Layered Trust Signals in E-Commerce Interfaces

What it is

My master's thesis in interaction design, submitted at NTNU in June 2026. It analyses how four e-commerce platforms build trust through design.

The thesis develops a framework I call Layered Trust Signals. It separates trust in the interface into three layers. Layer 1 is regulatory: the cookie consent dialogue. Layer 2 is institutional: payment logos, buyer guarantees and seller reputation. Layer 3 is the interface itself, meaning visual density, information hierarchy, and devices such as scarcity and social proof.

The framework is used to compare Amazon, Temu, eBay and Apple's online store.

The brief

The brief was my own. My supervisor was Aliaksei Miniukovich.

Research on trust in e-commerce sits across several fields: privacy law, information systems, visual credibility, and research on dark patterns. The fields rarely cite each other. None of them describes how the signals in the different layers work together in one interface.

A user who meets a manipulative consent dialogue, then a credible payment page, then a visually convincing product page has not been through three independent assessments. The thesis is built to describe that interaction.

What I did

I built an observation protocol of 18 questions, six per layer, and applied it identically to all four platforms.

Each platform was observed twice, on 19 and 26 May 2026, in Chrome on desktop, from a Norwegian IP address and without logging in where that was possible. The same product category was used throughout: a portable Bluetooth speaker. All button text, guarantee wording and scarcity phrasing was transcribed verbatim. Where the two sessions gave different readings, I kept both.

Two sessions were not planned as replication. Temu changed its consent flow between them, and that was the reason I went back and checked the other three.

Temu had the most compliant consent dialogue of the four. Symmetric buttons, rejection in one click, advertising cookies off by default. The same platform had the most persuasive devices on its product page: seven in the first viewport, among them a scarcity warning and a struck-through recommended price. The two layers are designed by different logics.

Amazon and Apple showed no consent dialogue at all to a Norwegian visitor. On Amazon the only opt-out sits behind a login.

On three of four platforms, the legally binding guarantee was not the most visible trust signal at the point of purchase. What sat at the top was weaker: algorithmic badges, sales counts and self-stated claims.

Apple broke with what the research would have predicted. The product page does not follow the e-commerce standard, has no reviews and no urgency signals, and is still perceived as the most credible. The brand does the work before the interface gets a chance.

The term I propose for the pattern is trust compensation: a weakness in one layer can be covered by strength in another. The limit is the point where use continues but trust does not.

What I would change

The thesis does not measure trust. It describes how trust is designed. That was a choice, and it is also the largest limitation: I can say what the platforms do, not what users are left with.

With more time I would have added an empirical part. User testing where participants move through the same four flows and rate trust along the way would have made it possible to test the claims rather than infer them from the literature. Eye or pupil measurements would have added a layer on the product pages.

The simplest measure would have been a second observer. Qualitative interface analysis is interpretive, and the protocol was only applied by me. A second coder on a sample of the observations would have shown how much of the findings sit in the protocol and how much sit in me.

Facts

Type

Graduate thesis in Interaction Design (MIXD)

Institution

NTNU, Faculty of Architecture and Design, Department of Design

Supervisor

Aliaksei Miniukovich

Submitted

June 2026

Platforms analysed

Amazon, Temu, eBay, Apple's online store

Observation period

19 to 26 May 2026