A conversion is not a lead, and your report cannot tell the difference
Google Ads counts a conversion when an action you defined happens: a form submitted, a call answered past a minimum duration, a tap on a phone number. That is all it can know. It has no view of whether the person was real, whether they were in your service area, whether they submitted the same form twice, or whether they hung up the moment they realised they had dialled the wrong company.
So the dashboard reports forty-one conversions, the sales team says it got maybe twenty usable enquiries, and both are telling the truth about different things.
This matters more than it used to because bidding is automated. Smart Bidding does not optimise toward leads — it optimises toward whatever fires the conversion tag, and it is extremely good at finding more of exactly that. If a third of your conversions are junk, you are paying an algorithm to go and find more junk, and it will.
Across the Tampa Bay accounts we audit, the counted conversion number typically overstates sellable enquiries by thirty to sixty percent. In storm-adjacent categories it is worse.
The four fake conversions: bots, mis-dials, out-of-area, duplicates
Four things account for nearly all of the gap, and they have different fixes.
Bot and spam form fills. Automated submissions hit contact forms constantly, and an advertised page is found quickly. The tells are consistent: a submission at 3:14am, a name with no vowel pattern, a message containing a URL or the words “increase your traffic”, an area code that does not exist. These fire your conversion tag exactly like a real enquiry.
Wrong-number and mis-dial calls. If your tracked number is one digit from a pharmacy, a school or a well-known local business, you will collect calls meant for them. Google counts a call conversion by duration, and the threshold is yours to set — set it short and a confused thirty-second conversation clears it.
Out-of-area callers and forms. Genuine people, genuine interest, physically outside where you work — Pasco and Hernando enquiries reaching a business that only crosses the bay, or Sarasota jobs a Pinellas crew will not drive to at that ticket size.
Duplicates. One person, three conversions: they submit the form, the confirmation page reloads, they call in the same session. Whether that counts once or three times is a setting, and on a lead-generation form it should be “one”, not “every”. Plenty of accounts we open have it the other way round, because the default suits ecommerce.
Florida’s storm cycle and the post-hurricane lead surge
When a system comes up the Gulf, roofing, restoration, tree and water-damage accounts see form volume double or triple within days. It looks like the best week the account has ever had.
Two things are happening underneath. Out-of-state storm chasers move into the auction with budgets set for a short campaign, lifting click costs across every related keyword whether or not you bid on damage terms. And the enquiries arriving are largely people gathering estimates for an insurance claim, a large share of whom are not buying from anyone for weeks — some never, because the adjuster denies it.
Conversions rise, cost per conversion may even hold steady, and close rate falls off a cliff. Running target CPA, the system reads that surge as success and pushes harder into it.
The response is not to pause. It is to keep the qualified-lead measurement running through the surge specifically, so that when the auction normalises you know which campaigns produced revenue and which produced paperwork — the same demand-window problem we describe in why roofers and HVAC companies rank differently.
Snowbird enquiries in July: real demand on the wrong timeline
An enquiry from someone sitting in Michigan asking about screen enclosure replacement for their Clearwater condo is a real lead. It is just a real lead for November.
This is a large share of off-season form volume in Pinellas and coastal Hillsborough, and it does two kinds of damage. It inflates the period’s conversion count, which flatters the report. And when sales calls and hears “we’re not down until after Thanksgiving”, the lead gets marked dead — so it is never counted when it does convert five months later either.
These are not junk. They need a third bucket: qualified but deferred. Ask for a timeline on the form, route anything beyond ninety days into a follow-up sequence rather than the call queue, and decide deliberately whether that conversion should feed your bid strategy at full value, reduced value, or not at all.
Presence versus interest: the location setting that spends your budget statewide
Every campaign has a location targeting option deciding whether you reach people in your targeted area, or people in it or interested in it. The interest version means somebody in Ohio researching Tampa can see and click your ad.
For a hotel or a relocation service that is exactly right. For a plumber it is a slow budget leak that never surfaces as anything more obvious than a slightly disappointing month.
The default has changed more than once over the years and varies by campaign type, so check it rather than assume. While you are in there, check the radius: a thirty-mile ring drawn from a St. Petersburg address reaches into three counties and across two bridges, which is a different business than the one you run. A list of specific cities and ZIP codes is almost always tighter than a radius, and it makes the out-of-area analysis far easier to read.
