or joined to spend.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Blanks cluster under one form, often one added or rebuilt recently | A test submission from a tagged landing page, then the record | A form added without the hidden parameter fields the other forms carry | Add the hidden fields and the capture script the other forms use |
| Web leads carry a source but no click identifier, or arrive blank when the visitor converted on a second page | The record of a test lead that landed with the parameters and converted elsewhere | Click identifiers and parameters not captured at the form, so the record arrives with nothing to join to | Persist the identifiers on arrival and pass them through the form |
| A block of records sharing a creation date and a batch tag read import or nothing | The import's source, campaign and lifecycle values as decided before the load | Lists loaded with no source value, or with a generic import value instead of the event or partner | Backfill with the real event or partner; make source a required import decision |
| Sources that cannot be grouped: several spellings of one campaign, a platform name in the medium, mixed case | The parameters on the links against the convention and the builder's log | Campaign parameters typed by hand or spelled differently, so the source cannot be grouped or joined to spend | Generate every link from the builder; map the strays once |
| Contacts whose source names a campaign that has no record, or a record with no members | The campaign record list against the campaign values on the links | No campaign record, or one named differently from the links, so members are never associated | Create one record per effort named exactly as the links; review unmatched values |
| Original source changes over time on the same record, or reset on a migration date | The record's field history for who wrote original source after creation | A sync or migration that let a later touch or the load write over original source | Make original source write-once; carry it in a field the load cannot set |
Fix, verify and prevent each cause
A form without the hidden fields
The fundamentals guide describes how source fields are populated from the website: campaign parameters are read from the landing page URL, held in a cookie or local storage for the session, and written into hidden form fields at submission, so they arrive on the contact with the form data, along with the landing page, the referrer and the form name, which cost nothing to store and explain the parameters when the naming turns out to be ambiguous. A form built by an agency from a template, a brochure download added by a different team, or a form moved to a new tool simply does not carry those fields, and every record it creates arrives without a source while every other form's records arrive with one.
Fix: add the hidden fields to the form, with the same capture script and the same field mapping the other forms use, and store the form name with every submission so the next gap can be grouped.
Verify: open a tagged landing page, submit the form, and read the record: source, medium, campaign, landing page and form name all present.
Prevent: keep a list of every form that creates CRM records with the fields each one writes, and make the tagged test submission part of the check every new or rebuilt form passes before it goes live.
Identifiers and parameters not captured at the form
A form can carry the hidden fields and still write nothing into them. The guide to connecting the CRM to ads and analytics names the two capture problems that account for most failures. Persistence: a person lands on one page with the identifier in the URL and converts on another without it, so the identifier has to be stored in first-party storage on arrival and read back at submission. And forms that do not pass hidden fields through: embedded, third-party and multi-step forms often strip unknown fields, and the fix is usually a form setting or a small script, verified by submitting a test lead and reading the record. The same guide describes the click identifier that Google Ads documents appending to landing URLs when auto-tagging is on, captured the same way and stored on the lead as the key for sending outcomes back; campaign parameters follow the same path and are the fallback when a click identifier is missing. A record with no identifier and no parameters is one the ad platform can never learn from and the report can never group.
Fix: store the identifiers and parameters on arrival, read them back at submission, and pass them through whichever form tool sits in front of the CRM.
Verify: click a real tracked link, land, move to a second page, submit, and confirm the record holds the identifier and the parameters from the first page.
Prevent: repeat that test after every change to the site's templates, forms, consent banner or form tool, and read the monthly share of web leads with no identifier as a capture alarm.
Lists loaded with no source, or with import as the source
The tutorial on importing contact lists makes source a decision taken before the load: every new record needs an original source that says where it really came from, such as the event or partner by name, never a generic import, together with the campaign the people join and their member status, so that the list can later be reported by attendance and pipeline. Its table of what goes wrong puts the symptom plainly: records get no source or campaign, and reports credit import or show a large blank segment. The lists that cause this are the ordinary ones, a conference's badge scans, a partner's registrants, a seller's spreadsheet, loaded by someone who had the file and not the decision.
Fix: find each sourceless batch by its tag or its creation date, backfill original source with the real event or partner, add the people to that effort's campaign record with the status that describes them, and set the import template so that source and campaign are required before a load.
Verify: the source report for the period shows the event or partner where it showed import, and a test load of a few rows without a source value is refused or sent to review.
