A structured framework
for turning page audits into
product decisions.
A repeatable product intelligence system, weekly design opportunity audits, persona-mapped observations, benchmark analysis, structured recommendations, CRO validation plans, and production-ready Jira tickets, applied across four affiliate platforms.
What this is
Most design research is reactive, a problem is flagged, someone investigates it. This body of work is different: a proactive, structured programme of weekly page audits applied across the Catena product portfolio, each following the same repeatable framework regardless of which product or page is being reviewed.
Each audit produces a complete design opportunity document: the page purpose, the likely user personas, the primary opportunity, the observation, why it matters commercially, a benchmark reference, a formal recommendation, the CRO data needed to validate it, a suggested Jira ticket with acceptance criteria, and a priority/effort assessment. These documents are not design artefacts, they are product strategy inputs that can be actioned by a PM, a developer, or a CRO team without additional explanation.
"The goal is not to identify design problems. It is to identify where a specific change would produce a specific commercial outcome, and to document that clearly enough that it can be built."
The framework
Every audit follows the same structure. Page purpose establishes what the page is supposed to do commercially. Persona framework identifies which user types are affected and how. The primary opportunity names a single specific thing that could be improved and why. The observation is factual, what is visible on the page, not what should be done about it. The why it matters section connects the observation to commercial impact: not "the UX is poor" but "this creates avoidable friction for first-time evaluators, which is a hypothesis that clearer pre-click qualification could improve qualified outbound clicks." The benchmark reference grounds the recommendation in how comparable pages solve the same problem. The recommendation is specific and prescriptive, not a list of possibilities. CRO data to validate defines exactly what to measure and how to confirm the hypothesis. And the Jira ticket translates the recommendation into acceptance criteria that engineering can implement and QA can test.
This framework was designed to close the gap between design thinking and product delivery, making it impossible for a recommendation to sit in a document without a clear path to implementation.
Bonus.com Online Casinos audit
The page headline promised "Best Online Casinos That Pay Real Money" but the opening ranked offers were sweepstakes casinos, LoneStar Casino, Crown Coins, WOW Vegas. Users had to infer they were not viewing regulated real-money options because the distinction wasn't made before the first CTA. The recommendation was a top-of-page classification module routing users into two clearly labelled paths: regulated real-money casinos, or sweepstakes alternatives. Sweepstakes offers must never present as the default real-money answer without explicit labelling.
Bonus.com Sweepstakes Casinos audit
The page opened with strong editorial trust signals but the initial cards were repetitive and offer-led, without explaining why each ranked choice was best for a specific need. One listing showed a lower rating than several lower-ranked listings with no visible rationale, a trust signal that works against confidence. The recommendation: a "Best for" label on each top card, 2-3 normalised comparison fields, a one-line ranking rationale where rank and rating appear to conflict, and a "Compare top picks" jump link to a condensed comparison table.
Eligibility before rankings
The page led with a long ranked list before answering two questions every user arrived with: "Is this legal in my state?" and "Is this play-for-fun only or prize-redeemable?" The recommendation was a compact eligibility module before the rankings, with CTA language defaulting to "Check availability" until state context was confirmed. Benchmark: Casinos.com and ATS bring eligibility into dedicated sections before comparisons. Priority: High. Effort: Medium.
PlayUSA Online Casinos audit
The page headline promised "Best Online Casinos That Pay Real Money" but the opening ranked offers were sweepstakes casinos. The real-money vs. sweepstakes explainer appeared much further down. The recommendation was a state-aware branching module before any rankings that routes users to regulated real-money options or clearly-labelled sweepstakes alternatives. Sweepstakes offers must never present as the default real-money answer without explicit labelling.
State Gambling Guide framework
PlayPennsylvania served breadth before decision structure, many modules competing at once rather than a clear comparison path. The recommendation was a reusable comparison-first module for state guide templates: text-only, two colours (default and dynamic status accent), dynamically adapting to selected category. No images, no gradients. The constraint was deliberate: visual simplicity forces the comparison structure to do the work.
PlayNJ intent-led comparison
The top list presented similar bonus cards without helping users understand which operator was best for their specific intent. The recommendation: 4-6 intent chips (Best bonus, Fast payouts, Best app, Low deposit, Live dealer, Loyalty rewards) that re-order or highlight relevant operators and expose 2-3 comparison attributes per card. Intent selection must not hide compliance text or bonus terms. Full tracking spec included for intent selection, CTA clicks, and promo-code copy actions.
Each of these four audits was produced proactively, not in response to a brief, but as part of a weekly research cadence. The consistent structure means any PM or CRO partner can read one audit and immediately know what to do next, without a design presentation or a follow-up meeting. This is design operating as a product intelligence function, not a service function.
CRO experiment analysis
Alongside the opportunity audits, a parallel stream of UX test proposals was produced, each analysing an existing A/B experiment result and proposing what the next test should be and why.
The sticky CTA plugin experiment had reported a pass with +29.5% session ALC and +19.4% conversion rate uplift. The analysis flagged that the documentation did not isolate the UX drivers behind the improvement, whether the uplift came from persistent CTA visibility, stronger brand salience from the featured operator, or reduced scroll depth required to reach a conversion action. The recommendation was to validate each driver independently rather than treating the pass as a design conclusion. Without knowing which factor drove the result, the next iteration has no informed direction.
The sticky CTA banner position experiment (moved from top to bottom of viewport) had resulted in -9.7% session ALC and -7.6% conversion. The experiment was marked neutral with a retest recommendation. The analysis identified a confounding variable: the measurement period coincided with a major NFL playoffs weekend, which distorts baseline traffic quality, user intent, and conversion propensity. The recommendation was a properly controlled experiment with a pre-specified baseline period that avoids high-intent sports events.
Dynamic contrast engine, accessibility at system level
The dynamic text contrast validation document represents the most governance-forward work in the entire body of research. Regional sites across the Catena portfolio were rendering white text on light backgrounds and black text on dark backgrounds, not because designers made poor choices, but because CMS-driven content changes background colours dynamically, and no system existed to detect when text contrast had fallen below accessible thresholds.
The proposed solution was a rule-based contrast engine that automatically determines the correct text colour (black or white) based on the background at render time. The objective was not to fix individual instances but to make the problem impossible at the infrastructure level, so that marketing teams could inject new content without engineering oversight and still produce accessible output. This is accessibility governance enforced by system logic, not by review process.
Outcome
Qualitative assessment of the work.
The Product Intelligence Framework is designed to turn design research from a reactive activity into a proactive operational function. Without it, problems tend to surface when they become visible: through traffic drops, stakeholder complaints or conversion declines. With it, they are identified structurally, before they reach that threshold.
The framework's main purpose is reducing the friction between design thinking and product delivery. By structuring every audit to end with a production-ready Jira ticket, including acceptance criteria, priority assessment, and success metrics, the work could move from observation to implementation without requiring a designer in every subsequent meeting. This is what design operations maturity looks like in practice.
The CRO experiment analyses applied hypothesis precision rather than result optimism. The analysis flagged an NFL playoffs confound in the sticky banner test, so a flawed conclusion would not be treated as a validated direction. That kind of rigour, treating a positive result with the same scrutiny as a negative one, is rarer than it should be, and more valuable than most organisations recognise.