Predictive Models
Four to six weeks, $18,000 to $35,000, fixed. A model that predicts no-shows, readmissions or lost customers from your own history, tested on cases it has never seen, with scores sent to the worklist or report your staff already use.
Prefer to skip the call? Get a written proposal
Built for
- Clinics losing visits to no-shows
- Care management teams deciding whom to call first
- Hospitals working to lower readmissions
- Health-tech companies that want to see churn coming
Why it's worth it
Today
You find out about the empty slot, the readmission or the customer who didn't renew after it happens. When staff work their lists in order, they can't tell who most needs a call. For a hospital, Medicare's readmissions program can also cut up to 3% of base payments.
After
Each morning staff see who is most at risk, ranked, in the worklist or report they already use, so reminders and calls go first to the people most likely to miss a visit, be readmitted or not renew. You know how often it's right, in plain numbers, before anyone relies on it.
What you get
Week 1
The decision
Who acts on the prediction, when, and what they do differently. We also confirm your data can support the model.
Weeks 2–3
Model and validation
Built on your history and tested on past cases it never saw. For a no-show model, we check how many of the 100 patients it ranked highest went on to miss their appointment.
Weeks 3–5
Into the workflow
Scores show up where staff already look, such as the EHR worklist, the CRM or a ranked list each morning.
Handoff
Monitoring and documentation
A check that warns you when the model's accuracy starts to slip, plus documentation your team can maintain.
Tyler has built this before
He designed and implemented the customer health score that a healthcare software company uses to track how each client account is doing; 97% of its clients renew each year. He also built a churn model that a national distributor launched across the country.
Price
One fixed price, written into your proposal. After launch there's no monthly fee from us; support is optional at $500 to $1,500 a month.
One model
For one decision staff make every day
Price: from$18,000
Up to $35,000, depending on how much cleanup your data needs and how scores reach staff.
One time
4–6 weeks
A model for one decision, such as which patients to remind first, scored where your staff work.
What's included in One model
- A week-one check that your data can support the model
- A model trained on one to two years of your history
- Accuracy tested on past cases and reported in plain numbers
- Accuracy and bias checked for each patient or customer group
- Scores in your EHR worklist, CRM or a daily list
- A check that warns when accuracy slips, documentation and 30 days of fixes
Payment and terms
- Under $20,000, half at signing and half at delivery. From $20,000, 40% at signing, 30% at validation and 30% at acceptance
- Paying by purchase order? There's no deposit: we invoice half at kickoff and half at acceptance, or from $20,000, 25% at kickoff, 35% at validation and 40% at acceptance
- If week one shows your data can't support the model, we stop there and tell you why in writing, and nothing beyond the signing payment is due
- Bias and accuracy reported for each patient or customer group
- 30 days of fixes after acceptance; support after that is optional at $500 to $1,500 a month, priced by the size of what we built. It covers the check that warns when accuracy slips, fixes, updates when a source system changes, and replies by the next business day
- You own the model and code on final payment
Start and finish dates are set in writing in your proposal.
Tyler Kent does this work himself. He has 13+ years in data science and analytics, most of it in healthcare. About Tyler
Questions
How much data do we need?
Usually one to two years of history for the outcome you want to predict. The first week confirms whether you have enough. If it shows your data can't support the model, we stop there and tell you why in writing, and nothing beyond the signing payment is due.
How will we know it works?
Before anyone relies on it, it's tested on past cases it never saw, and you get the result in plain numbers. Accuracy is also reported for each patient or customer group, so you'll see if it works worse for some of them.
Will it work with our EHR or CRM?
Scores can be written back into your EHR or CRM if it accepts them, or delivered as a daily list. We agree with you in week one on how the scores reach your staff.
Does it need patient data?
Clinical models do, so they start once a BAA and our insurance are in place. A no-show model reads each patient's appointment history, so it needs a BAA too. Customer models, such as a renewal model built on account data, don't need patient data.
What about forecasting census or visit volume?
Capacity and Staffing Forecast is built for that. It starts with a backtest, which runs the forecast on your own past counts, so you know it beats your current method before you pay for the larger build.
What happens after launch?
The model and code are yours on final payment, with a check that flags when accuracy drifts. Thirty days of fixes are included, and support after that is optional at $500 to $1,500 a month.
How an engagement worksQuestions about prices, payment and patient data
Book a 20-minute call
Tell us which decision you would make differently if you saw it coming. If we can help, you'll get a fixed-price proposal within 48 hours.
Prefer to skip the call? Get a written proposal
Or email tyler@kentapplied.com, or connect on LinkedIn.