What you'll be able to do

Read a full Google Ads landing page teardown in about twenty minutes and name exactly where the money is going. Then fix it in the order that costs the least and moves the most: headline first, page second, tracking third, follow-up last. Follow-up is the one that quietly leaks again the moment you stop watching it.

  • Spot the ad-to-headline mismatch in under a minute.
  • Decide what earns a spot above the fold on a phone.
  • Stop Smart Bidding from optimizing toward junk leads.
  • Build a five-minute response chain without writing code.

What you need

  • Read access to Google Ads and Google Analytics.
  • Microsoft Clarity (free, unlimited session recordings and heatmaps).
  • A Zapier, Make, or n8n account for the follow-up chain.
  • Lyssna (formerly UsabilityHub) if you want to prove the hero is failing using real human responses.

The Setup: A $3,000/Month Account That Already Works, and Still Leaks

Your dashboard says nine leads last month. Your dispatcher says three of them were worth driving to. Nothing in the account looks broken, and that is the problem.

Here is the whole story in two sentences. The ad says “24/7 Emergency Commercial Refrigeration Repair, 90-Minute Response.” The landing page headline says “Welcome to Northgate Mechanical, Your Trusted Partner for HVAC Solutions Since 1998.”

The ad was probably written by a freelancer in 2023. The page was written by a designer whose brief was “make the brand look professional.” Neither person ever read the other’s work.

That organizational distance, not incompetence, is what you are hunting in a Google Ads landing page teardown. Three thousand dollars a month buys roughly 300 clicks at $10. At a 2.9% landing page conversion rate that is 9 leads, 3 qualified, and about 0.9 closed jobs at $6,500 average.

Revenue lands near $5,850. Gross profit at 45% is roughly $2,632 against $3,000 of spend. The campaign is underwater while the dashboard looks calm.

Now say what is working, out loud. Campaign structure is sane, so match types and negatives are not the villain. There are 87 Google reviews at 4.7 stars sitting unused near the footer.

The 90-minute guarantee is real and genuinely differentiated. And the ad’s call extension converts better than the landing page does. That last one is the strongest signal in the account, and a hint: these buyers want to talk, not type.

Model the same $3,000 with a 6.5% page conversion rate and a 45% qualified rate. That is 20 leads, 9 qualified, about $333 cost per qualified lead, and roughly $9,360 gross profit. No part of that model needs more ad spend.

Leak 1: The Ad Makes a Promise the Headline Doesn’t Keep

This is message match, or “scent” in the MECLABS sense: one continuous thread from ad to headline to form to thank-you to follow-up. The blunt test is that the ad’s promise should appear within the first 40 words of the page. Here it appears nowhere.

Rewrite the H1 to “24/7 Commercial Refrigeration Repair in [City], 90-Minute Emergency Response,” with a subhead of “Licensed techs on call now. Same-day repairs, no overtime surcharge.” You are not writing new copy. You are moving the promise from one asset to the other. Roughly an hour of writing, zero dollars of build, and it is the highest-leverage fix in the entire teardown.

Prove the failure before you change anything. Install Clarity and watch ten recordings of real ad traffic on mobile. You will usually see people scroll past the hero, hunt for a phone number, tap the logo, and leave.

Then run a five-second test on the current hero and ask what the company does and how fast it responds. A handful of stranger responses will persuade the owner far more than any heatmap.

One wrinkle to check while you are inside the account: AI Max for Search and Performance Max final-URL expansion can push paid clicks to blog posts, category pages, and “About Us” with no human involved. That is machine-generated mismatch, the fastest-growing version of this leak. Set brand exclusions and negative keyword lists before you enable either.

The First Screen on Mobile: Hero, Proof, Form, and the 3-Second Cliff

Above the fold, four things belong: a real text headline with the emergency promise, a tappable phone number, a trust bar (4.7 stars, 87 reviews, license number, since 1998), and a three-field form asking name, phone, and “what’s broken?”

