Sample case study
Brightloop raised its price without a test, then found a sharing problem
BrightloopLanguage-learning app with AI speaking practice, sold to individuals and schools380,000 paying users · $46M annual recurring revenue · Leadership team of 6
Sample case study. Brightloop is a fictional brand and the data is simulated to show how Cufo works. These are not customer results.
Brightloop's customers already learn together on one login, and it costs them lost progress and a second plan. The price rise from $79 to $99 was made without a test and checked afterwards: renewals fell by less than feared, but the check is not proof. The onboarding call for large school districts missed its estimate, and the family plan is proposed as the first offer to be tested before launch. That decision is still open.
- support tickets in 12 weeks about progress overwritten on a shared login
- 574
- of 18,400 survey respondents learn alongside a partner, child or parent
- 23%
- of pricing-page visitors who open "Can I share my plan?" leave without buying
- 71%
- 90-day school seat activation, below the 55% to 65% expected
- 52%
Situation
The situation
Brightloop is a language-learning app with AI speaking practice. It has 380,000 paying users and $46M in annual recurring revenue. A leadership team of six runs it.
In mid-October 2026 the CEO faced one big question: should the app launch a family plan before the January peak? 31% of annual subscriptions start between 26 December and 15 February, and the January campaign locks on 15 December. Two other things sat open. One was the price rise to $99 made in June. The other was school seats that were bought but not used.
The decisions here are the CEO’s.
1Data
See: What was collected
Six sources went in: three kinds of customer words and three kinds of behaviour records. Cufo redacted 2,340 emails and 410 phone numbers from tickets and removed duplicates before reading anything.
Sources and records
- Annual learner survey18,400
- App store reviews3,200
- Support tickets12,600
- Product usage380,000
- Billing and subscriptions380,000
- Pricing page and paywall458,000
- In all1,252,200
Gaps Cufo flagged
- The survey reaches active subscribers only, so accounts that share a login are almost invisible in it.
- The survey was last exported on 22 September, 27 days before the sample date.
- 1,180 Portuguese and German tickets are waiting for translation before they can be coded into themes.
- The 41,800 shared-login accounts are a model estimate from usage patterns, not a confirmed count.
- App store reviews carry no account link, so they cannot be matched to billing.
2Analytics
Think: What it means
Cufo grouped the survey, reviews and tickets into themes, then checked each against billing and product usage. In the sample, a review panel of five roles argued the family plan question over two rounds. The weekly counts below cover the 12 weeks to 18 October 2026.
Themes
Each chart shows mentions a week over the 12 weeks before the case, oldest on the left.
Shared logins overwrite progress
574 mentions
People on one login say a partner's or child's lessons wipe their streak and reset their level. 191 of these tickets ended in a refund request.
Asking for a family or shared plan
514 mentions
Couples and parents ask for one plan on one bill. Some say they will switch at renewal. A few say a family plan would change nothing for them.
Second-plan discount requests since the $99 renewal
138 mentions
Couples who renewed two annual plans at $99 each ask for a household discount.
School seats not activating
410 mentions
Teachers say invite emails land in spam and nobody showed them how to assign lessons. Seats are up 34% since June.
In customers’ words
Illustrative lines from the simulated data, with personal details removed.
We share one login. Every morning my husband's streak overwrites mine. Is there a fix that isn't buying a second plan?
Support ticket
We renewed two annual plans this month at $99 each. Last year they were $79. Is there a loyalty or household discount?
Support ticket
The onboarding call was with our district coordinator, not the teachers. We still don't know how the dashboard works.
Support ticket
The finding
Customers already learn together on shared logins, and it costs them progress and money.
Evidence
- Fact4,232 of 18,400 survey respondents (23%) say they learn alongside a partner, child or parent.
- Fact940 support tickets in 90 days mention sharing. 574 of them (61%) are about progress overwritten on a shared login.
- Fact41,800 of 380,000 accounts (11%) show two or more learners. This is a model estimate from usage.
- Fact286 of 3,200 app store reviews mention sharing, family or kids. 203 of them complain about paying twice or losing progress.
Confidence
Medium
The survey and ticket samples are large, but the survey only reaches active subscribers. Demand is measured. Whether a family plan would make money is not.
Considered and set aside
- Taking stated intent as proof of demand. 14% of survey respondents said they would definitely pay for a family plan. Stated intent overstates buying, so Cufo reads 14% as a ceiling.
- Asking customers what they would pay. Willingness to pay is unreliable when asked, so the planned test measures it from behaviour.
- Using the 81% renewal of households that pay for two or more plans as proof that a family plan would renew better. Those households chose to pay twice, so they were already committed. Solo annual subscribers renew at 68%.
- Blaming seat growth alone for weak school activation. It explains part at most, because activation fell for small districts too.
3Experiments
Try: What was tested
Two changes were made without a test and checked afterwards. The family plan would be the first offer change tested before launch.
The tests
-
Test 1 of 2Complete
Annual price $79 to $99: renewal check-back
Hypothesis Raising the annual price from $79 to $99 costs 1.5 to 3 points of renewal.
- Design
- Matched groups. Two groups matched as closely as possible, but not assigned at random.
- Measure
- Annual renewal rate. Higher is better.
- Dates
- 1 Jul to 30 Sep 2026
- The rule
- Adopt only if the difference is clear at p below 0.05, in the right direction.
- Rule set in advance
- No, it was set after the start
What the numbers say
Clear loss for the variant
69.5% for the control (9,869 of 14,200) against 68.4% for the variant (19,084 of 27,900), a difference of 1.1 points against the variant; the test can tell this apart from chance (p = 0.02): the variant did worse.
