The page you requested doesn't exist

We could not locate the page: /%20p/-perform-30-100-better-by-ab-testing

Try one of the featured posts or search the archive below.

Featured Posts

Jan 14, 2025

Bad news? Send AI. Good news? Be human

We don’t think AI has bad intentions, so we are up to 2.6 times more likely to accept bad news given by them (e.g. accept a higher-than-expected price).

Thomas McKinlay

Nov 22, 2022

Where you should display your price

Scientific research found that we perceive prices as cheaper when they are below products (vs above), increasing sales.

Thomas McKinlay

Sep 06, 2022

Handwritten thank you notes increase sales

Give a note thanking customers for their purchase. They must be handwritten, but photocopies work just as well as originals. In experiments, they increased future customers spending 2x.

Thomas McKinlay

Special case study: The 2 steps to effective chatbots

How to design an ecommerce AI chatbot your customers will love, using insights from our Wharton Blueprint for Effective AI Chatbots.

Thomas McKinlay

Apr 14, 2026

What biases AI agents to choose your product

Optimizing your website for AI buying agents can drastically increase the likelihood they choose your product (e.g. tweaking a product title increased choice by 80 percentage points).

Thomas McKinlay

Apr 07, 2026

AI-driven traffic isn’t (yet) as good as it’s hyped to be

Research on 164m sales across 973 ecommerce websites finds that AI traffic drives more revenue per session than paid social, but less than any other channel. This gap is smaller for complex products.

Thomas McKinlay

Mar 31, 2026

The optimal text/visuals balance

Images perform worse on social media when the text is too large or when the text and imagery refer to different concepts.

Thomas McKinlay

Mar 24, 2026

Stacked (vs single) discounts work better

Stacked discounts (e.g. 15% sale + 10% online coupon) led to up to 16% higher purchase intentions vs a standard discount of the same amount (25%).

Thomas McKinlay

Mar 19, 2026

Premium

Case study: When and how to allow customization

Allowing people to customize their experience (e.g. preferred shampoo scent) boosts satisfaction and reduces cancellations by up to 46%.

Thomas McKinlay

Mar 17, 2026

‘AI-designed’ hurts sales

People are up to 29% less willing to buy products designed by AI (e.g. a chair, a snack flavor), compared to those designed by humans.

Thomas McKinlay

Mar 10, 2026

When people are most likely to switch brands

People are most likely to switch brands when they’ve tried a few products, but are not yet category experts. 54% of those who leave never return.

Thomas McKinlay

Mar 03, 2026

Which brand elements make you unique

People were over 2x more likely to correctly remember and associate characters and logos (vs colors) to the correct brand.

Thomas McKinlay

Feb 24, 2026

Keep your ads away from toxic content

People view a brand up to 19% more negatively if their online ads are placed next to unsafe or objectionable content (e.g. racist or homophobic content).

Thomas McKinlay

Customer ExperienceCustomer Experience

+3+3

Feb 19, 2026

Premium