Right now the 90-minute promise lives inside a hero JPEG that does not paint until 4.1 seconds, and the number is not clickable. Nielsen Norman Group found users spend roughly 57% of page-viewing time above the fold and about 74% in the first two screenfuls. That footer testimonial carousel is decoration.

Form friction is where most landing page conversion rate optimization projects earn their keep. Nine fields become three: single column, 16px inputs so mobile Safari does not zoom, a numeric keyboard on the phone field, inline validation, and a button that reads “Get a Tech Dispatched” instead of “Send Message.”

“Budget Range” and “Company Size” move to the phone call, where a human can ask them in context. For a $6,500 job, asking budget on the first touch filters out everyone who does not know their budget yet, which is most of them.

The speed budget is LCP under 2.5 seconds, INP under 200ms, CLS under 0.05, measured at the 75th percentile. INP replaced First Input Delay in March 2024, so any audit still reporting FID is stale. Google’s own research found roughly 53% of mobile visits are abandoned past three seconds, and Deloitte’s retail work tied a 0.1 second speed gain to about an 8% conversion lift (directional for B2B, not a promise).

Diagnose in this order: CrUX field data first, because that is what Google actually sees, then Lighthouse, then a tag audit. Here that meant a 2.8MB hero PNG compressed to a 140KB WebP with high fetch priority and explicit dimensions, 17 Google Tag Manager tags deferred to scroll or a three-second delay (two chat widgets and a heatmap were blocking paint), and Google Fonts swapped for two system-stack weights. Target: LCP from 5.4 seconds to under 2.3.

Proof placement is free money. Move the review block next to the form and add one job-photo case study directly below the fold: “Replaced a failed walk-in compressor for a 12-unit restaurant group in four hours.” Boring, specific, verifiable proof beside the ask beats a logo strip every time.

If your testimonials read like they were written by a machine, that is its own problem, covered in our breakdown of social proof after AI.

Leak 2: Tracking Integrity, or Stop Teaching Smart Bidding to Chase Garbage

The most expensive mistake here is also the most common. The primary conversion action was a page view on the thank-you page plus a separate form submit event. That double-counts, and worse, it hands Smart Bidding a soft signal to optimize toward.

Reduce it to one primary conversion called generate_lead, with everything else marked secondary. Smart Bidding is a mirror, not a fix. Point it at a soft conversion and it will cheerfully scale low-intent traffic with beautiful efficiency metrics.

Also watch terminology drift. GA4 renamed “conversions” to “key events” in 2024, so make sure you and your client mean the same number before you argue about it.

Clean the click-to-session link. Auto-tagging (gclid) stays on, but manual UTMs were also appended to the Google Ads final URLs, which is why GA4 showed both a “google / cpc” row and a “Google Ads” row. Remove the manual UTMs.

Then accept that GA4 and Google Ads will still disagree, because GA4 retired first-click, linear, and position-based attribution for web in mid-2023, leaving data-driven as default and last-click as fallback. Plan for the gap instead of panicking about it.

For B2B, the single highest-leverage bidding input is offline conversion import. Wire CRM stages back into Google Ads with the GCLID: qualified around day three, quoted around day seven, closed_won between day 30 and 60.

Extend the conversion window to 90 days click-through, or Smart Bidding will systematically undervalue slow-closing, high-value keywords. Pair it with Enhanced Conversions for Leads using hashed email and phone.

Two adjacent items erode the numbers quietly. On Meta, a pixel alone under-counts, so you need the Conversions API with event_id deduplication, ideally through a server-side container.

And if you touch EEA or UK traffic, Consent Mode v2 mistakes cost real volume: strict consent typically hides 20% to 40% of observable conversions. Model it, do not pretend it is not happening.

Add call tracking too, since calls are usually 40% to 60% of local service leads and are almost always invisible in the report. If measured pixel loss creeps past 15% to 20%, that is when server-side tagging via a GTM server container, Stape, or Addingwell earns its keep. Our notes on AI ad creative tracking mistakes cover the same failure pattern from the creative side.

