Lantern’s evidence review

Fictional case

It is January 2027. Ana Lim has four decisions to make for Lantern Coffee Co. and some evidence behind each: whether to move the shop’s parcels to a faster, dearer courier; which subject line to use for gift-box emails; whether to keep running all three workshop formats; and whether to offer fortnightly deliveries to subscribers. She also wants to plan a proper customer survey for 2027. She asks you to say what the evidence shows, how sure she can be, and what it can’t tell her. Lantern, Ana and every figure are fictional.

What you have:

The courier trial (16 November – 18 December 2026): each order was sent with Swiftline, Lantern’s current courier, or Parcelbird, chosen at random, and Ana tracked a random sample of the parcels: business days to delivery for 64 Swiftline and 58 Parcelbird parcels. A parcel is late if it takes more than 3 business days. Swiftline charges $5.20 a parcel; Parcelbird $5.65. Ana plans for about 450 parcels a month in 2027.
The gift-box email (12 November 2026): three subject lines, each sent to 1,500 people chosen at random from Lantern’s mailing list of about 6,000 (newsletter sign-ups as well as customers); the number in each group who clicked through to the gift-box page within a week. All three emails offered free delivery on gift boxes; only line C’s subject line said so.
Workshop satisfaction (2026): scores out of 10 from participants in Lantern’s barista workshops, by format (in person, online, hybrid). Participants chose their own format.
The subscriber survey (December 2026): all 520 subscribers were asked whether they would rather receive their coffee every two weeks instead of monthly; 312 replied.

What you will produce

A comparison of the two couriers: the late shares, the difference in average delivery time with its 95% confidence interval and test, whether the trial was big enough, and what the faster service would cost and save.
A reading of the email test that doesn’t over-claim, including what testing pairs after the fact would do.
A one-way ANOVA of workshop satisfaction by format, with its conditions, effect size, follow-up comparisons and limits.
An interval for the subscribers who prefer fortnightly delivery, the bounds that non-response leaves, and which decisions those bounds can settle.
A sample size for the 2027 customer survey and the number to invite.
A memo to Ana recommending what to do on each question, what each mistake would cost, and what to measure next.

Loading the project…