Every cancelled autopay is recurring savings Jar stops seeing. When a user taps cancel, this shows what they have built in their own numbers and offers a smaller step that fits their reason. Real offers from Jar's system, and the cancel button always one tap away.
Jar's system works out the offers it can make (lower amount, a 7-day pause, weekly instead of daily, or selling part of the gold). The model picks the two that fit the reason and writes the screen. The cancel button always stays.
Eight checks: valid output, real offers only, offer fits the reason, cancel stays visible, fits the screen, right language, no invented numbers, approved claims only.
Eight checks: valid output, real offers only, offer fits the reason, cancel stays visible, fits the screen, right language, no invented numbers, approved claims only.
If this screen keeps 8 in 100 users who try to cancel, Jar keeps ₹78 Cr of savings a year. Even at half that, ₹39 Cr.10,000 cancel attempts a day, ₹30/day autopay, 90 days of saving kept. Change the assumptions
Move the sliders to Jar's numbers. Jar runs 1M+ autopay transactions a day (public job post), so even a few points of saves are large.
Savings stay in gold that users own; for Jar, kept autopays protect the recurring flow that drives revenue. Run a screen above to see what writing 3,00,000 screens a month would cost.