Control: renewals due in July, at $79Variant: renewals due 1 Aug to 30 Sep, at $99Lines show 95% intervals
The groups were not assigned at random and the rule was not set before the test started, so these figures are a check, not proof.
Exact figures
Group Total Counted Rate 95% interval Control: renewals due in July, at $79 14,200 9,869 69.50% 68.74% to 70.25% Variant: renewals due 1 Aug to 30 Sep, at $99 27,900 19,084 68.40% 67.85% to 68.94% Difference, variant minus control −1.10 points, 95% interval −2.03 to −0.16 (Newcombe) Relative change 1.6% lower p-value, pooled two-proportion z-test 0.0215 (the rule needs below 0.05) Total: people or sessions in each group. Counted: how many of them did what the measure counts. All figures are computed on this page from those two counts.
What the brand did
Adopted The price was raised in June with no test, so this check was built afterwards from billing records. The two groups are different months, not random halves. A market-by-market comparison with last year's pattern pointed the same way in all five markets. Renewal fell by less than feared and revenue per renewal rose 25%, so the price stays. Five markets are too few for a reliable interval: read this as a check-back, not proof.
-
Test 2 of 2Planned
Family plan at $119, UK and Canada
Hypothesis Adding a $119 family plan raises contribution per pricing-page visitor without more than 31% of buyers having otherwise bought two or more plans.
- Design
- A/B test. People were split at random into two groups, one with the change and one without.
- Measure
- Contribution per pricing-page visitor. Higher is better.
- Dates
- Planned 9 Nov to 12 Dec 2026
- The rule
- Adopt only if the difference is clear at p below 0.05, in the right direction.
- Rule set in advance
- Yes, written into the plan before the start
What the numbers say
Not run yet
What the brand did
Planned Half of pricing-page visitors in the two markets would see the plan, and one checkout question would ask how buyers would otherwise have paid. This is a proposal that waits on the CEO, who is due to decide by 26 October.
Guardrails
- Family buyers who would otherwise have bought 2 or more plansLimit at most 31%. Not observed yet.Limit set
- Support tickets per 100 family accounts a monthLimit at most 2.5. Not observed yet.Limit set
4Strategy
Act: What was decided
Every decision is logged with the result it was expected to give, then checked later against what happened. Four checked decisions are shown here, with the one still open.
Decisions on the record
Tested: decided from a test. Untested: decided without one. Pending: not decided yet.
- Untested
Raise the annual plan from $79 to $99, for new subscribers from 1 July and renewals from 1 August.
Adopted without a test. At 60 days renewal fell by less than expected and annual-plan bookings rose 14%, inside the 9% to 15% expected. Annual subscribers look less price-sensitive than assumed, so churn estimates for pricing questions now start lower. Next check: 28 January 2027.
- Untested
Add one free streak repair a month for annual subscribers.
Both checks landed inside the expected range: 30-day retention for new annual subscribers rose 0.9 points and streak-related tickets fell 28%.
- Untested
Offer a 45-minute setup call to school districts above 500 seats.
Missed its estimate: 90-day seat activation was 52% against 55% to 65% expected. The calls reached district leads, not teachers.
- Untested
Move 20% of Brazil paid social budget to creators.
Paid acquisition cost in Brazil fell 11%, inside the 8% to 18% fall expected.
- Pending
Test a family plan at $119 in the UK and Canada, 9 November to 12 December.
Recommended in a decision memo dated 19 October. Finance would test at $139 instead, which likely earns more per buyer in year one. The CEO has not decided.
What changed next
- As of 19 October 2026 the CEO is due to decide on the family plan test by 26 October, so the result can be read before the January campaign locks on 15 December.
- Separate profiles on one account are built alongside the family plan from 26 October. Sharing tickets were 81 a week at the last count, and the goal is under 40.
- Customers with two or more learners will get an in-app note once the profile fix ships.
- The price check-back runs again on 28 January 2027.
- School seats are still an open question. Cufo proposes checking invite email spam rates first, then a 20-minute teacher session in the first week for half of new districts, judged on 30-day seat activation.
- Whether shared-login households would pay to split is still unknown. A short in-app question for those accounts is drafted but not sent.
Takeaways
What this case shows
A change made without a test can still be checked, if the check says what it cannot prove.
Measured demand is not the same as a profitable offer, so the family plan waits for the buyer mix.
Misses are logged next to wins, and each one shapes the next question.
More
More sample case studies
-
Sample case study 01
Fernhill Botanics
Clean skincare for sensitive skin, sold online
Refills of a best-selling serum fell. A reminder test was inconclusive; reviews and tickets pointed to a new pump that leaks or jams.
116pump complaints in the six weeks after the pump changed
-
Sample case study 02
Tidewell Home
Home goods sold online and in 40 stores
A fictional home goods retailer sees fewer first-time buyers order again. Two finished tests, three untested decisions and one open question follow.
17.8%of first-time buyers from April to June ordered again within 90 days. A year earlier: 23.5%
-
Sample case study 04
Larkfield Mobile
A challenger mobile network in 14 European markets
A fictional mobile network sent an app update to everyone at once. See how Cufo found the cause, stopped a risky test and planned a staged fix.
2.1% to 9.8%port-in eSIM failures, before 9 Oct and at the peak on 12 Oct
The demo shows the Fernhill Botanics case inside Cufo, with the same simulated data: open the demo.
Waitlist
Hear when Cufo opens
Cufo is in private development. Early access opens to a small group of brands first.