# Reading a price test > Revenue and contribution per visitor and retention, not conversion alone, and an arm that ran only in the exam peak. - Course: [Pricing: Set, Package and Raise Your Prices](https://productovi.com/courses/pricing.md) - Module: How do we test and review prices? - Lesson 30/32, ~28 min - Source: https://productovi.com/courses/pricing/reading-a-price-test ## Three arms, three answers **Alex Tutor:** Klasna's price test is over: eight weeks, 1 February to 28 March 2027, new visitors in Ukraine, the English 16-lesson package. **Nazar:** A confession first. In week 5 I added a third arm, C at 720, because B looked fine. So A and B ran all eight weeks and C the last four. **Leo:** I only need one number. Which price sold best? **Alex Tutor:** Careful what "best" means. Let's read the data three ways: conversion, money per visitor, and whether buyers come back. **Dataset (CSV):** [Klasna's price test: English 16-lesson package, February–March 2027](https://productovi.com/datasets/pricing/reading-a-price-test/klasna-english-price-test-2027.csv) One row per week and arm. New visitors in Ukraine were split at random: A and B 50/50 in weeks 1–4, A, B and C in thirds in weeks 5–8. Money in UAH. - `week`: Test week, 1–8 - `week_start`: Monday of the week - `arm`: A, B or C - `price_per_lesson_uah`: Price per lesson in this arm - `package_price_uah`: Price of the 16-lesson package - `visitors`: New visitors who saw this price - `buyers`: Visitors who bought the package - `sales_uah`: buyers × package_price_uah - `repeat_buyers_60d`: Buyers who bought another package within 60 days **Sofia:** Conversion first: A 232 buyers of 4,000 visitors, 5.80%. B 221 of 4,000. C 84 of 1,600, 5.25%. **Leo:** There you go: 620 converts best. Keep it. **Alex Tutor:** Conversion tells you how many said yes, not how much each yes was worth. The higher price can lose a few buyers and still earn more. > **Exercise:** Use the dataset. What is arm **B**'s conversion over all eight weeks? In %, to two decimals. ## Money per visitor **Alex Tutor:** Revenue per visitor is sales divided by visitors: it multiplies conversion by price, so it puts both in one number. Every visitor costs the same to attract, whichever price they see. $$ RPV = \frac{\text{sales}}{\text{visitors}} = \text{conversion} \times \text{package price} $$ *RPV: revenue per visitor. For contribution per visitor, use contribution per package instead of the price: 16 × (price × 0.975 − 390 − 30).* **Sofia:** A: 2,301,440 / 4,000 = 575.36. B: 2,369,120 / 4,000 = 592.28. B is ahead by 2.94%. **Alex Tutor:** Check A's figure yourself. > **Exercise:** Use the dataset. What is arm **A**'s **revenue per visitor** over all eight weeks? To two decimals. **Hanna:** And on contribution the gap is much wider. A 620 package leaves 2,952; a 670 package leaves 3,732. The extra 800 of price is almost all contribution. > **Exercise:** What is arm **B**'s **contribution per visitor** over all eight weeks? A 670 package leaves 3,732 UAH of contribution. To two decimals. ## Is the gap real? **Nazar:** But B converts worse. 5.53% against 5.80%. Isn't that a real loss? **Alex Tutor:** At 4,000 visitors an arm, a gap this small is noise: if price made no difference, a gap at least this large would turn up about 59% of the time (p ≈ 0.59). We can't say B loses buyers; we also can't rule out a small loss. **Alex Tutor:** So ask how big a loss would matter. B could convert 7.46% below A before revenue per visitor falls, and 20.9% below before contribution per visitor falls. The observed gap is 4.66%. **Alex Tutor:** Where does the 7.46% come from? Work it out. > **Exercise:** A's package costs 9,920 UAH (620 a lesson), B's 10,720 (670 a lesson). By how many percent could B's conversion be **lower** than A's before B's revenue per visitor falls to A's level? To two decimals. **Sofia:** And repeat purchases: 40.95% of A's buyers came back within 60 days, 40.72% of B's. They stay just as well. **Alex Tutor:** Check B's repeat purchases in the data. > **Exercise:** Use the dataset. What share of arm **B**'s buyers bought again within 60 days, over all eight weeks? In %, to two decimals. > **Exercise:** B converts at 5.53% against A's 5.80% (p ≈ 0.59). What is the right reading? ## The arm that came late **Nazar:** Now my C. Over the test it earned 604.80 per visitor, the highest of all. Why not 720? **Alex Tutor:** Because C only ran in March, when the exam season starts and every arm sells better. Compare it with A and B in the same four weeks. > **Exercise:** What is arm **A**'s revenue per visitor in **weeks 5–8** only? To two decimals. **Sofia:** Weeks 5–8: A 694.40, B 710.20, C 604.80. In the same weeks C is the lowest, and only a third of its buyers came back. **Alex Tutor:** C converted 5.25% against A's 7.00% in those weeks, a gap unlikely to be chance (p ≈ 0.04). Comparing arms that ran in different weeks mixes the price with the calendar. That's why arms aren't added midway. > **Exercise:** **True or false:** arm C had the highest revenue per visitor over the whole test (604.80), so 720 is the best price of the three. ## Your turn: the reading **Nazar:** Give me the reading in a few sentences. Leo will ask which price, and why not the one that sold most. > **Exercise:** Write Nazar the reading of the test in 3–5 sentences: which price the data supports, with numbers, why not 620 although it converted best, and why not 720. Key numbers: conversion A (620) 5.80%, B (670) 5.53% (p ≈ 0.59); revenue per visitor A 575.36, B 592.28; contribution per visitor A 171.22, B 206.19; repeat buyers A 40.95%, B 40.72%; arm C (720) ran only in weeks 5–8, where revenue per visitor was A 694.40, B 710.20, C 604.80. > **Key ideas** > > - **Conversion is not the goal.** A converted best (5.80%), but B earned more per visitor: 592.28 vs 575.36 in revenue (+2.94%), 206.19 vs 171.22 in contribution (+20.42%). > - **Check whether a gap is noise.** B's lower conversion (5.53%) is within noise (p ≈ 0.59); a loss big enough to matter would be 7.46% (revenue) or 20.9% (contribution). > - **Retention counts too:** repeat purchases 40.95% (A) vs 40.72% (B). > - **Compare arms over the same weeks.** C's 604.80 is the highest overall only because it ran in the March peak; in weeks 5–8 it is the lowest (A 694.40, B 710.20). > - Don't add arms midway; if you must, compare only the weeks they share. **Alex Tutor:** Next: the yearly pricing review, where tests like this one sit next to realized prices, costs and competitors. - Previous lesson: [Fair and lawful price tests](https://productovi.com/courses/pricing/fair-price-tests.md) - Next lesson: [The pricing review](https://productovi.com/courses/pricing/the-pricing-review.md) --- Productovi Learning. Free, no sign-up needed to read.