How would you improve retention for a SaaS where users often get what they need and leave? Also, how should an early startup calculate CAC and LTV? - (I will not promote)

I am working on Note2Tabs, a tool that turns guitar audio into editable tabs directly in the browser. You can transcribe a song, fix the notes and fingering, play it back, and use it for practice. We are getting users and people are clearly finding value in the transcription itself, but retention is something we want to improve significantly. The challenge is that transcription can be a very transactional use case. Someone has a song they want tabs for, they get the result, and they may not have an immediate reason to return. We are increasingly thinking about Note2Tabs less as just a transcription tool and more as a complete workflow for guitarists. Transcription gets someone through the door, while editing, playback, practice, creation and eventually sharing should hopefully give them reasons to keep using it. For anyone who has worked on a SaaS with a similar utility style entry point, what actually helped retention? Did you find success by improving the core product, adding adjacent recurring use cases, building habits around the product, email reengagement, sharing, community features, or something else entirely? I am also trying to get a better understanding of CAC and LTV at an early stage. We have mostly grown organically so far, which makes CAC difficult to interpret. Simply dividing our current marketing spend by acquired customers does not seem particularly useful when we have barely experimented with paid acquisition. LTV is similarly difficult because we do not have enough history yet to know what mature customer lifetime and churn will look like. For people who have been in this position, how did you establish credible CAC and LTV figures early on? Would you run small paid acquisition experiments specifically to get an initial CAC benchmark? And for LTV, would you use current retention and subscription data with assumptions, cohort based estimates, or wait until you have enough history for the number to mean something? Would love to hear how other early SaaS founders approached this, especially anyone building prosumer or creator tools.