Prevent: run every list through the tutorial's steps, with the source, campaign and lifecycle stageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer. values decided and written on the import before the file is cleaned.
Parameters typed by hand or spelled differently
A source that was captured but cannot be grouped is as useless to the report as a blank. The naming conventions guide explains why: values are read exactly as written, so a medium in two capitalisations is two values, a platform name in the medium field breaks the channel rule, and a campaign name typed freely cannot be split into market, objective, theme and period. Google's URL builders documentation describes adding campaign parameters to a link so analytics can read them from the landing page, as that guide and the campaign records guide state it; the convention around them is the work. The guide's rules are a short controlled list of medium values, a source that identifies the platform or the partner, lower case throughout, one separator inside a value and another between fields, and every link generated from one builder that logs what it made.
Fix: map the stray values already in the CRM to the convention once, in a mapping table that the source report reads, rather than editing records, and route every new link through the UTM link request procedure so nobody types a parameter by hand.
Verify: the source report groups each campaign under one value, and the builder's log holds every link used in the last campaign.
Prevent: add the monthly exceptions report the naming guide describes, listing mediums outside the list and campaign values that do not parse, and trace each one to its source. When the drift comes from agencies, platforms and teams each naming to their own pattern, the inconsistent campaign naming diagnosis above owns the fix.
No campaign record, or one named differently from the links
A source value is a label on a person; a campaign record is the CRM's account of an effort, with its members, their statuses, its cost and the deals that followed. The guide to campaign records in the CRM ties the two together by one rule: the record uses exactly the value that appears in the campaign parameter on the links, so a contact whose source reads a given campaign can be found as a member of the record with the same name. Members are added as a side effect of what people do, a form submission whose hidden field carries the campaign value, a registration from the event platform, a lead form connected through the integration, and a value that matches no record lands in a review list rather than creating a mistyped record or being dropped silently. Where no record exists, or the record is called one thing and the links another, the source is captured and the effort still has no members, no cost per member and no pipeline.
Fix: create one record per effort that someone will ask about, named from the taxonomy exactly as the links, with its status list agreed before it runs, and work the review list of unmatched values back to the records they belong to.
Verify: a test submission with the campaign value in its hidden field appears as a member of the matching record with the right status, and a submission with an unknown value appears in the review list.
Prevent: make the link and the record in the same request, by the same person, with the same name, as the campaign records guide suggests for the link request procedure, and never rename an old record for a new effort.
A sync or migration wrote over original source
Original source is written once and protected, and anything that writes it afterwards turns every customer into a customer of the last touch or of the migration. The tutorial on moving to a new CRM describes the loss exactly: fields that should be written once are written again by the import, so every record appears to have arrived on migration day from an import source, and its answer is to move original source and the parameters behind it as written, never set by the load, carried in clearly named fields of their own if the new system will not accept historical values on its own fields. The syncing guide's field ownership table says the same for the day-to-day case: original source flows from the marketing platform to the CRM on creation only and is never overwritten by a later sync, while most recent source is the field built to change.
Fix: read the field history for who wrote original source after creation, make the field write-once in the owning system, repair the sync or the load that wrote it, and restore the original values from the pre-migration export or the stored landing page and parameters where they survive.
Verify: a seeded contact with a known original source keeps it after a second visit from a different campaign, after the next sync and after any reload.
Prevent: keep original source in the field ownership table as written-once with one owner, and when it still reverts, the diagnosis for CRM values changing back after a sync reads the record's history to name the second writer.
Prevention: source captured at every entry point and a monthly blank-source report by creating form
Every cause above is an entry point that creates records without the fact that explains them. The prevention is a rule for each entry point, an owner for each rule, and one report that shows which rule failed. The table below is the fundamentals guide's design written as a checklist.
