Skip to content
ADIMIZE

Case study · Dumpster rental & junk removal

Busy on the phone. Invisible in search.

A dumpster rental and junk removal company in Fairfield County, Connecticut. Plenty of work coming in. Almost no way for a new customer to find them. This is what we changed, and what we are not willing to claim yet.

Full disclosure — we hold an ownership stake in this business. We ran this on our own revenue before we ran it on anyone else’s.

organic sessions, January 1 – July 31, 2025

1,618

organic sessions, January 1 – July 31, 2025

organic sessions, January 1 – July 31, 2026

4,877

organic sessions, January 1 – July 31, 2026

same window, one year apart

3.0x

same window, one year apart

Both figures are full January-through-July windows from GA4, one year apart. The work landed in May 2026 — the windows are shown whole so nobody can say we cherry-picked the slice.

The starting picture

Who this operator is.

Everything in this section describes the business before our work landed in May 2026. None of these numbers are a result of anything we did — they are the picture we walked into.

Three things were wrong. None of them were the phone.

Search was barely carrying anything. In the seven months before the work, organic search brought 1,618 sessions — while paid ads carried 15,510 over the same window. The phone stayed busy, but almost every new-customer visit was rented, not earned.

The town pages were near-copies of each other with the town name swapped. One page written once, then repeated for every town with a find-and-replace. Search engines and AI tools both throw that kind of page away.

Their business information was inconsistent across the web. The name, address, and phone details didn’t match across the directory profiles sites like Yelp, Apple Maps, and Bing keep about a business. When those don’t line up, search engines can’t confirm who you are.

Before our work · January 1 – July 31, 2025

Where sessions came from, before

Paid Search15,510
Direct2,785
Organic Search1,618
Organic Social216
Referral167
Cross-network162
Unassigned150
Every channel row from the GA4 traffic export, January 1 – July 31, 2025 — nothing omitted. Paid search carried nearly ten times what organic did. That is the dependence problem this whole engagement was built to fix.

This was our own book of business. We are not describing someone else’s mess. That is exactly why we know what it costs to leave it alone.

What we did

What we did.

Four moves. No tricks, no special files, nothing you need to be technical to understand.

1. Rebuilt the site so it loads fast and reads clean

The old site worked, but it was slow and hard to change. We rebuilt it so every page loads quickly on a phone and reads like a person wrote it. That is the floor. Nothing else works until the site does.

2. Replaced the copy-paste town pages with real ones

Every town page got rewritten to say something true about that town — actual prices, actual jobs we have done there, actual answers to the questions people ask before they book. If a page could have its town name swapped out and still make sense, it was not good enough.

3. Fixed the business information everywhere online

We went through the directory listings and made the name, address, and phone number match everywhere. This is boring, slow work. It is also the thing most owners skip, and nothing else lands until it is done.

4. Deleted the weak pages instead of adding more

The instinct is always to publish more. We did the opposite and removed the thin pages. A smaller set of strong pages beats a large pile of weak ones, every time.

Results

What actually changed.

Same window, one year apart — January 1 through July 31, pulled straight from GA4. The work landed in May 2026. Full windows, whole channels, no cherry-picked slice.

Organic Search sessions — year over year

January 1 – July 31, 20251,618
January 1 – July 31, 20264,877
1,618 4,877 organic sessions — 3.0x in one year. Average engagement time per organic session rose from 45s to 56s over the same windows.

Tracked click-to-call taps

January 1 – July 31, 2025 (untracked)(no tracking existed)0*
January 1 – July 31, 202681
Tracked click-to-call taps went from 0 to 81. The 0 is not zero calls — call tracking on the website was switched off before, so nobody was counting. The 81 is the first real count, not a jump. Calls are measured via the site’s ClickToCall event, live since the rebuild; no equivalent event existed on the old site. The 0 baseline means untracked, not zero.

