reallycurious

some people just need it for one time job others more regular. I guess for one time usage I should charge and for regular users require a credit card...but they will cancel it and extract as much as they want in the 14 day trial....

or I was thinking if it's a huge pdf file, charge per row of data extracted.

you mean like charge a $1 for 2 day trial of unlimited use?

or give 14 day trial before card is charged.

This feels like the scene in Godfather where Michael Corleone walks into the restaurant washroom, comes out with a gun and gets his vengeance.

I always thought that it was more generic tool since it's made for people who wants a tool to extract data from pdf files. But maybe that itself ends up being a niche.

Thank you, I will try it, such a good idea, like the manual conversion idea from above.

Current user base is widely distributed there's no strong user base from one industry although real estate is a possibility. Selling to that niche means I would need to create a separate marketing page and maybe even modify the software to their needs?

It costs $30/month for about 40 conversions. I didn't indicate that some data can't be extracted. I will add that warning in the dashboard.

The manual labor is interesting but it will cost far more than $30/month for the user, so I'm guessing that's a potential upsell?

Running the service costs $100 / month. But me having to work on it full time for free.

Successful number of jobs is around 70~80%. sometimes bad extraction is also caused by the user labeling things incorrectly.

Say my hourly rate is $30/hr, I end up spending 10 hours, $300, fixing one user's bad extraction. My hope is that this fix will transfer to other user's encountering similar bad extractions. The smallest plan earns $30/month. From this pattern I can see that what the average user pays won't cover my time spent on fixing bugs with the extraction system or maybe this is an early assumption? Also about 10% of bad extractions are simply impossible to do with just software alone, it really requires human to intervene and do it manually (something I can't do because it takes so long).

is this better than the terrassect OCR?