First, find where a submission stops: attempts against records received, the notification, the record system and the responder's view
The report comes from outside the business: a customer writes that they sent the form and heard nothing, or the owner learns of an enquiry because the person rang to ask. Neither says where the submission stopped, and a fix aimed at a guessed cause can leave the real one in place. As general practice, reply to the person through the details they gave when they wrote in, record the request by hand where the team works, and do not ask them to submit again until the path has been checked, because a second submission along the same path may be lost in the same way.
Then make the stage visible with one marked test submission. Follow the conventions in the martech test environments and test data guide: invented details, a marker that a person or a filter can recognise, report exclusions for test records in place before the submission is sent, and removal afterwards. Send it through the live form and write down the time to the minute, because each later check searches for that moment.
Now read four places in the order a submission travels, and note every place where the test is missing.
- Attempts against records received. Set the submit attempts recorded for the form, as the diagnosis of abandoned forms sets up its events, beside the entries that arrived in the inbox or the record system for the same closed days, and look for the test. A success message on screen with nothing behind it puts the loss at the form or at a later place, which the three places below separate. As general practice, where a tool keeps its own record of submissions, that record is a further point of comparison, though nothing below depends on it.
- The notification. If the form tells the business by message, ask whether the message left the site, was turned away by the receiving side or landed in junk, reading the sending record and the junk folder of the mailbox it goes to.
- The record system. Open the CRMThe system that stores an organisation's contacts, companies, deals and interactions, and the processes built around it. or the order or booking system that holds enquiries, whichever the business regards as its source of truthThe system whose figure is treated as authoritative for a given number, such as billing for revenue or the CRM for pipeline, with other tools reconciled to it. for them, and look for the test record by its marker, then read the error queue, the counts and a sample of records.
- The responder's view. Find the test record under its owner or queue, in the view the person who has to respond opens, and ask that person whether they can see it.
If the entries are present in the inbox and the record system and only the analytics count is short, the submissions arrived and the counting is the fault, which the diagnosis of form submissions not tracking in analytics covers. Otherwise each place where the test is missing points at a cause, and the table sets each one beside its check.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| On screen the visitor sees a success message, yet no entry turns up anywhere; recorded submit attempts outnumber the entries that arrived | The recorded submit attempts against the entries received, whether a confirmation reached the visitor on screen and by message, and whether the notification, the record system and the responder's view show a fault | A form that fails behind the visitor | Follow the fix in the abandoned-forms diagnosis |
| No notification in the mailbox or its junk folder, though the form reports success | The sending record for the notification, and who reads the mailbox it goes to | A notification that is refused, filed as junk or sent to a mailbox nobody reads | Authenticate the sending system; send the notification to a mailbox someone reads |
| Notifications arrive but the record system holds nothing new, or one old record with nothing added | The error queue, the counts on each side, and who authorised the connection | A connection that stopped, or a record the system refuses or folds into an existing one | Apply the matching fix in the sync diagnosis; record missed requests by hand |
| The record exists under an owner or queue, yet the person who must respond says they did not see it | The test record's owner, its queue and the view the responder opens | A record that sits unassigned or out of the responder's sight | Give it an owner and a fallback queue |
| Submissions stopped around a redesign, a change of form tool or a change to the form | The destination listed for the form against where a test submission lands now | A change that moved where submissions go | Restore or re-point the destination, then repeat the test |
Fix, verify and prevent each cause
A failure inside the form, behind the visitor
The check that separates this cause is the first comparison: submit attempts against records received, and a marked test that shows success on screen with nothing behind it while the notification, the record system and the responder's view show no fault of their own. Entries lost behind the visitor look like drop-offs, which is why a form abandonment rateThe share of people who start filling in a form, by entering a first field, but leave without submitting it. can read worse than the form deserves. The diagnosis of abandoned forms owns this cause, with its example of a filter that blocked company addresses it did not recognise; it starts from visitors who do not send the form, and this page starts after they have.
Fix: follow the fix in that diagnosis for the fault the check finds. Where spam protection is needed, the forms guide names a hidden field, a check made on the server, or a provider offering an accessible alternative as the right choice, and the confirmation pages guide covers what the visitor should be told.
Verify: a fresh marked test arrives in the inbox and the record system, and the visitor sees a confirmation.
Prevent: keep the test on the schedule set out in the prevention routine below.
A notification that is refused, filed as junk or sent to a mailbox nobody reads
When the form tells the business by message, the check is the one the diagnosis of order and account emails not arriving puts to a customer's missing receipt: read the sending record to see whether the message went out or was refused, and look in the junk folder of the mailbox it goes to for one that was accepted. That page is written for transactional emailAn email triggered by a person's own action or account, such as a receipt, password reset or shipping notice, rather than a marketing send. sent to customers, so applying it to a form's notification to the business's own mailbox is general practice, by the same mechanism. It cites Google's email sender guidelines and the DMARC overview for authentication, and the DMARC failure diagnosis covers a sender that uses the domain in its From address without authentication. The third possibility is a mailbox that receives the message and is not read; the example in the landing page checklist sends a test submission to an inbox nobody had opened since a staff change.
Fix: authenticate the system that sends the notification, as the two diagnoses describe, and point the notification at a mailbox a named person reads.
Verify: a fresh marked test produces a notification in that mailbox, and the sending record shows it delivered.
Prevent: read the sending record for the form's notification when a sender, a domain policy or a member of staff changes.
