Most agencies sell traffic because traffic is easy to count. Conversion is harder to move and harder to fake. This case study uses Sunstone Digital Tech’s published Cody Pools results to show what a conversion-first engagement looks like, how the experiments are structured, and what evidence you should ask any CRO agency to produce before you sign.
A note on sourcing: the figures below come from Sunstone’s published case study cards and company profile. Per-test baselines, variant screenshots, and exact test dates sit in client project files and aren’t published on those cards. Where this article gives worked numbers, they’re labeled as arithmetic examples, not client data.
Cody Pools: a conversion problem, not a traffic problem
Cody Pools is a pool builder, a category where the sale is high-ticket, research-heavy, and local. A visitor rarely buys on the first session. The conversion event is a lead: a form submission, a call, or a consultation request. Everything upstream (rankings, Maps visibility, paid clicks) only matters if a meaningful share of those sessions turns into a qualified conversation.
That makes the goal a conversion-rate goal. More sessions at a flat conversion rate raise cost per lead linearly with spend. A higher conversion rate lowers cost per lead on the traffic you already pay for or earn. Sunstone frames its conversion rate optimization work the same way: A/B testing and behavioral analysis to turn existing visitors into revenue without increasing spend.
The published outcome for the engagement:
- 312% more qualified leads
- 189% higher conversion rates
- $3.17M in revenue
What the headline numbers say when you decompose them
A 189% higher conversion rate means the post-optimization rate is 2.89x the baseline. As arithmetic only: a site converting at 2.0% before would convert at about 5.8% after.
A 312% increase in qualified leads is 4.12x. If traffic had stayed flat, a 2.89x conversion rate would yield 2.89x leads. The gap between 4.12x and 2.89x is roughly 1.43x, which points to added volume or better lead mix contributing alongside the conversion gains. That fits Sunstone’s positioning of CRO as one part of a system with local and organic visibility. More visibility feeds the funnel, and a better-converting site keeps more of what arrives.
This decomposition is an inference from two published figures, not a reported attribution split. It still gives you a useful habit: when an agency quotes a lead lift, ask how much came from rate and how much from volume.
Baseline metrics and the analytics used to diagnose friction
Before any variant is built, a CRO engagement needs a measurement baseline that holds up. For a local service or home-improvement business, that baseline typically includes:
- Session-to-lead conversion rate, segmented by device, traffic source, and landing page
- Form start rate versus form completion rate (the gap isolates form friction)
- Click-to-call events, tracked as separate conversions from form fills
- Qualified lead rate, meaning the share of leads that pass the client’s own qualification criteria
- Revenue or pipeline value per lead, so a test can’t “win” by inflating junk submissions
Instrumentation matters as much as the metrics. Conversion events are configured in Google Analytics (GA4) and checked against the CRM or call-tracking records so form fills and calls reconcile. A baseline window covering several full weeks smooths weekday and weekend swings, and it should avoid overlapping a promotion or seasonal spike.
Behavioral analysis inputs
Quantitative funnel data tells you where visitors leave. Behavioral tools suggest why. Heatmaps, scroll-depth maps, and session recordings surface patterns such as:
- Visitors scrolling past the hero without engaging because the offer isn’t clear above the fold
- Taps on non-clickable elements on mobile
- Form abandonment at a specific field (phone number and project-budget fields are common culprits)
- Message mismatch between the ad or search listing and the landing page headline
- Slow mobile load times on image-heavy pages
Each finding becomes a candidate hypothesis, ranked by expected impact, confidence, and implementation effort. The structure of a high converting website (clear offer, fast pages, visible calls to action) sets the checklist the diagnosis starts from.
How hypotheses and test variants are structured
A usable test record reads like a lab note, not a design opinion. The format Sunstone’s approach calls for has six parts:
- Observation. What the behavioral and funnel data showed.
- Hypothesis. A falsifiable statement: “Reducing the quote form from N fields to the minimum needed for qualification will raise completion rate without lowering qualified lead rate.”
- Variants. Control (A) is the live page. Variant B changes one defined element or one coherent group of elements. A third variant is added only when traffic can support it.
- Primary and guardrail metrics. Primary is conversion rate to qualified lead. Guardrails include bounce rate, page speed, and lead quality.
- Sample size and duration. Fixed before launch.
- Decision rule. What result ships, what gets iterated, and what gets discarded.
