Case Studies / Roofing

One creative doing all the work. No system to replace it.

A US roofing company at $90k a month had one video doing almost all its productive spend across 11 states and five regions. The creative was aging. There was no pipeline to absorb growth. Building the system came before scaling a dollar.

Vertical

Roofing

Platform

One creative doing the work of nine had a ceiling that was architectural, not tactical.

The client had projected scaling to two to three times the initial monthly spend. That kind of growth requires a creative pipeline that can absorb increasing spend without frequency pushing CPL to unsustainable levels. One aging video cannot do that. When it fatigues, the account drops with it because there is nothing behind it.

The regional architecture had the same problem. Five distinct geographic markets, Mid-Atlantic, New England, Central Pennsylvania, Northern Virginia, and South Pennsylvania, were running inside a single shared campaign structure. Different performance profiles, identical setup. Each region's signal was masking the others.

CPL had moved from $133 to $109 in the early weeks of active management, before the structural work was complete. That progress was real but fragile. It sat on top of one video and one campaign. The five months that followed were about building the infrastructure to sustain $109 and scale from there.

Core constraint

"One creative doing the work of nine is a frequency ceiling waiting to happen. The ceiling is not the creative. It is what is missing behind it."

In roofing, the sales cycle from lead to closed install runs four to eight weeks. CoM% for any given month reflects leads from the prior month. Optimizing to short-term CoM in this vertical produces decisions that look correct but often are not. January's 26% reflected pre-restructure activity. February's 22% reflected large-scope project backlog. The real number was March and April at 15%.

Starting conditions

Cost per lead

$133 (then $109)

Active creative doing productive work

1 of several

Region-specific campaign architectures

1 (Mid-Atlantic only)

CoM%, January

26%

Creative variants available to scale into

None

The Approach

Understand why the winner works. Then build a system to produce more of them.

The winning video had a structural formula. The goal was not to replace it. It was to isolate which components were driving performance, so every subsequent variant was grounded in what had already converted rather than being a fresh guess.

Cost of marketing %, monthly trend

January and February reflect activity from before the restructure. In roofing, installs close 4-8 weeks after the lead. The CoM% you see today is the account from last month. March and April reflect leads generated under the new structure.

Deconstruct the winning formula into testable variables

The original video had four components: a spokesperson-led hook (shingle throw), a timelapse installation sequence, a benefits-and-authority middle (30 years, 200,000+ installs, lifetime warranty), and a direct CTA. Each component became a variable to test independently — new scripts on the same footage, same script on new drone sequences, hybrids swapping one element at a time. The purpose was diagnostic: confirm which component was driving performance so the next brief could build from that understanding rather than another iteration of the same concept.

Build a mega brief for 11-state coverage

Each market needed versions of the winning ad with state-specific callouts in both script voiceover and text overlay. A mega brief format produced one master template and nine regional variants structured so an editor could produce all of them efficiently. Treating each state as a separate production would have made this impossible to sustain at scale.

Give each region its own campaign architecture

Mid-Atlantic, New England, Central Pennsylvania, Northern Virginia, and South Pennsylvania each received a dedicated monthly budget and an ABO structure: one broad adset on top Post IDs, one region-specific adset on highest-performing zip codes by lifetime CPL and close rate. County-level CBOs were queued for Mid-Atlantic and New England once the all-zips data had built enough signal to trust.

Introduce the 3-requirements qualifier in tight-zip regions

New England and Central Pennsylvania had tight serviceable zip lists. Every out-of-area lead represented real sales team time with no revenue potential. The qualifier hook filters that out before the click. Geographic precision mattered most in these regions, so the 3-requirements format was introduced there specifically rather than across all five markets.

Creative

Nine variants. One formula. Every variable isolated to produce an insight.

The brief was structured so that each variant changed one thing. That isolation is what made the data from each test readable — not as a scaling strategy, but as a diagnostic one. The output wasn't nine more versions of the same ad. It was a confirmed understanding of which component was doing the work, so the next brief could depart from that formula with something genuinely different rather than a guess.

Each card below represents one variant tested. Cards showing an ad are ones where the asset is available to share. Cards without a visual cover concepts where the footage isn't available here, but the strategic logic and result still stand.

Control

Original winner (0002)

Spokesperson throws a shingle toward the camera. Into timelapse installation. 30-year authority close. 50%-off CTA. The formula every variant was built from. Shortened by 1-2 seconds to improve hook-to-content transition.

Baseline for all nine variants

Key insight

The message changed the entry point. The formula held.

Three scripts ran on identical footage — social proof, urgency, ease of installation. The visual never changed. What changed was who leaned in first.

Winner

Script variants (x3)

Three new scripts on the same footage. Script 1: social proof open. Script 2: property value framing. Script 3: 200,000+ installs authority open. Same visual, different entry point.

Deployed in 1:1 and 9:16 across all regions

Winner

New drone footage sequences (x2)

Two new footage sequences from the client's archive. Fast-paced aerial installation shots, timelapse from full-roof perspective, finish shots. Same script and voiceover as the control. Testing whether visual freshness alone extended creative lifespan.

