TimeH2O Claims Planning
  1. iThe problem
  2. ⚙Set up
  3. 1Forecast claims
  4. 2Plan the staff
  5. 3Eight weeks later
  6. ✓Results
</>Source code All use cases

Claims Operations · Monday 09:00

TimeH2O forecasts next week's claims and cost, so you can staff for them.

What TimeH2O doesIt reads your weekly claim counts and costs. It forecasts each coming week, with a range. You choose how many adjusters to staff.

40 daysto close a claim. Most of that time it waits in a queue.
Too fewadjusters, and the queue grows
Too manyadjusters, and people sit idle
Weekly claims 5 years · counts and costs Known holidays in the weeks to come TimeH2O Reads your past weeks No training step 1 Forecast 8 weeks, with a range 2 Plan Staff for the high side 3 Result Average wait 3 days not 11 days

The app uses sample data made for this example. The numbers show how the steps work. They are not results from a customer.

Claims planning · Claims Operations · for a claims operations manager

Know next week's claim volume and cost, and staff for it.

The problemA claim can take 40 days to close. Most of that time it waits in a queue, because nobody knows how many claims will arrive next week or how much they will cost.
What TimeH2O doesIt reads 5 years of weekly claim counts and costs. It forecasts claims and cost per week, by class, for the next 8 weeks, with a range. There is no training step.
What you getYou staff adjusters for the side of the range you choose. Finance uses the cost range to plan the reserve. The actuary still sets the final reserve.

Set up the run

Choose what TimeH2O forecasts and your staffing plan.

ClassesAbout this many new claims a week
History TimeH2O readsWeekly claim counts and costs
5 years · 260 weeks for each class · rows
ForecastAnd the width of the range
The next 8 weeks, with an 80% range: 8 of 10 real weeks fall inside it
Staff home claims forAn adjuster takes on about 25 new home claims a week

The staffing plan follows home claims, as in the story. You can change the plan again in step 2.

Step 1 of 3

TimeH2O is forecasting the next 8 weeks

What TimeH2O doesIt reads the past weeks of each class. It sees the rise each winter, when pipes freeze and burst, and the dip in holiday weeks. Then it forecasts each coming week, with a range.

Claims expected, next 8 weeks–
Home claims expected–
Home claims cost expected–

New claims a week, by class

Past weeks Middle forecast 80% range Click a class to see its forecast.

Step 2 of 3

Use the TimeH2O range to choose how many adjusters to staff

What TimeH2O doesIt gives the middle forecast and the range for each week. You choose which side of the range to plan for. TimeH2O does not make the decision. Your plan does.

Home adjusters a week–
Adjuster-weeks paid–
Claims the team can take, 8 weeks–

Home claims a week

Past weeks Middle forecast and 80% range Claims the team can take Click a week to see it.

Step 3 of 3

The real claims arrive, week by week

What TimeH2O doesIt made its forecast on day one. Now the 8 weeks play out with the real claims. A red dot marks a busy week: more claims than the team can take. Those claims wait.

Weeks covered–
Busy weeks–
Average wait for an adjuster–

Home claims a week

Real claims, covered Busy week Claims the team can take Click a week to see it.

Results · 8 weeks later

Average wait for an adjuster11–
Weeks covered–
Adjuster-weeks paid–

Every plan on the same 8 weeks

Staff forHome adjusters a weekAdjuster-weeks paidBusy weeksAverage wait

If you staff for the high side, claims rarely wait, but some adjusters are free in a quiet week. If you staff for the middle, you spend less, but in a busy week claims wait. TimeH2O gives the forecast. Your team makes the decision.

The app uses sample data made for this example. The numbers show how the steps work. They are not results from a customer.

Source code · Python

Run this use case with TimeH2O in your own environment.

What the code doesIt makes sample data like this app: 5 years of weekly claims for 5 classes. TimeH2O forecasts the next 8 weeks with a range. Then the code staffs adjusters for the high side. To use your own data, change step 1 in the code to read your CSV.

1
Get a free API key at tabh2o.h2oai.com, then set itexport TABH2O_API_KEY=<your key>
2
Install the packagespip install pandas numpy requests
3
Save the code below as a file and run itpython claims_planning.py
claims_planning.py129 lines

  

  
The API runs TimeH2O for you, so you do not need a GPU. The code makes its own sample data, so its numbers differ from the numbers in this app.