First, find where the poor leads come from
The account cannot answer this question alone, because it sees submissions, not sales. Export a few weeks of leads from the CRM with the outcome sales recorded for each, such as accepted, rejected as poor fit, rejected as spam or unreachable, and join them to the click data: campaign and campaign type, network, keyword and search term where available, the landing page and the form. Leads captured with the click identifier or matched through enhanced conversions can be joined directly; otherwise use the landing page and time of submission with tagged final URLs.
Then read rejection by source. Spam that clusters in one campaign type or network, arriving at odd hours with nonsense in the fields, is a different problem from real people who are simply the wrong customers. Poor fit that clusters on certain search terms points to matching; poor fit spread evenly across Search points to the ads or the form. Count the accepted share per source, not just the rejects, since a large campaign will contain many rejects even when it is the best one. Keep this page's question separate from two neighbours: no conversions at all belongs to the diagnosis for clicks but no conversions, and money spent on unrelated queries belongs to the diagnosis for irrelevant search terms. When the same poor-fit reasons also arrive from directories, referrals and other platforms, the cause sits outside the account, and the diagnosis of enquiries from the wrong kind of customer traces it to the promise, the offer and the target that every channel shares.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Reported conversions rise while the share sales accepts falls | Which conversion action is primary for bidding; whether any CRM outcome is imported | Bidding rewarded for every form submission, with no qualified outcome returned | Import a qualified stage, make it primary and demote the raw form |
| Rejected leads cluster on searches about jobs, courses, prices or selling to the business | Search terms behind rejected leads; match types of the keywords that matched them | Broad matching reaching research, job and supplier searches | Narrow the match types for those themes and add negatives |
| Spam and casual fills cluster in Performance Max or Display, often at odd hours | Accepted share by campaign type and network; placement report | Performance Max or Display inventory producing low-intent or automated form fills | Apply the documented controls and judge the campaign on qualified leads |
| Nonsense entries, repeated submissions or the same details many times | Form submissions against the form's protection and server checks | Forms with no protection against automated submissions | Add a hidden field or server-side check, with an accessible alternative |
| Real people enquire who cannot afford, are outside the area or below the minimum | Ad copy and landing page for price, audience, area and minimums | Ads that never say who the offer is for, its price or its minimums | Add true, specific qualifiers to the ads and the page |
Fix, verify and prevent each cause
Bidding taught to collect forms
Smart BiddingGoogle Ads' automated bid strategies that set each auction's bid from conversion signals. sets a bid in each auction from a prediction of whether that auction will convert, so it learns from whatever the account counts, as Google's Smart Bidding help page describes. Counting every submission teaches it to find the people who submit most readily. The enhanced conversions guide sets out the leads variant: hashed contact details captured at submission, and the later outcome uploaded against the same hashed data so Google can match it to the click, a route Google documents in About enhanced conversions.
Fix: agree with sales what a qualified lead is, often a threshold in the business's Lead ScoringAssigning points to leads for fit and behaviour to decide which are ready for sales. model, upload that stage, make it the primary action for the lead campaigns and keep the raw form as secondary.
Verify: the import action records matched outcomes each week, and once bidding has had time to adjust, cost per qualified lead falls even if total forms drop.
Prevent: give the upload a named owner and a schedule, and alert when a week passes with nothing imported.
Broad matching into the wrong conversations
The match types guide explains, from Google's keyword matching options page, that broad match enters searches related to the keyword's meaning, including ones that share no words with it, and that the search terms report is the only record of what matched. A keyword about installing something can reach people looking for jobs installing it, courses on it, or suppliers who want to sell parts to the installer. With a form-based conversion behind it, the system has no reason to tell them apart.
Fix: move the themes that produced rejected leads to phrase or exact match, add negatives for job, training and supplier wording, and keep broad match only where a qualified outcome guides the bidding.
Verify: the rejected share for those themes falls without the accepted leads disappearing.
Prevent: read the search terms behind rejected leads each month, not only the costly ones.
Inventory where a form fill is cheap
Performance Max reaches inventory across Google's channels from a single campaign, as Google's Performance Max overview describes, and some of that inventory, along with Display placements, makes a casual or automated form fill easy. The Performance Max for lead generation guide puts a qualified outcome in place before launch, separates lead goals into their own campaigns, and lists the documented controls: brand exclusions, account-level negatives, final URL expansion off or limited, and account-level placement and content exclusions fed by the placement report. For Display on its own, the diagnosis for low-quality Display placements covers the placement side.
