I moved signup and paywall to the web so I could change price points, trial length, and copy without shipping a new build. Over the last month I ran small price steps of 5 to 10 percent on annual and monthly, plus toggled trials on and off.
To estimate elasticity, I looked at qualified reach to purchase, not pageview to purchase. For each price step I calculated elasticity as change in conversion over change in price, and I adjusted conversion for 7 day refunds and payment failures. I also tracked 30 day retention to catch trial churn that hides in day one conversion wins.
Two surprises:
- Annual conversion fell sharply above 84 dollars, but net revenue per visitor kept rising until 92 dollars when refunds spiked.
- A 7 day trial lifted initial conversion by about 18 percent but lowered 30 day paid subscribers enough to make net present value flat.
I am folding this into LTV by using cohort survival and ARPPU, then checking CAC by UTM to keep ROAS honest.
If you have done this, how do you model elasticity when trials and intro discounts are in the mix? Do you use net revenue after refunds, and what attribution window are you using to call a winner?
Use net paid after refunds and payment failures.
I run 5 percent price steps for a full week. Elasticity is conversion change over price change on users who reached the paywall. Not pageviews.
I use Web2Wave.com to flip prices and trials in minutes. Their AI gives a JSON and the SDK reads it, so changes show up fast. I call winners on a 28 day window.
Bracket test two or three price points and one trial. Let each run 3 to 5 days, then rotate.
The speed matters. I push changes on the web and they appear in the app instantly with Web2Wave.com. I judge on net revenue per visitor and day 30 paid rate.
Net revenue per visitor beats raw conversion every time
Model demand on paywall reach, not total sessions. Run 5 to 10 percent steps to map a curve, then fit a simple logit or use piecewise slopes. Compute net revenue after refunds and chargebacks. Layer 30 day retained payer rate to avoid trial inflation. Hold each step long enough to cover weekday and weekend. Check sample ratio mismatch and payment method mix. Feed the best price into an LTV forecast with 90 day revenue and compare to CAC by UTM.
For attribution windows, I use 7 day click and 1 day view for the media side, but I judge price tests on a 14 day revenue window.
Too many late refunds after day seven. It gives me a cleaner read without waiting a full month.
Net revenue after refunds is safer than raw conversion. It prevents silly wins from trial abuse.
We use 14 days for price tests. Short windows broke when refunds landed later.