Illustrative hypotheses for a local service landing page
These are representative of the test types used on lead-generation pages for local service businesses. They aren’t a list of Cody Pools’ individual results.
| Element | Control (A) | Variant (B) | Behavioral signal that motivates it |
|---|---|---|---|
| Hero section | Brand-led headline, generic CTA | Service-and-location headline, specific CTA | Low scroll-to-form rate |
| Quote form | Long form, many required fields | Shorter first step, progressive follow-up | High form-start, low form-completion |
| Trust signals | Reviews in footer | Review rating and project proof near the CTA | Hesitation before the form |
| Mobile CTA | Inline button only | Sticky click-to-call and quote buttons | Mobile tap-through lower than desktop |
| Service-page structure | Long narrative copy | Scannable sections, offer first | Short dwell time, early exits |
Statistical guardrails
Fixing sample size in advance prevents calling a winner on noise. A common rule of thumb (Lehr’s approximation, roughly 80% power at a 5% significance level) gives the visitors needed per variant as about 16 × p(1−p) ÷ δ², where p is the baseline rate and δ is the absolute lift you want to detect.
As arithmetic only: at a 3% baseline, detecting a 20% relative lift (0.6 percentage points) needs about 12,900 visitors per variant. A page with 2,000 monthly sessions can’t resolve that in a reasonable window, which is why low-traffic pages are often tested on bigger changes, or optimized by sequential before/after review with explicit caveats.
Other checks include verifying the traffic split is even (a sample ratio mismatch signals a tracking or bucketing bug), running tests through whole weekly cycles, and keeping UTM parameters and attribution consistent between variants so channel mix doesn’t skew the comparison.
Rollout, QA, and the retesting loop
A winning variant isn’t finished when the dashboard turns green. The rollout stage covers:
- QA across devices and browsers. Checking layout, form validation, and event firing on iOS Safari, Chrome on Android, and desktop browsers.
- Tracking verification. Confirming conversion events still fire in GA4 and that call-tracking numbers still swap correctly.
- Regression monitoring. Watching page speed, bounce rate, and lead quality for several weeks after launch to catch novelty effects fading.
- Backlog update. Each result feeds the next round. A form-length win, for example, raises follow-up questions about which fields to restore later in the funnel.
This is also where CRO connects to the rest of the stack. Local SEO and Google Maps work raise the number of high-intent visitors. Content and service-page work decide what those visitors read. Tested layouts then feed back into a conversion focused website template that can be reused across additional markets. Window World (187% organic traffic growth, 164% more organic leads) and Groundworks (264% more organic traffic, 198% more leads across 30+ major markets) show the visibility side of that system. Cody Pools shows what the conversion side adds.
Why a conversion-first agency is easier to evaluate
Directory listings and “best CRO agency” roundups rarely show the mechanics behind a lift. If you’re comparing agencies, ask for the following and treat missing answers as information:
- The baseline conversion rate and the date range it covers
- The hypothesis, variant description, and test duration for each claimed win
- The sample size and the significance or confidence threshold used
- How lead quality was protected (guardrail metrics, CRM reconciliation)
- Whether the reported lift is rate-driven, volume-driven, or both
- What happened to the result after rollout
Sunstone’s published proof points give you a starting set to test against: a 4.9/5 Google rating, more than 2,500 clients served, 6,000+ projects completed since 2018, and a homepage claim of a 73% lead increase in 60 days. Those are company-level figures, so ask for the project-level detail behind them during a proposal. For a broader scoring approach that covers process, conversion focus, and proof, see how to choose a web design agency.
What the Cody Pools engagement teaches
Four lessons carry over to most local service businesses.
Measure qualified leads, not form fills. A shorter form can raise submissions and lower quality at the same time. Cody Pools’ headline metric is qualified leads for exactly this reason.
Test where the data points. Heatmaps and funnel drop-off rank the backlog. Opinions about button color don’t.
Protect the test with arithmetic. Decide sample size and duration before launch, or you’ll end up rationalizing whichever variant looks ahead on day four.
Pair conversion work with visibility work. A 2.89x rate on a page that nobody finds is a small win. More qualified traffic into a high-converting page compounds both effects.
Client perspective
The published case study card reports outcomes, not a verbatim stakeholder quote, so none is reproduced here. When you speak with a prospective agency’s references, ask them three things: which specific tests changed their numbers, how the agency reported results, and whether they could trace revenue back to a change on the site. A client who can answer those has seen the process, not just the invoice.
Get a conversion audit before buying more traffic
If your pages already get visits and the leads aren’t following, the bottleneck is probably on-site. Sunstone Digital Tech starts with your baseline analytics, behavioral data, and lead-quality criteria, then proposes a prioritized test plan with defined hypotheses and success metrics. Request a free proposal and ask to see the test-record format described above.
