Nobody churns because they ran out of credits, they churn on the third render
A credit balance turns every click into a purchase decision. That's the whole problem and it has nothing to do with the price being too high.
In a tool with several studios a video render costs roughly 40x an image. One credit unit spanning both means the user is doing arithmetic before every button, and the arithmetic gets worse as the balance drops. The first render feels free. The third one has a price tag on it. People stop long before the balance hits zero, and that shows up in your dashboard as low usage, never as a complaint about cost.
Flat seats have the opposite failure. Your heaviest users cost you real money and you can't see them coming. So most of us bolt credits onto a seat and ship both problems at once.
The part I haven't solved is showing remaining capacity in a unit nobody has to convert. Minutes of video and number of images are legible. Credits are not, and the conversion table is where the trust goes.
If you sell credits, did anything actually stop the arithmetic, or did you just make the balance bigger?
Replies
The image vs vodeo cost difference makes one shred credit balance pretty confusing.
I'd rather see ''2 videos left'' or '' 10 images left'' than a random credit number.
@_raja_tashfeen Agreed, and the catch is that the number has to stay true. "10 images left" is only better than a credit balance if it survives someone switching to a 4K model on the next click, otherwise I've swapped an abstract number for a concrete one that turns out to be wrong. That's a harder engineering problem than the credit display, which is probably why most tools ship the credits.
Price the action, not the consumption. Mine is an AI flashcard generator, and it's one credit per generation whether that produces ten cards or a hundred. There's nothing to convert and nothing to optimise, so the arithmetic has nowhere to happen. The balance reads as "you can do this six more times", which is already the legible unit you're after.
That only works because my expensive and cheap actions sit within about 5x of each other. Yours are 40x apart and I don't think that's a display problem. A conversion table shows up when two economies got merged into one number.
Another thing that worth mentioned - the refund. The credit comes back on its own when the model returns nothing, and when it comes back well short of the count that was asked for. And anything that isn't a model call, the audio on the cards for instance, never costs anything. "The credit buys the AI call, everything else is free" is a sentence someone can hold in their head. A conversion table isn't.
@siarheihamanovich The 5x versus 40x line is the part I can't argue with. Two economies merged into one number is exactly what we did, and no amount of display work fixes it. The refund rule is the bit I'm taking today, because right now we charge for a run that comes back with a mangled face and the user can't tell our bug from their bad prompt, which costs more trust than the credit is worth. Flat rate per action is the other half and I don't think I can do it yet, our p95 run costs about 40x the median, so pricing every action flat means pricing images at video cost to stay safe.
The 40x difference between video and image makes a single credit system especially hard to understand. Have you considered showing capacity separately for each studio?
@simonclark55 That's where I've ended up, per studio capacity in each studio's own unit. What I haven't solved is the boundary, since one wallet funds all six. If the video studio says 4 renders left and the image studio says 60 images left, those are the same money described twice, and somebody is going to spend the 60 and then ask where their videos went. Separate numbers off a shared balance is a new way to lie, just a friendlier looking one.
The pricing model really changes how willing people are to experiment. Would showing remaining renders instead of credits make the cost feel less restrictive?
@jenniferdavis Less restrictive, probably not. Less exhausting, yes, and I think that's the bigger lever. The cost is the same either way, what changes is whether the person has to do a division sum before clicking, and it's the sum that stops people rather than the price. Nobody has ever written in to say a render was too expensive.