A connection to the CRM that stopped, or a record it refuses or folds into another
Here the notification may arrive while the record does not appear, or appears as an old record with nothing new on it. The check is the reading in syncing the CRM and the marketing platform: the error queue, the counts on each side and a sample of records. The worked pattern is the diagnosis for LinkedIn leads that do not sync to the CRM, which the syncing guide says applies to any form that writes into the CRM, and three of its faults, a connection that stopped, a record refused and a repeat submission folded into an existing record, carry over to a website form as general practice, the last under the CRM's deduplicationRemoving double counts when one conversion is reported by more than one source. rules.
Fix: apply the matching fix in that diagnosis, and record by hand the requests customers reported, as the opening section of this page describes.
Verify: send a marked test record through each affected form and look at the source recorded on it in the CRM, the check the CRM sync change checklist asks for.
Prevent: read the error queue and the counts on a schedule with a named owner, as the syncing guide asks.
A record that exists but sits out of sight of the person who must respond
The submission reached the record system, yet nobody replied. The check is to find the test record with its owner or queue and then to open the view the responder works from. Lead routingRules that assign each new lead to an owner, queue or next step based on territory, segment, score or source. decides who receives a new record, and a record can wait where nobody looks. This page follows the record no further than that: sales not following up on marketing leads owns routing that assigns nobody and leads that do not reach the screen sales uses, and the diagnosis of leads assigned to the wrong salesperson covers a lead that did reach an owner, but the wrong one.
Fix: follow the fix in those pages; the sales follow-up diagnosis gives each record an owner and adds a fallback queue.
Verify: a marked test appears under a named owner or queue, and the responder confirms they can see it.
Prevent: run the lead routing QA checklist with a test lead from each form after a change to forms, fields or the team.
A change that moved where submissions go
The landing page checklist asks for a rerun after a redesign and whenever the form or its destination changes, with a test submission confirmed to have arrived at the destination. The diagnosis of a rate that fell after a redesign lists forms among the things a rebuild changes, with new fields, changed rules or a changed tool; the check here is the destination listed for the form against where a test submission lands now.
Fix: as general practice, restore or re-point the destination so the two agree, then repeat the test.
Verify: the marked test arrives at the listed destination and its named owner sees it.
Prevent: rerun the destination check after a redesign and after each change to a form or its destination, and keep the destination written beside the form, as the routine below describes.
Prevention: a marked test submission through every live form, with a named destination and owner for each
The diagnosis of abandoned forms asks for a scheduled test through every live form, and the walk in checking conversion paths after a release makes it specific for an enquiry path. For the forms on a site, three practices turn that into a routine.
- Send a marked test submission through each live form after each release, and on a schedule. The tracking verification checklist asks for a test conversion, such as a form submission from a known address, that can be told apart and taken out afterwards, and its monthly pass over the key journeys is a place to attach the schedule. Follow each test through the four places above, and write down the result with its date beside the form.
- Write the destination and an owner beside each form. Name where the submission is meant to arrive, meaning the mailbox, the record system and the view the responder opens, and the person who answers for it. A staff change, a tool change or a rebuilt page then has someone whose job it is to repeat the test, and a connection or a mailbox no longer depends on one person's memory.
- Compare counts and alert the owner when they diverge. Set the submit attempts recorded for each form beside the entries received, read against that form's own history. Alert the owner when a test fails to arrive or the gap moves, instead of waiting for a customer to write.
When a customer does report a missing enquiry, treat it as one traceable failure. The five whys asks why until the answer is a process, rule or check that can be changed, not a person, and for a form that failed unnoticed the chain may end at a missing test, an unwritten destination or an owner nobody named.
A thatching firm's missing quote requests (illustrative)
Pavrisko Thatching finds where its quote requests stop
Pavrisko Thatching, an invented thatching firm, has one conversion: a quote request form that asks for roof photographs, a postcode and a phone number. In one week 3 households write to say they sent the form and heard nothing. The owner replies to each household through the details they gave when they wrote in, records each request by hand in the customer record system and asks nobody to send the form again.
A marked test submission sent at 10:15 on Tuesday, with an invented household and a marked address, shows where the requests went.
| Place | What the marked test showed |
|---|---|
| Attempts against records received | A success message on screen at 10:15, and no entry received in the mailbox or the record system; the next two rows show why |
| The notification | Nothing in the firm's mailbox or its junk folder; a refusal in the sending record with a reason naming authentication, since the site began sending as the firm's domain without authentication |
| The record system | No new record, because the connection that copied each request there stopped when the person who set it up left the firm |
| The responder's view | Nothing to see, so nothing to check |
The firm authenticates the sender, reconnects the integration under a named owner whose access will last and writes a destination and an owner beside the form. It sends a marked test submission through each live form after each release, and the next test arrives in the mailbox, in the customer record system and in the owner's queue.
Visitors who start a form and do not send it, or a conversion rate that moved with no cause anyone can name, are the subject of other pages in the conversion rate troubleshooting collection, which routes each symptom to its diagnosis.
Common mistakes
- Asking the person to send it again before checking the path. The first request stays unrecorded, and the second may meet the same fault.
- Trusting the thank-you message. The record, not the message on screen, shows that the enquiry arrived.
- Testing the form and looking no further than the screen. The notification, the record and the responder's view are separate stages, and a test is not finished until it has been followed through each.
- Checking the inbox and nothing else. An entry can reach a mailbox while the record system is empty, or the reverse.
- Fixing the stage that was guessed. The marked test shows where the submission stops, and a fix aimed elsewhere leaves the loss running.
- Leaving the destination in one person's head. A connection, a mailbox or a form can depend on someone who leaves, so write the destination and its owner beside the form.
- Treating a missing analytics count and a missing enquiry as one fault. In the first the enquiries arrived and the counting failed, and in the second the enquiries did not arrive.
