Past Analytics Excellence Awards

Suggest an idea

Author: Jim Kunce, SVP & Chief Actuary

Company: MedPro Group

 

Awards Category: Best Use of Server

 

Describe the problem you needed to solve 

MedPro Group, Berkshire Hathaway's dedicated healthcare liability solution, is the nation's highest-rated healthcare liability carrier - according to A.M. Best (A++ as of 5/27/2015). We have been providing professional liability insurance to physicians, dentists and other healthcare providers since 1899. Today, we have insurance operations in all 50 states, the District of Columbia and are growing internationally. With such great size of operations and diversity of insurance products, it is a challenge to connect systems, processes and employees with one another.

 

Regardless of an insurance carrier's size and scale, its long-term success depends on:

 

  • Continued new business growth
  • Consistent pricing and risk-evaluation
  • Unified internal operations

Our challenge was to lay the analytical foundation necessary for an ever-growing insurance company to execute on these three objectives. We identified the following three action items and linked them to the drivers of long-term success.

 

  • Fuel new business growth by:  Centralizing processes & remove system silos, link manual processes together.
  • Drive consistent pricing and risk-evaluation: Remove data supply bottle-necks & empower business analysts to self-serve.
  • Unify internal operations: Accelerate modernization & facilitate enterprise-wide legacy system integration.

 

Describe the working solution

"Fuel new business growth by centralizing processes & remove system silos, link manual processes together."

 

This solution has three parts to it.

  • First, we programmed our pricing algorithm using Alteryx to "learn" the insurability of a prospective customer.
  • Second, we overlaid this system on our CRM data to create sales recommendations nationwide.
  • Third, we deployed this recommender system with our Alteryx private gallery to provide real-time access to our sales teams.

MPG Private Gallery Snapshot.jpeg

 

Today, from anywhere in the country, our sales personnel can request a report for a customer they are prospecting and receive a consistent, reliable recommendation in a matter of seconds with little manual intervention.

 

"Drive consistent pricing and risk-evaluation:  Remove data supply bottle-necks & empower business analysts to self-serve."

 

In an insurance company, actuaries and underwriters are responsible for pricing insurance policies and evaluating insurance risks of applicants. These complex decisions rely on many data inputs - some of which are internally available, but in other cases come from external sources (e.g. government websites, third party resources).

 

Today, we have been able to significantly reduce the data supply bottle-necks by configuring the Alteryx server to be the bridge between the data sources and our actuaries and underwriters. Each person along the pricing and risk evaluation process now gets “analysis-ready” data consistently and timely from the private gallery, a virtual buffet of self-serve apps for all data needs.

 

"Unify internal operations: accelerate modernization -- facilitate enterprise-wide legacy system integration."

 

In 2015, MedPro Group decided to scale up investments in modernizing legacy systems to a new web-based system. The challenge was to move data from our legacy systems into the new web-based system and vice versa. Additionally, the software solution needed to have a short learning curve and be flexible and transparent enough that key business leaders managing this modernization would be able to perform the data migration tasks.

 

Alteryx was a great fit in this case. Not only were business leaders able to program processes in Alteryx in a relatively short timeframe, we scaled up with ease and accelerated modernization by deploying on the private server for analysts to use in a self-serve, reliable environment.

 

Describe the benefits you have achieved

"We have connected systems, processes and employees to one another and made the benefits of that interconnectivity available to every employee."

 

We have been using the private gallery and server since July, 2015. What started as a proof of concept and experiment is now a fully functional production-grade experience. The list of systems that have been connected, processes that have been automated and employees who are finding value out of our private gallery and server is growing rapidly.

 

Here's a view into some of the measurable benefits we have achieved in just nine months -

  • 94: The number of apps published to the private gallery to date.
  • 6951: The number of times an app has run on the gallery. That's 26 runs a day over 9 months!
  • 15: The percentage of employees who are served with this consistent, reliable self-serve platform.

 

And our goal? Move that needle to 100% with Alteryx in the months to come!

Author: Michael Barone, Data Scientist
Company: Paychex Inc

Awards Category: Best Use of Predictive

 

Describe the problem you needed to solve

Each month, we run two-dozen predictive models on our client base (600,000 clients). These models include various up-sell, cross-sell, retention, and credit risk models. For each model, we generally group clients into various buckets that identify how likely they are to buy a product/leave us/default on payment, etc. Getting these results into the hands of the end-users who will then make decisions is an arduous task, as there are many different end-users, and each end-user can have specific criteria they are focused on (clients in a certain zone, clients with a certain number of employees, clients in a certain industry, etc.).


Describe the working solution

I have a prototype app deployed via Alteryx Server that allows the end-user to “self-service” their modeling and client criteria needs. This is not in Production as of yet, but potentially provides great accessibility to the end-user without the need of a “go-between” (my department) to filter and distribute massive client lists.

