Cut junk form fills at the source
MarqIQ labels a junk fill after it has arrived; it never stops anybody submitting a form. So the work here is in two halves: use the analysis to find out which form and which source produce the junk and why, then put the control in the place that can refuse it, which is your own form tool. If you only want to know what MarqIQ counts as junk and on what basis, read What counts as a junk form fill? first.
Before you start
Plan required
Any plan except Starter, which does not include the Analysis screens. Switching fill-quality signals on needs a workspace admin, because it changes what runs on your website.
- The tracking script live on the pages the forms sit on. Without it there are no captured fills to judge.
- Administrator access to whichever tool your forms are built in, because every control that refuses a submission lives there.
- An export of the records you are about to remove, taken before you remove anything. MarqIQ cannot put back a record it never wrote.
Size it before you change anything
Open Insights in the left-hand menu, then Analysis.
Pick the tool your leads land in along the top.
Open the Form fills tab.
Read the four tiles across the top.
Captured is every fill MarqIQ recorded. Recorded twice is the same person on the same form within two minutes, counted once. Junk gives the count and the share. Real fills is what every rate on the page is measured on.
Check the line under Captured before you quote any of it.
It reads “every fill MarqIQ recorded”, or “the most recent; older fills are not shown” once the analysis has hit its cap of 60,000 captured fills. Those are different denominators and the tile says which you have.
Repeats are not junk
A double-fired form event and a visitor who double-clicks both produce the same fill twice. MarqIQ removes those first, before anything is judged, and counts them separately. Where repeats are 10% or more of captured fills, What to do next says so — any report of yours that counts raw fills is overstating volume by about that much.
Find which reason dominates
The What the junk is card lists each reason with its count and up to three masked example addresses, largest first. The reason decides the control, so work from the largest one down.
| Reason | The control that actually stops it |
|---|---|
| Test entries | Your own team. Agree one test address, exclude it from lead counts and reporting |
| Throwaway mailboxes | A blocked-domain list on the form, or an email confirmation step before the lead is created |
| Email addresses that cannot exist | Format and domain validation on the form field |
| Made-up names | A required field with a minimum length rarely helps. An email confirmation step does |
| One visitor, many email addresses | A per-browser submission limit on the form |
| Automated floods | A hidden trap field, a challenge, or a per-form rate limit |
| Filled in by a program, not a person | A challenge on that form. This reason is only read once you switch fill-quality signals on |
None of these controls is MarqIQ’s
MarqIQ adds no hidden trap field to your forms, runs no challenge, blocks no domain, reads no IP address and applies no per-browser or per-form limit to a submission. An unknown field can be rejected by a CRM’s own form, and an embedded HubSpot form cannot be reached at all, so the risk outweighs what it would add. Every control in the right-hand column is set up in the tool your form is built in.
Find which form and which source it comes through
Read the Where the junk comes from card.
It lists sources with at least 30 fills, the twelve with the highest junk share and no more. A source is called out only when its junk share is at least 1.5 times the overall share or at most half of it, so a source a few points off the average is left alone rather than accused.
Read the Junk column on the Your forms card.
Up to 40 forms, biggest first. Each row also gives the share found in your tool, the share that went on to buy, the source most of its fills arrive from and the date of the last fill. A form with at least ten fills and none in 60 days carries a gone quiet chip.
Read What to do next at the foot of the tab.
It names the worst source and the worst form by name, with their own figures, rather than telling you to check your forms.
Check the Month by month chart for the month junk jumped.
It shows the last twelve months as real fills and junk. A jump usually lines up with a campaign starting or a form changing, which is the fastest way to name the cause.
Tip
A source called out for junk that is also paid traffic is the one to act on first. You are paying for those fills twice: once for the click, and again in the sales time they consume.
Switch on fill-quality signals
This is the one thing MarqIQ adds, and it adds evidence rather than a barrier. A competent bot types a plausible name and a real-looking address, so on content alone it reads as a person. How the form was filled in is harder to fake.
Open Tracking Script in the left-hand menu.
The page is headed Script & Deploy.
Find the card headed Spot form fills made by programs.
It appears only when the script this build serves carries the collector. Only a workspace admin can change it.
Turn the switch on.
It applies to fills from the next time a page loads your script. Once fills start arriving, the card says how many have been captured with signals.
Come back to Form fills a day later and look for the Filled in by a program, not a person reason.
Fills captured before you switched it on carry no signals and are never judged on this.
What it records, and what it does not
- Three things about each fill: how long after the page loaded it was submitted, how many times the visitor pressed a key, clicked or touched — a count, nothing more — and whether the browser says it is driven by automation.
- On the server, MarqIQ reads the browser’s own identifying header once and reduces it to one word. That header itself is not stored. No IP address is read, stored or looked up. Nothing that was typed is kept.
- A fill is called automated on any one of four things: the browser declared itself automated, it came from a headless browser, it came from a program identifying itself as a bot or a script, or it was submitted within a second and a half of the page loading with no key press, click or touch at all.
- A fast fill by a person using autofill still involves a click, so it is never flagged. A fill with no signals is not judged at all.
- While the switch is off, this code is not in your script. Turning it off restores the old script on the next page load and stored signals are ignored.
- It labels fills. It never stops anybody submitting a form.
Note
Consent is unchanged by this. Nothing is sent unless the existing consent gate already allows the fill itself to be sent.
