The strongest signal here seems to be that 32 of 42 respondents either had a one-off job or already had enough credits. That sounds less like churn from a bad product and more like a subscription that does not match the job. Have you considered selling non-expiring credit packs as the default, then offering a subscription only to accountants with recurring volume? I would also split retention by customer type before changing price; tax-season users and bookkeeping firms probably have very different curves.