Library

In a restaurant test, one mint with the bill raised tips by about 3%. Two mints raised tips by about 14%. One mint plus a second one offered as a personal extra raised tips by 23%.

A lender in South Africa mailed 53,000 loan offers. A photo of a woman raised loan take-up by men as much as a large rate cut. Fewer example loans also helped.

An audit of 2,241 US companies found that the average reply to a web lead took 42 hours. Firms that replied within one hour were nearly seven times as likely to qualify the lead.

When a Hershey's Kiss cost 1 cent and a Lindt truffle cost 15 cents, 73% chose the truffle. When both prices went down by 1 cent, 69% chose the free Kiss.

In a classroom test, a print-only offer at the same price as print plus web made 84% of students choose print plus web. Without that offer, only 32% chose it.

A test with 689,693 pharmacy customers compared 22 text reminders. On average, texts raised vaccine visits by 6.8%. The best texts said that a vaccine was waiting for the person.

In a clothing catalog test, a dress priced at $39 sold more than the same dress at $34. It also sold more than the dress at $44.

A Facebook app test found that automatic updates to friends raised new users by 246%. Personal invites raised them by only 98%, because people sent few invites.

A Yahoo experiment with about 1.6 million customers of a large retailer found that online ads raised sales. The extra sales in stores were more than six times the extra sales online.

In a hotel test, a card that said most guests reuse their towels raised towel reuse from 35% to 44%. A card about guests in the same room raised it to 49%.

A field test at a travel website found that general ads brought back more visitors than ads with the exact hotel they saw. Exact-product ads worked better only for people who were ready to choose.

Field tests at a test-prep company found that a first name in the subject line raised opens by 20% and sales leads by about 30%. Unsubscribes went down by 17%.

A study of about 7,000 New York Times articles found that positive articles were shared more. Articles that caused awe, anger or worry were also shared more. Sad articles were shared less.

A review of 25 large online ad experiments found that ad effects are very small compared with normal changes in sales. Most ad tests are too small to show if the ads paid back.

A study of 15 Facebook ad experiments compared real test results with the usual report methods. The report methods were often wrong, and sometimes they showed three times the real effect.

When eBay stopped its ads for searches on its own name, almost all of that traffic still came through the free search results. Most of the paid clicks were clicks that eBay already had.

A 10-week Facebook test with 34,849 advertisers compared 2 AI ad writers. The AI that learned from real clicks raised the click-through rate by 6.7%.

A 21-month Pandora experiment with almost 35 million listeners found that each extra ad per hour cut listening time by about 2%. A short test can miss most of this cost.

A field test with 8,966 telecom customers found that a chatbot that disclosed human help got longer messages. The authors estimate that human work cost 38% to 70% more.

A Spotify test with 852,937 users found that personal podcast picks raised listening by 28.90%. But the mix of podcasts for each user became less varied.