How exactly we examine your e-shop
The internet is full of tools that spit out a thousand “errors”. We built EshopMedic on the opposite principle: every finding must be defensible — when we are not sure, we stay silent. Here is the whole methodology, no secrets.
The nine areas the check-up scores
You get each one as its own number — 100 = no findings, higher is better — and can open the concrete findings behind it. The shop’s overall score is their summary.
Whether people find you on Google — titles, meta descriptions, H1, canonical, structured data, social previews.
Product descriptions, image alt texts, prices inside articles, expired promotions, empty and duplicate copy.
We load every link and image: dead links, broken and oversized images, error pages, SSL and domain expiry.
Google PageSpeed for mobile and desktop plus real visitor data (Chrome UX) when you have enough traffic.
Omnibus on discounts, company ID, terms, complaints procedure, health claims under EU 1924/2006, cookie bar.
Whether ChatGPT and friends can see you: robots.txt, content readable without JavaScript, schema.org Product/Offer, company profile.
Once Search Console is connected: which queries you appear for and where the quick wins are — positions 11–20.
SPF, DKIM and DMARC — so order confirmations do not land in your customers’ spam.
We walk the path from product to payment and report where it breaks.
Four rules every check follows
1. We measure what the customer sees — not what a robot sees
Our scanner presents itself as a regular browser, including the headers a server uses to decide image formats. It sounds like a detail, but it is the difference between truth and a false alarm.
2. When we are not sure, we stay silent
An ambiguous finding gets dropped. A false alarm is worse than a gap in the report: it costs your time, erodes trust, and next time you will not even open the report. That is why, for example, we only report a price next to a product link when we can attribute it unambiguously — otherwise nothing.
3. Every limit is visible
When capacity lets us test 500 of 543 links, the report says so: “your plan does not cover the rest”. We never pretend to have checked everything when we checked a part. Silent caps are common in this industry — and a check-up that hides its own scope is useless.
4. We say who can fix each finding
The report splits findings into those you fix yourself in the administration and those caused by the template or the platform — for these you get a ready-made message for support, just copy and send. And if we upgrade the scanner between your two scans, the report says so: new findings from new checks do not mean your site got worse.
What we check
| Area | What exactly we test |
|---|---|
| Availability | We load every discovered page and verify it responds. We distinguish a real error from the platform’s anti-bot protection. |
| Links | We test all of your own links; for links to other sites we only report unambiguously dead targets (404/410) — social networks redirect and rate-limit robots, and that is not your fault. Links with technically invalid format (spaces, double protocols) are reported separately. |
| Images | We load every image. Broken ones, oversized ones (over 500 kB as they actually reach the browser) and images without alt text — knowing that an empty alt="" is the standard way of marking a decorative image, not an error. |
| Prices | On product pages we read the real price (schema.org). Then we go through articles and other pages looking for prices written next to product links — if they differ from the price list, we report it. Nobody remembers to fix prices inside articles after a price change. |
| Law & ads | Omnibus (lowest price in 30 days for discounts), mandatory information, health claims under EU Regulation 1924/2006, expired promotions, cookie banner. Exactly the things Google suspends Merchant Center accounts for and consumer authorities fine. |
| SEO | Titles (measured the way Google truncates them), meta descriptions, H1, canonical, structured data, page language, social previews, noindex. |
| Speed | Google PageSpeed for mobile and desktop, with concrete advice on where you gain the most. |
| Feed & infrastructure | XML feed for comparison sites, SSL certificate and domain expiry, SPF/DMARC (so your e-mails do not end up in spam). |
How we compute the score
Each of the five areas starts at 100 and loses points for the findings in it. A critical finding weighs roughly 3.5× more than a warning — and it is the only kind that also lowers the second half of the score, so in the total the gap is wider still. Deductions follow the share of affected pages, not the raw count: 43 broken links across 120 pages is as sick a site as 11 across 30. Past a certain share the deduction stops growing — one broken thing must not sink a whole site. The total is 60 % the average of the five areas and 40 % how many critical findings remain. And honestly: fixing one page out of eight usually will not move the number — what moves it is a finding disappearing entirely. The formula is shown right in the report next to your result.
What we deliberately do not report
The list we might be proudest of. Every item is a real false alarm we once produced ourselves — and then built a rule so it never comes back:
- Struck-through reference prices in product cards — a topic for the Omnibus check, not a “wrong price in the text”.
- Instagram, Facebook, YouTube and Maps as “broken links” — these services answer robots differently than people.
- Tracking pixels (1×1, invisible) as “images without alt text”.
- Anti-spam honeypots in forms as “fields without labels” — screen readers do not see them, and that is correct.
- Decorative images with
alt=""oraria-hidden— the standard says they are meant to be skipped. - “You save €4”, “free shipping over €40”, “€25/kg” — amounts that are not a product price.
- Prices with ambiguous attribution — when we are not sure which product an amount belongs to, we do not report it.
Security & privacy
- No passwords. We never ask for administration access — we read only public pages, just like your customer.
- No changes. The check-up writes nothing to your site, orders nothing, fills in no forms.
- Gentle pace. Pages are loaded in small batches so the check-up does not strain your server.
- Your data. You see the results; we keep scan history for the “what changed” comparison and you can delete it at any time.
The mini check-up is free, no registration, results immediately.
Run the free check-upHow we calculate "What might this be costing you?"
The box uses only the numbers you enter yourself (orders per month and average order value). We apply deliberately conservative ranges of orders at risk per finding type: broken links 0.2–0.5% each (max 10), broken pages 0.5–1.5% each (max 5), broken images 0.3–1% total, e-mails at spam risk 1–4%, pages hidden from search engines 1–3%. The total is capped at 25%. It is a ballpark estimate, not a promise — the real impact depends on where exactly the error sits. Legal findings (Omnibus, mandatory information) are not counted; for those we state the statutory fine risk, which is set by law, not estimated.