1 in 13 storefronts

A banner that offers a choice does not always give reject the same visual weight as accept. Reject is often a link inside the text, and where both buttons exist, accept is often the one that stands out.

"Visually favoured" is a rule applied to the screenshot: a filled button counts as more prominent than an outlined one, then colour and size decide. It describes what is on screen, not what a site intended or how visitors choose.

Positions are relative to the screen, so a corner banner that spans the width of a phone screen may be classified as bottom centre. The same banner can appear in different positions at different screen sizes, which is why the dataset keeps one row per site and viewport.

Rows

3,174, one per site and viewport

Sites

1,587 storefronts on three Turkish SaaS ecommerce platforms

Columns

44

Viewports

desktop 1041 x 750, mobile 390 x 844

Licence

CC BY 4.0

The columns come in groups:

• Site: coded platform, sector, market and declared language.
• Banner: type, position, share of the screen it covers, backdrop.
• Vendor: a coded vendor and whether the banner is provided by the ecommerce platform itself or by a third-party consent management provider.
• Controls: which of accept, reject, settings and close are on the first screen, and how reject is offered.
• Colour and prominence: theme and hue of banner and buttons, how much the banner stands out from the page, and which button is visually favoured.
• Accept vs. reject: same colour, same style, difference in font weight, label contrast.
• Wording: the category of each button label, never the label itself.

No row names a website. Platforms and vendors are coded, measurements are grouped into ranges, and the file holds no domains, screenshots, banner text or cookie names.

Read the datasheet or the column definitions (columns.json).

Setting up a consent banner? Paste this into your own AI assistant, add a description or a screenshot of your banner, and compare it with what the data shows. It tells you what is common and what is rare, not what the law requires.

You are helping me review the cookie consent banner on my ecommerce store.
Data: Consent banners on ecommerce storefronts v1.0.0 (CC BY 4.0),
https://doi.org/10.5281/zenodo.24090345
CSV: https://raw.githubusercontent.com/ceaksan/storefront-consent-banners/v1.0.0/consent-sites.csv
Column definitions: https://raw.githubusercontent.com/ceaksan/storefront-consent-banners/v1.0.0/columns.json

Read the column definitions first. Then compare my banner, described below, with the
sites in the data, overall and in my sector: whether reject is offered on the first
screen and how, whether accept and reject look alike, where the banner sits on desktop
and mobile, how much of the screen it covers, and the colour contrast of the buttons.

Rules:
- Tell me what is common and what is rare. Common is not the same as right: the most
  common practice in this dataset may still be problematic.
- Do not judge legal compliance. The rules come from KVKK and from my CMP tool's
  documentation; point me to them instead:
  https://kvkk.gov.tr/Icerik/7353/Cerez-Uygulamalari-Hakkinda-Rehber and
  https://www.mevzuat.gov.tr/mevzuatmetin/1.5.6698.pdf , and to my CMP tool's own docs.
- Use only this data and what I tell you. Say when the data cannot answer something.

My banner: <describe it, or attach a screenshot>

I opened each homepage in a fresh browser with no stored cookies and captured the visible area before scrolling or interacting with the page. When a banner did not appear during the automated capture, I retried the capture or captured it manually in a regular browser. I did not accept or reject cookies.

I labelled every screenshot myself, including the banner, its backdrop and each control. I measured fill colours from screenshots. I used computed styles for button label colours and font properties only where background colours matched the screenshots. Otherwise, I sampled label colours from screenshots. Some font values came from a second visit.

Read the numbers with these limits in mind:

• A curated sample: the example stores featured by the three platforms, not every store on them and not Turkish ecommerce at large.
• Captured in November 2026. A banner can change, be tested or depend on location.
• I emulated phone screen sizes in a desktop browser rather than capturing on real phones.
• I was the only annotator, so agreement between independent annotators was not assessed.
• "Visually favoured" is a rule on what is on screen. I did not test whether visual prominence affects visitors' choices.
• The data says what is common, not what is lawful.

The datasheet has the full method, every column and the limits in detail.

Every Medium Term Programme since the 2024-2026 one has scheduled the alignment of KVKK with the GDPR and moved the target each time (2024 Q4, 2025 Q4, 2026 Q3, now 2027 Q3), so the rules that apply to these banners are still in transition. Medium Term Programme 2027-2029.

DOI

10.5281/zenodo.24090345 (v1.0.0)

All versions

10.5281/zenodo.24090344

Licence

CC BY 4.0

Attribution:

Ceyhun Aksan, "Consent banners on ecommerce storefronts", v1.0.0, https://doi.org/10.5281/zenodo.24090345 (CC BY 4.0)

Citation tools can read CITATION.cff.