Fix: apply those controls, point each asset group at a service page whose form qualifies, and read the Performance MaxA Google Ads campaign type that runs one goal across all Google inventory from shared assets. campaign on qualified leads for the whole account.
Verify: the spam share from that campaign falls and account-level qualified leads hold or rise.
Prevent: keep the placement review on the monthly calendar and treat a falling acceptance rate as a reason to pause, as the guide's stopping rules do.
A form any script can submit
If submissions arrive with nonsense names, copied text or the same details many times, the form is being filled by software, and every one of those fills may be reported as a conversion. The forms guide recommends spam protection that does not shut people out: a hidden field, a server-side check or a provider with an accessible alternative, rather than a challenge only sighted people can solve. It also builds the form from the next step, drawing on Nielsen Norman Group's form usability recommendations, so a postcode that confirms coverage or one qualifying question can earn its place.
Fix: add the protection, fire the conversion only on a confirmation that a valid submission reaches, and drop flagged submissions before they count.
Verify: nonsense entries disappear from the CRM and the reported conversion count falls towards the real one.
Prevent: test the form after every site release, including a deliberate junk submission to confirm it is stopped.
Ads that invite everyone
Some poor leads are genuine people the ad never warned off. The guide to pre-qualifying ad copy uses plain statements of price, audience, minimum order, service area, eligibility or lead time to tell the wrong people not to click, and requires each to be true, current and repeated on the landing page. It also explains that a qualifier in one headline may rarely show, so a message that must always appear is pinned, and that click-through and Quality Score may dip as a result.
Fix: write the qualifier that matches the commonest rejection reason, pin two or three phrasings to the same position, and state the same terms on the page.
Verify: the qualification rate rises and cost per qualified lead falls, read from sales outcomes after they have had time to arrive.
Prevent: update qualifiers whenever prices, minimums or service areas change.
A prevention routine for lead quality
Lead quality decays quietly, because the account's own columns look healthy while the sales team's day gets worse. Hold a short monthly review with sales. Bring the accepted share and cost per qualified lead by campaign type, network and keyword theme, the top rejection reasons in sales' own words, and the date of the last successful outcome import. Each rejection reason should map to one owner: bidding and imports, matching and negatives, inventory controls, the form, or the copy.
Before launching any new campaign type, confirm three things: the qualified outcome is primary for it, the form it sends people to is protected and qualifies, and the ads state who the offer is for. After any change to the website, submit a test lead and a deliberate junk entry, and confirm the first is recorded once and the second is stopped. When volume falls after a quality fix, read qualified leads before reacting. The other symptoms an account can show, from ads that do not appear to rising costs, are listed in the Google Ads troubleshooting section.
Zibberfold Roofing traces its enquiries to the source
Zibberfold Roofing, an invented roofing firm, adds a Performance Max campaign beside its Search campaigns. Monthly enquiries rise from 160 to 330, while booked roof surveys fall from 52 to 44. The marketing lead joins the CRM outcomes to the click data. Search enquiries are accepted at 46%; Performance Max enquiries at 9%, and 61% of the rejected Performance Max ones are marked spam: invented names, repeated phone numbers, most between 1am and 5am. The placement report shows the fills coming from a cluster of mobile apps. Within Search, rejections cluster on broad-match searches about roofing jobs and roofing courses. The form has no protection, and the primary conversion is any submission. The firm adds a hidden field and a server-side check, uploads booked surveys through enhanced conversions for leads and makes them primary, excludes the app placements, moves the job and course themes to negatives, and adds a pinned headline: "Pitched And Flat Roofs, Homeowners Only". Two months later enquiries are 190 a month, booked surveys 63, and cost per booked survey falls from £118 to £79.
Common mistakes
Judging the account on conversions while sales judges it on customers. Blocking spam at the form but still counting the blocked fills as conversions. Turning off Performance Max before giving it a qualified signal. Writing a qualifier into one unpinned headline and assuming it shows. Reacting to the drop in volume that a quality fix was meant to cause.