Measuring true lead rate in under an hour
You do not need new software for this. You need a spreadsheet and one uninterrupted hour.
- Export the last ninety days of conversions, or the equivalent CRM records filtered to paid search.
- Take one hundred consecutive records. Not a sample you choose — consecutive, or you will unconsciously pick the good ones.
- Tag each: sellable, out of area, wrong timeline, duplicate, bot or spam, mis-dial.
- For calls, listen to the first fifteen seconds — that is where the mis-dials live, and it beats reading notes. Florida is an all-party consent state, so if you record, your call tracking needs the announcement enabled.
- Divide sellable by one hundred. That is your true lead rate.
- Recalculate cost per lead as spend divided by sellable leads rather than conversions.
That last number changes the conversation. An account reporting $94 per conversion at a 55% true lead rate is buying leads at $171 — which may still be fine, but it is a different decision than the one the dashboard was inviting.
Run it per campaign. The rate is rarely uniform, and the campaign with the worst headline cost per conversion is frequently the one producing the cleanest leads.
Feeding qualified leads back with offline conversion import
Measuring it is worth something. Feeding it back is worth more, because it changes what the bidding chases.
The mechanism is offline conversion import. Turn on auto-tagging, capture the click identifier from the landing page URL into a hidden form field, and store it against the record in your CRM. When a human decides the enquiry is real, upload that outcome back against a separate conversion action — “Qualified Lead” rather than “Form Submitted”. Uploads can be sent no earlier than six hours after the click and must fall inside your conversion window, so a weekly routine is usually the right cadence.
If that plumbing is beyond your setup, enhanced conversions for leads achieves much the same thing using hashed contact details instead of the click ID, and it is a shorter project.
Two cautions. Automated bidding needs volume; a qualified-lead action firing four times a month cannot steer a bid strategy, and in thin accounts the better move is to keep the form conversion primary but assign different values to qualified and unqualified outcomes and bid to value. And whatever you import, make exactly one conversion action primary — accounts marking five as primary are averaging signals rather than sharpening them.
Configuring this properly is the substantial part of what PPC management should be doing for a lead-generation business, and it is the difference between a Google Ads account that compounds and one that plateaus.
What to change on the landing page and the form
Roughly half the fake-conversion problem is fixable on the page, before the bidding ever sees it.
- Add an invisible honeypot field and a scoring-based CAPTCHA. Puzzle CAPTCHAs cost you real leads; invisible checks do not.
- Make the form endpoint reject direct posts. A meaningful share of spam never loads your page at all — it posts straight to the handler.
- Require a ZIP code and validate it. This filters out-of-area submissions at the door and gives you a clean field to analyse later.
- Ask for a timeline. One dropdown separates the November snowbird from the burst pipe.
- Validate phone format. It removes a large slice of automated junk by itself.
- Stop counting everything as the same conversion. A newsletter signup, a chat and a quote request are three events, and only one is worth bidding on.
- Match the ad to the page. Vague ads produce accidental clicks and confused submissions; specific ones produce fewer conversions and more leads.
Most of these are an afternoon of work. If the page itself is the constraint, our landing page design work starts here, and the common conversion mistakes piece covers the structural side.
What good actually looks like by industry in this market
True lead rates vary enormously by category. These are the bands we typically see in Tampa Bay accounts — a starting benchmark to argue with, not a standard:
- Emergency home services (HVAC, plumbing, electrical): 55–70%. Call-heavy, so mis-dials and hang-ups are the main loss.
- Roofing and restoration: 45–60% normally, falling to 25–35% in the weeks after a named storm.
- Considered home improvement (windows, remodels, pool enclosures): 50–65%, with timeline mismatch the largest single loss.
- Legal and injury: 30–50%. The highest click costs in the market and the heaviest form spam — a punishing combination.
- Professional and B2B services: 35–55%, dragged down by job applicants and vendor pitches using the contact form.
If your number sits in the band and your cost per sellable lead works, the account is healthy and the reporting was describing the wrong thing. If it sits well below, the fix is nearly always one of the four causes above rather than a bidding change.
Either way the sequence is the same: measure the real rate, clean the form, tighten the geography, then feed qualified outcomes back so the algorithm optimises toward leads you can sell. If you want that audit run on your account, get in touch and we will tell you which of the four is costing you the most.