Kasteller turns the exports you already have into validated machine-learning predictions. You log in to a private dashboard and see a score and the reasons behind it for every customer, lead, appointment or machine, refreshed every month.
Each one is a custom LightGBM model with automated feature engineering, trained on your own history and checked with cross-validation before you ever see a score.
Gyms, salons, SaaS, ISPs and other subscription businesses
Predicts which members or subscribers are likely to cancel in the next 30–60 days, so you can reach out before they do.
Stripe, Chargebee, Mindbody or CRM exports
Insurance agencies, real estate teams, solar installers, B2B agencies
Ranks your open leads by how likely they are to close, so your team calls the right people first.
HubSpot, Salesforce or Pipedrive deal history
Dentists, chiropractors, therapists, vets, salons
Flags upcoming appointments most likely to no-show, so you can confirm, remind or double-book with confidence.
Scheduling system exports
Healthcare clients: a Business Associate Agreement (BAA) is available on request.
Manufacturers and trucking fleets
Predicts which machines, trucks or batches are likely to fail or produce scrap, so you can schedule work before it costs you.
Maintenance logs, sensor readings, quality records
A CSV or Excel export from Stripe, your CRM, scheduler or maintenance system. No integrations needed.
We engineer features, train the model and measure accuracy with cross-validation, so you know how much to trust it.
Your private dashboard ranks every record by risk, with the top reasons behind each score. Export to CSV anytime.
Upload fresh data each month and get updated predictions, tracked against previous runs.
Every engagement starts with a free audit of your data so you know what accuracy to expect before you pay anything.
Tell us about your business and, if you like, attach a sample export. We will reply within two business days with what we can predict, the accuracy we expect and a fixed price.
Your files are used only to build your models. They are stored in a private database, never sold or shared, and visible only to you and Kasteller. Your dashboard is behind a password-protected login, and each client can only ever see its own organization's predictions.
Yes. Email salh2kn@gmail.com and we will delete your uploaded files, predictions and account. You can also ask us to remove raw exports as soon as a model is trained.
Every model is evaluated with k-fold cross-validation on your own historical data: the model is repeatedly trained on part of the history and scored on the part it has not seen. We report the result (for example AUC, the chance a true positive is ranked above a true negative) on your dashboard, so you know how much to trust each run.
Usually a plain export you already have: membership or billing history for churn, CRM deal history for lead scoring, past appointments for no-shows, maintenance and sensor logs for equipment. A year of history is ideal; the free audit tells you if you have enough.
Yes. For no-show prediction in dental, chiropractic, therapy and other healthcare settings we sign a Business Associate Agreement (BAA) on request, and we recommend sending de-identified exports where possible.
No. You send an export, and you log in to a web dashboard that works on desktop and phone. Every table can be exported to CSV for your own tools.