Sped up x2-x3 to maintain frame-to-frame engagement

Learning

State callout versions

Region-specific callouts in voiceover and text overlay, produced from the master template across all 11 states. Location relevance pre-qualifies geographic intent before the click. Applied across all five regions with varying emphasis by market density.

Applied across all five regions

Learning

3-requirements video (NE and C. Pennsylvania)

Three requirements to qualify: US resident, homeowner, roof older than a defined threshold. Introduced in tight-zip regions where out-of-area leads were costing the sales team real time with no revenue potential. Qualifier format pre-filters before the click.

Tight zip regions only. Set rate improvement was the target.

Tested

Hook swap (spokesperson change)

Replaced the shingle-throw hook with a different spokesperson open from the client's existing content library. Testing whether the hook was the primary driver of the original's performance, or whether the rest of the formula would hold with a different entry point.

Testing hook vs. formula as the key performance driver

Results

The creative system sustained $240k. One video never could have.

The account scaled from $90k to $240k over five months without frequency becoming a ceiling. That is not a CPL story. It is a creative operations story. The CoM% settled at 15% in March and April once the pipeline normalized under the new structure. Set rates in New England and Mid-Atlantic held above 27%, above the 20-25% roofing benchmark on Meta.

$90k → $240k

monthly spend

Scaled over 5 months with a continuously growing creative library giving the algorithm fresh signals as budget increased

15%

Cost of marketing

March and April. Down from 26% in January. The gap reflects the 6-8 week sales cycle, not creative underperformance in Jan-Feb.

27%+

Set rate

New England and Mid-Atlantic. Roofing benchmark on Meta typically runs 20-25%. Regional architecture and qualifying creative drove the improvement.

Variants produced

From one proven formula, each isolating a specific variable. The isolation confirmed which element was driving performance — that knowledge is what briefs new, conceptually distinct creative rather than more variants of the same concept.

Reading CoM% correctly in a long-cycle vertical

The figure you see in any given month reflects leads generated 6-8 weeks earlier. January and February look elevated because large-scope installs from the pre-restructure period were still closing. Optimizing to monthly CoM in roofing produces decisions that look coherent at the top of the funnel but are wrong for the business.

What I Took From This

Scale reveals the gaps in your creative system.

A winning creative is a starting point, not a system. The goal of early account management is to understand why the winner works structurally, then build a methodology that produces more of them. Isolating individual variables across multiple variants is a diagnostic tool, not a production model. It answers a specific question: which element was doing the work? Once you have that answer, the next brief shouldn't produce more iterations of the same concept — it should use that confirmed insight to brief something conceptually distinct that opens new audience territory. The isolation produces the knowledge. Coverage is what you do with it.

Multi-region accounts need region-level architecture before any scaling begins. Running Mid-Atlantic and New England inside the same broad campaigns, despite meaningfully different performance profiles, was the single biggest structural limitation on scale. Separating them made the account simpler to manage, not more complex, because each region's performance became readable on its own terms rather than masked by the others.

Roofing CoM% must be read in trailing 60-90 day windows. Optimizing to monthly CoM in an account with a 6-8 week sales cycle produces decisions that look coherent at the top of the funnel but are wrong for the business. The January 26% was not a problem to fix. It was a description of what happened before the restructure. Patience in reading the data is a strategic input, not a passive posture.

Variable isolation: how each variant was built

Before the work

One video carrying all productive spend across 11 states, aging with no replacement pipeline

Five regions sharing one campaign structure despite different performance profiles

No region-specific callouts in creative across a five-region geographic footprint

Scaling ambition with no creative architecture to absorb the spend without frequency becoming a ceiling

CoM% read monthly in a vertical where the relevant window is 60-90 days trailing

After the work

Nine video variants from one formula, each isolating a specific variable — not to scale all nine, but to identify which element was driving performance and brief from confirmed insight

Five region-level CBOs, each with its own monthly budget, adsets, and performance baseline

Region-specific callouts in voiceover and text overlay across all five regions

Creative library expanding monthly, enabling scale from $90k to $240k in spend without frequency becoming the ceiling

CoM% evaluated in trailing 60-90 day windows with lagged project context accounted for explicitly

→

How to extend a winning creative before it becomes account dependency

The Creative Lifecycle playbook explains how to preserve a proven signal while expanding it into new angles, hooks, formats, and creators before saturation turns into decay. Read it →

→

Why region-level architecture matters before scaling spend

The Creative Coverage playbook covers the three-campaign model, how to separate testing from scaling, and when to give each region its own campaign infrastructure rather than running a shared structure that masks performance differences between geographies. Read it →

→

The Creative System — how coverage, hooks, and lifecycle connect

This case ran through all three layers: coverage to build the pipeline, hooks to qualify the right buyer in each region, and lifecycle to extend the winning signal before it fatigued. The Creative System hub maps how the three playbooks work together. Read it →