Case studies / McKellips & Sons
McKellips & Sons Case Study
When Google Trusted a Local Furnace Company Over the Furnace Manufacturer
Over the course of a 90-day sprint, Rygo Labs built the diagnostic content structure and information architecture behind McKellips's furnace, AC, and installation pages from the ground up. The client brought management in-house shortly after, making this case study a unique look at a complete build compressed into a single quarter, without a multi-year optimization arc or ongoing retainer. Over a year later, the results have held up completely untouched, proving the long-term durability of GEO (Generative Engine Orientation) frameworks in AI search.
Engagement Length
3 months
Type "what causes a furnace to blow cold air in Michigan" into Google, and the AI Overview doesn't send you to Carrier , the company that built the furnace. It sends you to McKellips & Sons, a West Michigan HVAC company, and lists Carrier's own content below it in the source panel.
That's not a fluke result. It's one of twelve queries in the furnace, AC, and installation funnel where McKellips is now the source Google's AI Overview reaches for first, ahead of national manufacturers, regional competitors, and even a Reddit thread with years of upvotes behind it.
This case study is about how three service pages got there in three months.
The Problem Wasn't the Pages. It Was the Order of the Questions.
McKellips already had a furnace-repair page before this project started. It wasn't thin, and it wasn't badly written.
A homeowner standing next to a furnace that won't heat doesn't ask one question. They ask a sequence, each one ruling something out:
- Why is my furnace blowing cold air?
- Is this the thermostat, or something mechanical?
- Could a dirty filter really cause this?
- Why did it start warm and then go cold?
- Is it the pilot light or the ignitor?
- Do I keep resetting it, or call someone?
- Is this a repair, or do I need a new furnace?
We rebuilt the page around that sequence, not around "furnace repair" as a single topic. Each question became its own section, in the homeowner's own phrasing, answered directly before any mention of McKellips as a business.
Google isn't matching one page to one query. It's matching individual sections of one page to seven or eight separate queries, because each section was built to stand alone as the answer.
We call this the Precision Search Magnet framework, or PSM: a specific problem, for a specific person, in a specific place, answered before anything else. This is the same framework Rygo Labs builds into every diagnostic-heavy service account: map the real question sequence a buyer walks through before they call, then structure the page so each section can be lifted and cited on its own. It isn't specific to furnaces. It's specific to any business competing against a bigger, more generic name in a moment where the buyer is scared and needs an answer fast.
The Proof: Where McKellips Is Winning, and Against Whom
In three months, McKellips's furnace page, AC-repair page, and installation page went from no AI Overview presence to the cited source across nearly the entire diagnostic funnel for furnace and AC problems in West Michigan:
Furnace Repair Page AI Wins
AC Repair Page AI Wins
fthsdfsfd
HVAC Installation Page AI Wins
This isn't a claim of a perfect sweep, it's eleven wins and one close miss, and the miss shows exactly where the next round of work should go: Moore Mechanical currently has a more direct answer to that specific failure mode, and that's the next section to rebuild.
Selling against a bigger name in your market? [See what a diagnostic content audit would find in yours.]
The Page That Proves It's the Structure, Not the Site
There's a fourth page in this story (as well as an entire build that never materialized), and there is a visible difference between what we provide, and how our copy works over time.
McKellips's indoor air quality page was next on the build plan. It's the one page in this cluster that was never rebuilt around our PSM framework. This page was built around classic SEO structure and best practices from previous years.
It doesn't show up in a single AI Overview for indoor air quality queries in West Michigan. That’s the power of the AI panel and where Google is moving.
Think of it like a controlled laboratory experiment. If you run two identical samples under the exact same conditions and change only a single variable, you know with certainty which variable caused the result.
That’s the exact environment we had here. McKellips had four primary service pages on the exact same domain. Three were rebuilt using our PSM question-sequence framework, while the fourth (Indoor Air Quality) was left untouched on traditional SEO copy. Same domain, same authority, same brand signals. The three restructured pages immediately swept Google's AI Overviews, while the control page captured zero.
So we don't have to guess why it worked. We can see it.
Same domain. Same business. Same brand signals Carrier and Trane were up against.
Why the Manufacturer Lost This Particular Fight
Carrier is one of the largest furnace manufacturers in North America, likely the actual company that built the furnace in the reader's basement. Its furnace-repair page is titled "Why is My Furnace Blowing Cold Air?", almost the exact same question McKellips answers under "What Causes a Furnace to Blow Cold Air?" Same header. What's underneath it is where they split. Carrier lists ten causes that would read identically in Phoenix or Grand Rapids, because the content has to work everywhere. McKellips's answer names the exact regional failure mode, frozen exterior intake and exhaust pipes, a problem specific to Michigan winters that a Texas-facing manufacturer page has no reason to mention in detail.
When Google's model is choosing between a technically correct but generic answer and a technically correct answer with the missing regional variable filled in, it's picking the one that resolves more of the reader's actual uncertainty. That's the mechanism, not a ranking trick: McKellips's page answers a question Carrier's page doesn't know to ask.
Trane, another major furnace manufacturer, loses the DIY-troubleshooting query a different way. Its troubleshooting guide never actually organizes around the homeowner's real question, there's no header asking "should I try troubleshooting my furnace myself," just a general checklist and a blanket warning about the danger of DIY repair. McKellips's page has that exact question as its own section, and draws the specific, actionable line, which checks are safe to do yourself, which ones involve gas or wiring and need a licensed tech, and that's the distinction Google's AI Overview repeats back almost verbatim.
Google isn't just crawling a page, it's building what we call a Character Profile of the business behind it: who McKellips serves, where, and how they solve problems, assembled from every page and every consistent answer. Carrier and Trane have profiles too, but theirs are built for every furnace owner in North America. McKellips's is built for West Michigan specifically, which is the only version of authority that actually matches someone searching "furnace blowing cold air in Michigan."
None of this required a secret header trick. Carrier used almost the same question McKellips did and still lost, because the advantage was never the header.
From Diagnosis to Decision, Without Losing the Reader
The furnace page explains what's wrong. The installation page answers what to do about it. On their own, that's two pages. Together, they're one funnel: a homeowner who's just ruled out the easy fixes on the furnace page lands next on the exact next question they're asking, whether this is worth fixing at all.
McKellips ranks first among eleven sources for "should you repair or replace your HVAC system," ahead of Moore Mechanical's own piece built specifically to answer that question. That's the same diagnostic structure carried across the full decision, not two pages that happen to both rank. It's why the informational page turns into a sales conversation before the homeowner ever picks up the phone.
Built in a Quarter, Still Working a Year Later
These three pages went live in the final months of 2025. Nobody at Rygo Labs has touched them since the account went in-house shortly after.
That's the argument for organic content as an asset instead of an expense. A Google ad stops working the day you stop paying for it. A page built around the right question sequence keeps answering those questions for as long as people keep asking them.
What This Means Beyond McKellips
If your business sells against manufacturers or national chains, the takeaway isn't "write more content." It's that the manufacturer's disadvantage is structural, and it's the same disadvantage every national brand has against a business that can lean into local signals.
That gap doesn't close with volume. A hundred generic pages don't outrank one page that answers the specific version of the question your market actually asks. McKellips didn't outspend Carrier. It answered a question Carrier's content wasn't built to answer.
If you're in a regulated or manufacturer-adjacent local service and you're not showing up in the AI Overview for the diagnostic questions your customers are actually asking, that's the gap worth checking first , not your ad budget, not your backlink count.