Leak 3: Speed to Lead, the 60-Second Text and the 5-Minute Call

Today the form routes to a shared inbox nobody owns, and the median first response is 19 hours. Harvard Business Review’s study of lead response found firms that contact a lead within an hour are roughly 7x more likely to hold a meaningful qualification conversation.

The same research found the average first response across firms was about 42 hours, and roughly 23% never responded at all. Older research points the same direction with a sharper edge: minutes matter far more than hours. Treat the exact multipliers as directional. Treat the direction as law.

The fix is a four-step chain you can build in an afternoon:

  1. Form submit fires an SMS to the lead within 60 seconds: “Got it, [Name]. A tech will call you in under five minutes. Reply STOP to opt out.”
  2. Simultaneous SMS and push notification to the on-call dispatcher.
  3. CRM deal created automatically with source set to the GCLID.
  4. If no human contact is logged within ten minutes, escalate to the owner.

Voicemail goes to Slack. After hours gets the auto-text plus a 7am call queue. One prerequisite people discover only after it fails: A2P 10DLC brand and campaign registration for US SMS numbers. Budget one to three weeks.

Unregistered messages get filtered silently, and you will never see an error, just an eerie quiet.

Then measure it. Median first-response time becomes a weekly KPI reported next to cost per lead, not a vibe. And resist automating the entire thing, because an auto-text gets a reply in 60 seconds but qualifies nobody.

The combination that works is text in 60 seconds, human call in five minutes, and a log of which one produced the booked job. The reason so many of these chains rot is covered in our piece on why CRM automation fails.

The 90-Day Sequence, and the Honest Build-vs-Buy Math

Order matters more than effort. Weeks 1 to 2: tracking only. One primary conversion, CAPI, Enhanced Conversions, call tracking. Do not touch the page yet, or you will never know what worked.

Week 2: Clarity sessions plus a five-second hero test to document the comprehension failure. Weeks 3 to 4: the new landing page, running 50/50 against the old one, with offline conversion import live from day one.

Week 5: speed-to-lead automation. Weeks 6 to 12: test one variable at a time, headline first, then offer, then form length, with two to four weeks minimum per test and a sample size declared in advance. If you are testing button colors, you are avoiding the actual work.

The honest DIY case. A page builder, Zapier, and a free Clarity account genuinely handle steps one, three, and five. Tooling runs $50 to $250/mo. Setup is realistically 30 to 50 hours, then 6 to 15 hours a month of upkeep.

At $150/hr, that is $900 to $2,250/mo of invisible spend. The failure mode is not the build. It is the decay: a site migration breaks the form handler in month seven, GTM quietly accumulates four more tags, a plugin update reverts your primary conversion, and nobody notices because leads still sort of arrive.

Managed help typically runs $1,500 to $6,000/mo in retainer, or $3,000 to $15,000 for a one-time rebuild. Faster, and you inherit someone else’s tracking conventions. But you still own lead follow-up no matter who builds the page.

Do not outsource speed-to-lead to an agency and then route the leads to an unowned inbox.

So frame it honestly. DIY for the first 90 days is a sensible choice, and it teaches you what you are actually buying. Past 90 days, the question is not “can I do this cheaper.” It is “will I actually do this in month eight, on a Tuesday, when a technician is out sick and two jobs are running behind.”

If the answer is no, the retainer is cheaper than a leak that runs about $6,700 in gross profit every month, per the model above.

Where to go next

Load your own landing page on your phone, on cellular data, and time it. Then read your ad headline and your page headline back to back out loud. If they do not match, you already know your first fix, and it is the cheapest one on this list.

Then open Clarity, watch ten sessions, and write down the median first-response time for your last twenty leads. Those two numbers will tell you more about your real cost per lead than any dashboard tab.

You now have the full teardown, and most of it is genuinely doable with a page builder, a Zapier account, and a free afternoon. If you would rather have the page, the tracking, and the automated follow-up built for you in one fixed-scope pass, our Growth Sprint is a two-week $1,500 build with no retainer: see the Growth Sprint.

Cover photo by Georgie Devlin on Pexels.