Hi, I made a loom to explain, hope thats ok. I can send screen shots if needed. Understanding Cost Calculations for Pickaxe Usage | Loom
Trying to get the costs of a ‘run through’ of my pickaxe per user which is a series of questions to help then with their brand strategy, we will use this pickaxe as a lead magnet and want to know the general cost per guest user. I’m having issues with the definitions and many numbers and names that don’t match - please explain.
Thanks
Hi @biba,
Thank you so much for taking the time to record the Loom. I can hear how carefully you’re trying to understand the numbers, and it’s totally normal to feel confused at first. The admin dashboard shows a lot of technical data that isn’t meant to be used for cost-per-user estimation, so you’re not doing anything wrong.
Let me walk you through each point you raised and simplify the whole thing.
This part is correct. You spent $0.37 for 35 test runs, which averages to about 1 cent per run.
That already gives you a solid benchmark: your lead magnet is extremely low-cost.
Your Pickaxe is using Haiku, which is one of the cheapest high-quality models. That’s why the Loom mentioned “3,000 to 30,000 messages per $15” for Haiku. In your case, the numbers match that range.
Sometimes:
• the model returns extremely short output
• or it’s a very old message
• or the analytics backend didn’t record a tiny value
When that happens, you’ll see “No analytics for this submission.”
This is normal and doesn’t affect your cost at all.
This is also normal. Costs vary because the model considers:
• how long the user typed
• how long the model’s reply is
• which reasoning path it took
• which Knowledge Base chunks it fetched
• whether Actions were triggered
• how long your system prompt is (because the model rereads it every time)
Two runs can be structured the same but still cost slightly differently.
Nothing to fix here. This is simply how all LLMs work.
Every time the model replies, it reprocesses your entire prompt to understand how to behave.
If your prompt is long, that increases cost.
This is fully under your control.
This is often the biggest hidden cost for builders, so trimming the prompt helps.
If your Pickaxe uses any Actions (email, webhook, PDFs, etc), each Action call adds:
• one extra LLM message
• extra tokens
• extra reasoning
If your flow triggers 2–3 Actions in one run, the cost naturally increases.
If no Actions fire, the run costs less.
Forget the token numbers for cost calculations.
Pickaxe credits are very simple:
1 AI message = 1 credit.
So the easiest way to estimate your cost per lead is:
• Run your Pickaxe from start to finish
• Count how many answers the Pickaxe sends
• That is your cost in credits
The analytics dashboard is meant to give you a big-picture view, like how many runs happened and how much it cost overall. It won’t show exact cost per user, because every user behaves differently. Some people type long answers and use more credits, while others keep it short and cost almost nothing. Over time these differences usually balance out, which lets you estimate an average cost per lead instead of trying to track each person individually.
Haiku is extremely cheap per token.
So even if a Haiku run uses more credits (more messages), the underlying token cost is still tiny compared to more advanced models.
This is exactly why Haiku is recommended for lead magnets.
I recently wrote a detailed post on maximizing profit and choosing the right AI model for your use case. It breaks down costs, performance, and good defaults in plain language.
Here’s the link in case you want to check (highly recommended):
It should make the “3,000 to 30,000 messages per $15” math much easier to understand.
Your Loom showed that your entire 35-run test batch cost you only $0.37.
That is right in line with Haiku’s expected pricing and a very good sign for using it as a lead magnet.
If you ever want to walk through this together in real time, I’m more than happy to hop on a quick call and go step by step. We can look at your flow, break down the exact cost per lead, and even explore ways to make your setup more profitable and efficient for lead generation. You’re building something valuable, and I’m here to help you make the numbers work in your favor. 