How to price your AI SaaS - 2 mistakes we made

Today I decided to make less money...

• My pricing page before

• My pricing page now

Up until now, figuring out pricing for my SaaS has been unbelievably difficult. I've spent days playing with different models and I just couldn't find one that I was happy with.

So far I've made 2 big mistakes, and hopefully, you can learn from them.

For context, my tool Ferndesk has two core features:

• It can writes and publishes help articles from your codebase, your changelogs, videos, internal docs etc

• It scans scans your support inbox + product releases and find gaps and outdated content in your KB

Mistake #1. Pricing based on costs

I am running a very tight ship at Ferndesk.

100% bootstrapped, still solo, zero outside funding.

So I've been super wary about how much I spend.

Running both core features at a really high standard is VERY expensive.

Thinking about how much I \could\ lose from power-users made me overthink....

"What if I have 0% margins from one power user, or worse, negative?!"

So I started to penny-pinch on features, and went for usage based pricing.

Example: I was charging for every article drafted, regardless of whether they were good enough to publish or not.

But customers aren't stupid.

When they visit my pricing page, their first thought is- "so I'll be charged even if the output is bad?"

That's no bueno.

So instead, I'm switching to outcome-based pricing.

$0.5 for every article PUBLISHED - not drafted.

Even though this will undoubtedly be WAAAAYYYY more expensive on my end - this removes a HUGE cognitive burden from my customers.

Mistake #2. Introducing unnecessary complexity

• Every single line of copy...

• Every single concept introduced...

• Every single clause...

Gives customers something extra to think about.

Stack up enough of them, and they get overwhelmed, and leave.

Simplicity >>>>> than everything else.

ESPECIALLY if you're building something new and don't have the proof to back it up.

If your users ask,

"What do I get for $X?"

then you should be able to say "You get Y" in plain english.

In my case I did two things wrong:

Complexity #1. Introducing "edit credits"

I created this new concept called edit credits.

• 10 for a brand new article

• 5 for an edited one

The goal was to make editing articles cheaper than creating new ones. I wanted to protect users from being charged for one-line edits for example

But by introducing a new concept - I instantly lost an overwhelming amount of clarity on my pricing page.

Users would visit and ask, "What the heck is an edit credit???"

I knew this was a problem when even I couldn't tell how much usage I was consuming as I dogfooded.

So I killed the concept and replaced it with articles published.

Now users instantly know what they're getting when they visit the pricing page.

Also makes it easier to talk about my product in sales conversations. "Oh, if you sign up, you can publish X amount per month."

Not- "oh if you sign up, you can maybe publish X articles or edit X articles with this magical thing called credits"

Complexity #2. Adding tiered pricing when I really didn't need it

I split my pricing into 3 tiers- Solo, Startup and Business with increased limits on each plan.

Again, I fell into the trap of doing something because it's what everyone does.

Again, every additional word on your page is extra complexity that must be explained to the user. Introducing tiers wasn't making things easier for me.

At my former company Senja, we only really started growing when we killed all tiers and opted for a single plan.

So I did the same and went all in on one base plan + metered pricing.

I only just made this change, so only time will tell if this truly makes things better for me.

All I will say is, as a user of my own product, I am much much happier with this model.

TL;DR - how should you price your AI product?

I highly suggest doing the following, especially if you're building something new.

1. Kill as much complexity as you can.

Do you really need those extra paid tiers?

Can you pick one tier and meter it.

Do you really need to introduce a new concept? Can't you anchor it to something your customers already know?

2. Focus on outcomes, not on running costs.

You will almost certainly lose more money this way.

Margins may be tighter, and you will have to watch out super carefully for abuse.

But in return, you will get happier customers who know you care about their success with your product.