 

Step 1: ETL

  • I have an app that runs every month after our main company data sources have been refreshed:

51.png

This results in several YXDBs that are used in the models. Not all YXDBs are used in all models. This creates a central repository for all YXDBs, from which each specific model can pull in what is needed.

  • We also make use of Calgary databases as well, for our really large data sets (billions of records).

52.png

Once all the YXDBs and CYDBs are created, we then run our models. Here is just one of our 24 models:

53.png

  • Our Data Scientists like to write raw R-code, so the R tool used before the final Output Tool at the bottom contains their code:

54.png

The individual model scores are stored in CYDB format, to make the app run fast (since the end-user will be querying against millions and millions of records). Client information is also stored in this format, for this same reason.

 

Step 2: App

  • Since the end-user will be making selections from a tree, we have to create the codes for the various trees and their branches. I want them to be able to pick through two trees – one for the model(s) they want, and one for the client attributes they want. For this app, they must choose a model, or no results will be returned. They DO NOT have to choose client attributes. If no attribute is chosen, then the entire client base will be returned. This presents a challenge in key-building, since normally an app that utilizes trees only returns values for keys that are selected. The solution is to attach keys to each client record for each attribute. My module to build the keys in such a way as I described is here (and there will be 12 different attributes from which the user can choose):

545.png

  • Here is what the client database looks like once the keys are created and appended:

56.png

  • The model keys do not have to be as complex a build as client keys, because the user is notified that if they don’t make a model selection, then no data will be returned:

57.png

  • Once the key tables are properly made, we design the app. For the model selection, there is only one key (since there is only one variable, namely, the model). This is on the far right hand side. This makes use of the very powerful and fast Calgary join (joining the key from the pick-list to the key in the model table). For the client table, since there are 12 attributes/keys, we need 12 Calgary joins. Again, this is why we put the database into Calgary format. At the very end, we simply join the clients returned to the model selected:

58.png

 

Step 3: Gallery

  • Using our private server behind our own firewall, we set up a Gallery and Studio for our apps:

59.png

  • The app can now be run, and the results can be downloaded by the end-user to CSV (I even put a link to an “at-a-glance” guide to all our models):

591.png

  • The user can select the model(s) they want, and the scores they want:

592.png

And then they can select the various client criteria:

593.png

Once done running (takes anywhere between 10 – 30 seconds), they can download their results to CSV:

594.png

 

Describe the benefits you have achieved

Not having to send out two dozen lists to the end-users, and the end users not having to wait for me to send them (can get them on their own).  More efficient and streamlined giving them a self-service tool.

Author: Mandy Luo, Chief Actuary and Head of Data Analytics

Company: ReMark International

 

Awards Category: Best Use of Predictive

As a trained Statistician, I understand why "70% data, 30% model" is not an exaggeration. Therefore, before applying any regression models, I always make sure that input data are fully reviewed and understood. I use various data preparation tools to explore, filter, select, sample or join up data sources. I also utilize the data investigation tools to conduct or validate any statistical evaluation. Next, I would usually choose 3-5 predictive modeling candidates depending on the modeling objective and data size. I often include one machine learning methods in order to at least benchmark other models. After the modeling candidates finish running, I would select the best model based on both art (whether the coefficients look reasonable based on my understanding of the data and business) and science (statistical criteria's like the goodness of fit, P-value and cumulative lift etc.).  I am also often using the render function for model presentation and scoring/sorting  function for model validation and application.

 

Describe the problem you needed to solve 

ReMark is not only an early adopter in predictive modeling for life insurance, but also a true action taker on customer centricity by focusing on customer lifetime analytics (instead of focusing on 'buying' only). In this context, we need to 'join up' our predictive models on customer response, conversion and lapse in order to understand the most powerful predictors that drive customer activities across pre and post sales cycle. We believe the industry understand that it is insufficient to only focus on any single customer activity, but is still exploring how this can be improved through modeling and analytics, which is where we can add value.

 

Describe the working solution

Our working solution goes with the following steps:

  1. Match over one year post sales tracking data back to sales payment data and marketing data (all de-personalized)
  2. Build 3 predictive models: sale(whether the purchase is agreed or not), conversion (whether the first premium bill is paid or not), 1 year persistency (whether lapse happened at month 13 or not).
  3. Compare model results by key customer segments and profiles
  4. Expert to visualization tool (e.g. Tableau) to present results
  5. Model use test: scoring overlay and optimization strategy

 

Describe the benefits you have achieved

  • We could create customer segments' not just based on tendency to 'buy', but also tendency to 'pay' and 'stay'.
  • We could further demonstrate ReMark's analytics and modeling capabilities covering the whole customer lifetime value chain without missing out