2023-10-02 Architecture WG Meeting Notes
Date
ZOOM Meeting Information:
Monday, Oct. 2, 2023, at 11:30am PT/2:30pm ET.
Join Zoom Meeting
Meeting ID: 790 499 9331
Attendees:
- Sean Bohan (openIDL)
- Jeff Braswell (openIDL)
- Peter Antley (openIDL)
- David Reale (Travelers)
- Josh Hershman (openIDL)
- Yanko Zhelyazkov (Senofi)
- Faheem Zakaria (Hanover)
- Satish Kasala (Hartford)
- Tsvetan Georgiev (Senofi)
- Ken Sayers (AAIS)
- Ash Naik (AAIS)
- Brian Mills (AAIS)
Agenda:
- How the analytics node functions in the PoCs vis-a-vis the state regulators
Notes:
- ND = 1 DOI
- would have EP
- would submit EP
- Carriers return data to Analytics node
- AN would combine data, generates report
- DOI gets report
- NE CAT = N States (10?)
- how to run POC to make it as effective as possible
- immediate - asking for conssitent set of data covering different areas
- doesn't make sense to have nodes for states
- appropriate approach - analytics node performing data call request to collect data,
- so carriers provide data for mult states
- carriers prep data per state
- only need one AN to provide interfaces for the state
- as an add-on, one or two standard, possibility what it would look like for DOI to make a request
- simple filter
- limitations on report they get back
- AN gets data from carriers from var states, perform reports on behalf of states
- Will every DOI get the same report?
- format of the report is the same, relative data is the same, EP is the same, results look the same, report generator doesn't care about state
- thing that scale, amount of effort, different EP returns same results - if diff formatted report, requires work
- format/output of the report - standardize it to make it trackable
- possible to put filters on input data, "x coverage" or "y locations" filters that effect how report processed
- when you have to change code on the report it is more work
- thinking: EP is a collection of data for subsequent reporting, the state Regs could request diff filters of data
- first step, content more than format
- carriers - dont want DOIs to have a dash with BI capabilities
- for POC, this data is for the POC, one time use of the data
- lets say Maryland running EP, will they need to have an end clause to list out the counties? Or "in the state of Maryland and then AN filters out counties"
- if it is for a particular cat, ID areas in state by cty/zip where damage occured, thats in EP, filtering happens after that
- because CG will be involved in doing the work to help, timely, bring them into the picture
- data standards to put out will also inv some input from states
- good idea of what things look like, NAIC standards there, additioonal data elements in play
- conceptually how we approach, formats not locked
- Reports are ephemeral - dont have anything saved outside of carrier's node
- dont want stitching together of various data calls and making an external data store to ask other questions
- Need clarification from Dale - not just approving query - also THE STATED PURPOSE
- "looking at x data in this format, for y purpose"
- not just approving mechanics of extraction, approving data product
- Regulators can ask for a specific purpose
- every data call in the POC will be for this POC, strictly for that report, blanket covenant
- stated code for purpose of data call
- can be process instead of tech solution
- not just an FYI - binding, this is what it is used for, like a TOS
- shouldnt be filtering in the data analytics node
- only what it is said to do
- flexible means needed so they can ask for what they want
- involve in next conv on data side, get regs involved, fact we will have help
- more news ssoon on that - agree on data elements
- what output format to DOI?
- default model at the moment - want to have that conversation, get input from the states
- excel or PDF?
- multiple DOIs dont get each other's reports
- pull it out of the S3 bucket
- exclusivity? Only so DOI could see it, didnt make it for mult DOIs, wont see if you dont consent
- constraint - not assoc with a state you cannot consent to it
- designed right now - each DOI would have own node, in this we would simulate all DOIs having own node
- issue - not really serving the function as a proof - would have to set up mult nodes
- not really doing what it says it is doing if there are caveats
Time | Item | Who | Notes |
---|---|---|---|