Take the over-exposed form off the field
On HubSpot and Marketo, MarqIQ also reads your tool’s own forms and what the people they brought in went on to do. Three findings there are worth acting on beside the junk.
- A form that brings people in who do not buy: at least 30 people came through it, none is on a deal, and its conversion sits below half the rate across all your forms even at the optimistic end of the range. Rename it for what it is, or route its fills somewhere that is not sales.
- Intake resting on one form: 70% or more of everyone a form brought in came through a single form, on at least 100 people, and with more than one form recorded. One broken embed moves the whole month. Set an alert on that form’s daily submissions and measure a second route in.
- Live forms nobody fills: three or more forms are live and brought nobody in. Each is a place a real fill can be lost silently. Open each one, fix or move the ones that should work, and archive the leftovers.
Note
These three read your tool’s own record of form conversions, not what MarqIQ’s script saw. A workspace that never installed the script still has them; a workspace on any other tool does not.
Clean up the junk you already have
MarqIQ will not do this part. The changes it can propose today fill empty first-touch fields on HubSpot and Marketo people; it removes nothing and it edits no existing value. So the clean-up is yours, and the order matters.
Build the same list in your own tool using the reasons above.
Work one reason at a time. A list built from “test address or disposable domain” is defensible; a list built from “looks like junk” is not.
Export that list before you touch it.
Take the export even if you are only going to suppress rather than delete. It is the only copy you will have.
Read a sample of 20 rows yourself.
An unusual real name is never treated as junk by MarqIQ, and a personal mailbox such as Gmail is never junk either. Check that your own list keeps both of those promises before you act on it.
Suppress rather than delete wherever your tool bills by marketable contacts.
Setting a record to non-marketing removes the cost without removing the record, its history or its deals.
Delete only what you are certain of, and only after the export.
A delete cannot be reversed by MarqIQ. Nothing in MarqIQ wrote those records, so nothing in MarqIQ can put them back.
Check the controls worked
Wait for a fortnight of traffic through the changed form.
Open Form fills again for the same tool.
Compare the Month by month chart, not the overall share.
The overall junk share includes every month of history, so it moves slowly. The current month’s bar is what tells you whether the control is holding.
Check the named form’s own Junk figure on the Your forms card.
That figure is the one your change was aimed at, and it has its own fill count beside it.
Take care
Quote the denominator whenever you report this. “Junk is down to 4%” means nothing on its own. “4% of the 3,180 fills captured this quarter, against 19% last quarter” is the same figure and can be checked.
If it did not work
The tab says no form fills have been captured yet
Why: The analysis reads fills MarqIQ’s own script captured. Either the script is not on the pages your forms sit on, or it is not firing.
Fix: Open Setup & diagnostics in the left-hand menu. It checks the script install, the fills that arrived and the consent setting, and it names a figure it could not read rather than showing a zero. Where Wait for consent before sending anything to MarqIQ is on under Tracking Script, a banner that never reports a decision holds every fill back; that setting is off unless somebody turned it on. Records already in your CRM do not appear on this tab either way: this half of it is about fills MarqIQ’s own script saw.
The Captured tile says older fills are not shown
Why: The analysis reads the most recent 60,000 captured fills. Your workspace has more than that.
Fix: Nothing to fix, but change what you quote. Every figure on the tab is measured on those 60,000, so say so, and narrow the Created range when you want a period you can name.
The Spot form fills made by programs card is not on the page
Why: Either the tracking script this build serves does not carry the collector yet, in which case the card is hidden entirely, or you are not a workspace admin, in which case the card says only an admin can change what the tracking script does.
Fix: Ask a workspace admin to open the page. Where they cannot see the card either, the collector is not in the served script yet and there is nothing to switch on.
You switched signals on and no fill is labelled automated
Why: Signals apply from the next page load onwards, and a fill with no signals is never judged. It is also a genuine possible answer: your junk may be typed by people rather than posted by scripts.
Fix: Give it a few days of traffic, then read the reason list again. Where the other reasons carry the whole count, the controls in the table above are the ones to reach for.
A real customer appears in the junk list
Why: One of the seven reasons matched their record. The commonest is one browser having submitted four or more different addresses, which happens on a shared machine at an event or on a reception desk.
Fix: Read the reason shown against them. MarqIQ never treats an unusual real name or a personal mailbox as junk, so those two are not the cause. Exclude the record by hand and, where a shared machine is the cause, expect it again at the next event.
A busy launch minute was counted as a flood
Why: A flood is eight or more fills on one form inside a sliding minute from eight or more different addresses. A genuine launch can clear that bar.
Fix: Check the door test: where one source accounts for three-quarters or more of that minute, only fills from that source are flagged, so a real visitor arriving from elsewhere is not swept up. Where your launch genuinely sent eight people through one channel in a minute, read the month chart to confirm it was one minute and not a pattern.
The junk share fell but your CRM still fills with rubbish
Why: The analysis measures fills MarqIQ captured on your website. A record created by an import, a list upload, a sales tool or another integration never passed through a form.
Fix: Read Source truth on the same Analysis screen for where those records come from, and Audience health for how many of them can be emailed at all.
You want MarqIQ to block the junk rather than label it
Why: MarqIQ labels fills after the fact by design. It adds no trap field, no challenge, no domain blocking and no IP reputation check, and it never refuses a submission.
Fix: Put the control in the tool your form is built in, using the table above. Use MarqIQ to decide which form and which source to put it on, and to measure whether it worked.
Next steps
Related articles
Was this article helpful?