Paid Search sessions — the deliberate step-down

January 1 – July 31, 202515,510
January 1 – July 31, 20263,544
Paid sessions went from 15,510 to 3,544 across the same windows. That was not decay — it was the controlled throttle described in the paid ads section below, done on our own revenue while the organic base was built underneath it.

AI search

AI search: what actually worked.

This is the part most owners are being sold badly right now. It is worth reading slowly.

People now ask ChatGPT, Copilot, and Perplexity for local recommendations. Google does it too, with AI Overviews — the written answer that shows up at the top of the results page before any of the blue links. Being the business the answer names now matters as much as being a link on the page.

So a wave of “AI SEO” started getting sold. Special files to add to your site. Code only AI can read. Rewriting your pages to sound machine-readable. This site had already added some of it.

Google publishes its own guidance on this, and it is direct: its AI features run on the same ranking and quality systems as regular search, and from Google’s side, optimizing for AI search is still just SEO. There is no separate playbook to buy. Read Google’s guidance yourself — it is short, and it is free.

So we removed the AI-specific tricks this site had added, and put the effort somewhere duller: direct answers to real questions, actual numbers, and clear signals about who the business is and where it works.

What we can and cannot show you

Here is the honest limit, and you should hold every agency to it. Most AI assistants do not pass clean referral data. That means nobody — not us, not anyone — can show you a precise “ChatGPT sent us this many visits” number. Any agency claiming otherwise is guessing and hoping you do not ask.

One small, early signal we can show: GA4 began attributing an “AI Assistant” channel this year, and it recorded 22 sessions in the 2026 window. The channel does not exist in our 2025 export at all. That is a small number and we are not extrapolating anything from it — it is here because it is real, measured, and new.

The best available stand-in is Bing. Bing is the index behind Copilot, so when Bing starts surfacing your pages, Copilot starts using them. That is the number we track — and it needs a deeper export than the ones this page is built on, so it is not here yet. We would rather publish this page with a gap in it than publish a number we cannot stand behind.

The work that gets you cited by AI is the same work that gets you ranked. Anyone selling you a separate AI package is selling you the same job twice.

Paid ads

What we did with the paid ads.

This is not a story about ads being bad. Paid ads were never the problem here.

The ads worked. They were running and they were producing volume. The problem was dependence. Paid placement was carrying the new-customer load close to on its own, because the search side had nothing underneath it — no trust, no consistent listings, no pages worth ranking. Every lead was rented. Stop paying, and it stops the same afternoon.

Two things were quietly broken

The website click-to-call conversion action was inactive. In plain terms: when someone tapped the phone number on the site and called, Google was never told it happened. Google’s automatic bidding decides where to spend based on what it can see converting. It could not see the main way these customers were actually converting. It was optimizing in the dark.

The account was also set to run on search partners — other websites Google shows your ads on, which are not Google search results. Those clicks look like search traffic on a report. They do not behave like it. How much of the budget was going there is a number we are still verifying, so we are not putting it on this page yet.

What we did about it

We throttled the paid spend deliberately while the organic side was built up. Not a shutdown — a controlled step down, on purpose, with a plan behind it. The session numbers are in the results section above. And we did it on our own revenue. We would not have tested this on a client first.

Where this is headed

The whole point of building the search side is that paid stops being life support and becomes a lever you choose to pull — busy season, a new service, a town you want to break into.

That is the goal for every account we take on. Paid ads running with working conversion tracking, pointed at people who are actually ready to buy, sitting on top of an organic base that keeps producing when the ads are switched off.

For your shop

What this means for your shop.

  • Depth beats volume. One real page beats ten copies of the same page with the town name swapped.
  • AI search and regular search are the same job. Do not pay twice for it.
  • Your business listings have to match everywhere before anything else you do will work.
  • This takes months, not weeks. The work happened in May. Search does not move on your timeline.
  • And if your phone is already ringing, that is not proof you are findable. It usually means you are living off repeat customers.

Want an honest read on where your business actually stands?

We will tell you what we would change, in what order, and what it costs. If we cannot help, we will say that too.

Or read the rest of our proof.