How it works
How OrderLeak turns your pages into an action plan.
Three steps, a fixed method, a structured report at the end. Here's precisely what happens between the moment you point us to your pages and the moment you receive your recommendations.
01You point us to your pages
Paste the URLs of the pages to analyze: homepage, product page, collection page. The cart and checkout can't be reached by an address: add full-page screenshots of them instead, and they count among the included pages. One credit covers up to 4 pages of the same site; beyond that, one extra credit per further 4. Everything stays in your history.
Only public pages are analyzed: the ones any visitor can already see. No connection to your back office, no access to your customer or sales data.
02We capture and analyze
Each page is captured full-page, desktop and mobile, separately. The mobile experience is never inferred from desktop. Each capture is then evaluated against a fixed method across six axes: value proposition, calls to action, trust and social proof, friction, visual hierarchy, mobile UX.
03You receive a structured report
The report is organized page by page, desktop and mobile separated, ready to share with your team or your contractors.
The method
A fixed grid, not a case-by-case opinion.
OrderLeak doesn't improvise different criteria for each page: the six axes and their scoring grid are fixed. What changes from one report to the next is what the analysis finds on YOUR page, never the method itself. That's what guarantees a consistent level of analysis, whether you're auditing one site or several.
Value proposition
If a visitor doesn't understand what they gain in 3 seconds, they leave. We check that they stay.
What is sold, to whom, and why buy here rather than elsewhere. The analysis reads the main headline, the subheading, the visual and the whole first screen, on desktop and mobile alike, and checks that the promise is clear without scrolling. The page's code is read alongside: the number of main headings and their text, the page title and description. A first screen that only gives the brand name, or a promise any store could sign, is flagged with the exact sentence at fault.
Calls to action
A button no one sees is a button no one clicks. We track every CTA that fades into the background.
Clarity, visibility, contrast, wording and placement of the buttons that move a purchase forward: add to cart, buy, order. The question is simple: can the button be reached without hunting for it? The code confirms what the screenshot shows: purchase wording present in the page, an add-to-cart form, buttons with no name a screen reader can read. On mobile, the analysis first checks whether the main action fits on the first screen.
Trust and social proof
Before buying, people doubt. We check that reviews, guarantees and trust signals show up at the right moment.
Reviews and ratings, guarantees, returns and shipping, secure payment, transparent pricing, real photos, ways to get in touch. The code is searched for what a screenshot can miss: free-shipping or returns wording, payment logos (Visa, PayPal, Apple Pay, Klarna…), review badges (Trustpilot, Trusted Shops…), structured product data with a rating and review count. A signal present in the code but invisible on screen is reported as such, not counted as a given.
Friction
Every extra field, every extra step loses buyers who were ready to pay. We count them.
Anything that slows a buyer down or makes them hesitate: a cluttered page, too many choices, pop-ups, confusing navigation, hidden search, long forms, missing information at the moment of decision. The code counts the fields in the largest form, spots a missing search field, fields with no label or no autofill, email and phone fields typed as plain text, and the third-party scripts the page loads.
Visual hierarchy
The eye follows a path. If it gets lost, so does the sale. We spot every detour that leads it astray.
How layout, typography, spacing, colour and contrast guide the eye: is there a clear reading path from the promise to the button? The analysis flags whatever pulls attention away from the main action: several buttons of equal weight competing, a decisive element lost in the page, an image that fills the screen and says nothing. The code checks the heading structure and skipped levels.
Mobile UX
Mobile forgives nothing: small screen, hurried thumb, short patience. That's often where conversion quietly breaks down.
Assessed on the mobile capture, at 390 × 844, a typical smartphone screen: tap-target size, thumb reach, legible text, what fits above the fold on a tall narrow screen, a persistent buy button, no horizontal scrolling. The code checks the mobile viewport tag and flags blocked zoom. This axis does not count towards the desktop score.
The score
How the score is calculated.
- 01
Each axis is scored from 0 to 100 against real stores, not an ideal page: 85 and above is a reference in its sector; below 35, the page is broken on that point. The number of findings does not set the score, and something that is simply absent is not penalised for its absence.
- 02
A device's score is the mean of its axes. The global score weights mobile and desktop by each one's real share of your store's traffic, as measured by Chrome; without that data, each counts for half. The report always states which weighting it applied.
- 03
A finding's severity is capped by the analysis's confidence: an uncertain finding cannot be presented as critical. Findings that contradict what the code measured are removed before delivery.
- 04
The estimated gain combines findings with decreasing weights instead of adding them up, and stays capped at +25%. It only becomes an amount with your own traffic, conversion and average-order figures.
The deliverable
The formats
Global and per-axis scores
An immediate read on performance, from 0 to 100, broken down by category and device.
Prioritized findings
Each issue is qualified by its severity, its exact location and a concrete recommendation.
Desktop and mobile
Two distinct analyses per page: the mobile experience is never treated as a scaled-down desktop.
Measured technical signals
Response time, page weight, images without alt text, structured data, trust cues: read from your page's own code, verifiable.
Revenue projection
The estimated gain is computed from your visitors, your conversion rate and your average order value. Without those figures it is not shown.
Sized experiments
For every genuinely uncertain call: the hypothesis, the metric, and the visitors per variant needed to settle it.
Polished PDF export
A clean deliverable, ready to share with your team or your contractors.
A score benchmark will join the report once our corpus of real audits is large enough to mean anything: the observed distribution of our own audits, across all sectors, stated with its sample size. It is not a published sector average, and while the sample is too small the report shows nothing rather than a figure nobody should act on.