| Entry point | Source rule | What goes to review | Owner |
|---|---|---|---|
| Web forms the business controls | Hidden fields filled from the landing page; landing page, referrer and form name stored | Nothing; a form that cannot capture is fixed before launch | Whoever owns the forms |
| Embedded, third-party and multi-step forms | Identifiers persisted on arrival and passed through; a test lead read after every change | Web leads with no identifier and no parameters | Whoever owns the site |
| Ad platform lead forms and integrations | The campaign value mapped to the source and campaign fields at creation | Values matching no campaign record | Whoever owns the integration |
| List imports | Source names the event or partner; campaign and status decided before the load | A file arriving without a source decision | Whoever runs the import |
| Manual creation by sales | A source picklist filled at creation for offline leads, with the event or partner named | Records created with the source left blank | Sales operations |
| The sync and any migration | Original source written once, on creation, by one owner; never set by a load | Records whose original source changed after creation | Whoever owns the sync |
Then run the report. Once a month, list the records created in the period with no original source, with a generic source, or with an original source that changed after creation, grouped by creating form, import batch or system user, and route each group to the owner in the table. The grouping is the whole point: a total tells the team it has a problem, a group tells it which door is open. The data quality guide runs the same discipline for event data in a weekly routine that takes well under an hour, and the two reviews are best scheduled together, because a campaign missing from the source report and a campaign missing from the channel report usually share a cause.
Seed the entry points so the report has something to catch. Keep one tracked link per platform and one test submission per form in the check every site release, form change, consent banner change or integration change has to pass, and read the record each time rather than the form's confirmation page. Keep the mapping table between the ad platforms' campaign names and the campaign values in the CRM short, and keep it shrinking, by naming new campaigns from the convention so the join between spend and outcomes needs no translation. Keep the UTM entry and the naming guide's vocabulary lists where the people who build links can see them. And keep the campaign record list tidy, with finished records marked as finished so the pick lists stay short and nobody chooses last year's event for this year's leads. A rebuilt or moved site can change forms and capture scripts together, and the tutorial on moving a website without breaking the CRM sets the order of those changes and the tests that prove them.
Finally, protect what was captured. Original source is write-once with one owner; most recent source is the field that moves; and a report that needs the latest touch reads the second, never the first. When several of these habits are missing at once, the CRM section places the guides that build each of them side by side, and the troubleshooting hub separates record, workflow and sync faults when the symptom is not yet clear enough to say which this is. A person held twice, with the source on one record and the deal on the other, is the duplicate contacts diagnosis's subject and shows up in this report as a blank that a merge would have filled.
A commercial laundry service finds its missing sources (illustrative)
Zundrapple Laundry Services
Zundrapple Laundry Services, an invented commercial laundry serving hotels and clinics, is asked by its finance director which campaigns produced this year's 212 new contracts. The CRM holds 5,840 contacts created in the year; 2,310 have no original source and 1,460 read import. The marketing lead runs the report of blank and generic sources grouped by creating form and finds four groups. 1,740 blanks came from a brochure-request form the design agency built for the clinic sector, without the hidden fields the hotel form carries. 410 blanks came from the hotel form itself but only where the visitor landed on a pricing page and converted on a second page, because the capture script was never added to the pricing template. The 1,460 import records came from two trade-show lists and a linen supplier's referral list, loaded with import as their source. The remaining 160 blanks had once held a value: a sync change in March let the marketing platform's latest-touch value write over original source, and for contacts whose latest touch carried no parameters the field was emptied.
The laundry adds the hidden fields and the capture script to the clinic form and the pricing template, backfills the three lists with the event and partner names and adds the people to those efforts' campaign records, and makes original source write-once in the CRM with the sync restricted to creation. The 160 overwritten values are restored from the stored landing page and parameters on 131 records and set to an explicit unknown on 29. Rerun, the report shows 44 blank sources out of 5,840, all from leads sales created by phone without choosing a source, and a required source picklist on manual creation closes that door. The paid search campaign that had appeared to produce 38 clinic contracts turns out to have produced 71, and the monthly blank-source report becomes a standing item in the marketing lead's first week of each month.
Common mistakes
- Reading the total instead of the groups. A share of blank sources is a worry; a group under one form is a fix.
- Letting a person fill the blank. A salesperson's guess becomes the data and the report credits word of mouth.
- Trusting the form's confirmation page. The page says the form worked; only the record says what was captured.
- Loading a list because the file was ready. Source, campaign and status are decided before the load, not backfilled after the report.
- Fixing the spellings in the records. Map the strays once and generate every new link from the builder; edited records drift again by next month.
- Naming the campaign record differently from the links. The source is captured and the effort still has no members.
- Letting the sync write original source. Every customer becomes a customer of the last email they clicked.
- Treating a blank source as a reporting problem. It is a capture gap at an entry point, and the entry point is where it is fixed.
