Frequently Asked Questions


How does line of business and linetype prediction work?

How does the claim combination logic work in loss run files?

How does Groundspeed detect duplicate loss run claims?

How does Groundspeed detect duplicate loss run summary records?

How does Groundspeed identify loss run claim occurrences?

What insurance documents does Groundspeed support?

What documents doesn't Groundspeed support?

Why do some files fail to ingest?

What ACORD forms doesn't Groundspeed support?

I sent a ZIP file and then later sent another, but they're not coming back under the same account?

How is data populated in extracts for fields that can be drawn from multiple documents?


LOB/Linetype Prediction

How does line of business and linetype prediction work?

The Line of Business (LOB) and Linetype Fields

For claims, Groundspeed populates two related fields in the loss detail output to describe the relevant class of insurance business and coverage: line of business and linetype.

Line of business describes the class of insurance business the claim pertains to, and can include umbrella, package, or excess lines.

Linetype describes the underlying coverage type relevant to the claim, and does not include umbrella, package, or excess classifications.

In practice, these fields will always be the same unless the LOB is umbrella, package, or excess. In which case, the linetype will specify the underlying coverage type relevant to the claim.


How Groundspeed determines LOB and Linetype

The Groundspeed pipeline uses capture values, heuristics, and machine learning models to identify the line of business and linetype for a given claim in the following order:

  1. Utilizing the Coverage Type captured from original Loss Run document:
    1. If a coverage type value is present in the document we will capture it and normalize the data. Example: For the following input values: GL, CGL, General Liability, the output will show General Liability.
    2. If a value is captured but cannot be normalized, it will display in the “Line of Business Raw” field, but #2 or #3 below will be used for the “Line of Business” and “Linetype” Fields.
      1. For example “Liability,” which is too vague to be assigned a normalized value.
  2. Predict using one of a set of heuristics:
    1. Policy/Claim Number: if policy or claim number prefix is “WC” or “GL,” then the LOB/Linetype will be set accordingly.
    2. Monoline WC carrier: if carrier tied to claim is recognized as a WC-only carrier, lob and linetype will be set to WC for that claim. Groundspeed maintains a list of carriers in this space and frequently updates that list.
    3. Losstype: if products completed, premises, or prem/ops captured as a claims losstype, a lob/linetype of “General Liability” will be inferred.
  3. Predict using Groundspeed AI:
    1. When there is no explicit coverage type in the original document and no heuristic applies, Groundspeed uses a series of models to predict LOB and Linetype values.
    2. These models leverage the entire captured claim record to assign a lob/linetype and are trained using our vast contributory dataset.
    3. If a predicted value is low confidence, the models will abstain from making a prediction, leaving the lob/line type as unknown.
      1. For example, we often see claims with no claim description or phrases like “N/A”.
    4. Given that our contributory dataset reflects the real world with a mix of various lines, some lines are better represented in the data than others. Those lines with higher volumes provide more consistent and high confidence predictions, while those lower volumes tend to result in lower confidence scores.



Claim Combination Logic

How does the claim combination logic work in loss run files?

Overview

The Groundspeed claim combination logic combines individual claim records that belong to one accident but were split in the source document.

To enable combination the logic performs two steps: identification of claims to merge and combining the data from the claim records.

Rules

  • Claim Is Summary = FALSE
    • Description: A record must be a claim detail record and not a summary record.
  • The below list of fields must match to combine claims:
    • GS Occurrence ID
    • File ID
    • Line of Business
    • Carrier
    • Policy Number
    • Claim Number
    • Claimant Name
    • Date of Loss
    • Loss Description
    • Loss State
    • Insured Name

Record Combination Logic

  • All financial values from the records are summed.
    • Indemnity Paid
    • Indemnity Reserves
    • Indemnity Incurred
    • Medical Paid
    • Medical Reserves
    • Medical Incurred
    • Rehabilitation Paid
    • Rehabilitation Reserves
    • Rehabilitation Incurred
    • Expense Paid
    • Expense Reserves
    • Expense Incurred
    • Total Paid
    • Total Reserves
    • Gross Total Incurred
    • Total Reserves
    • Net Total Paid
    • Net Total Reserves
    • Net Gross Total Incurred
  • All non-financial values are retained and merged into the final, singular record.
    • Example: Claimant Age will return the age as is without summing the values.



How does Groundspeed detect duplicate loss run claims?

Overview

The Groundspeed claim detail duplication detection logic works to identify individual claim records that are duplicates.

Rules

All fields must match in the record to detect a duplicate:

  • Account ID - the account to which the claim belongs
  • Line of Business - the line of business of the claim
  • Carrier - the carrier responsible for the claim
  • Claim Number - the cleaned claim number
  • Policy Year - the policy year of the claim
  • Date of Loss - the date on which the claim occurred
  • Valuation Date - the date on which the claim financials were updated
  • Claimant Name - the name of the claimant
  • Claimant Number - the number of the claimant
  • Loss Type - the classification of the claim within the LOB
  • Medical Incurred - the incurred medical financials for a claim
  • BI Incurred - the incurred bodily injury financials for a claim
  • PD Incurred - the incurred property damage financials for a claim
  • Expense Incurred - the total incurred expense financials for a claim
  • Indemnity Incurred - the incurred indemnity financials for a claim
  • Indemnity Reserves - the indemnity reserve financials for a claim
  • Indemnity Paid - the indemnity paid financials for a claim
  • Gross Total Incurred - the total incurred financials for a claim
  • Total Reserves - the total reserves financials for a claim
  • Total Paid - the total paid financials for a claim



How does Groundspeed detect duplicate loss run summary records?

Overview

The Groundspeed claim summary duplication detection logic works to identify claim summary records that are duplicates.

Rules

  • All fields must match in the record to detect a duplicate summary record:
    • Line of Business - the line of business of the claim
    • Carrier - the carrier responsible for the claim
    • Policy Year - the policy year of the claim
    • Summary Claim Count - the number of claims listed for the summary record
    • Policy Number - if policy number is present, policy number is included in the mandatory fields. If policy number is not present it is not included in the mandatory fields
  • At least 1 of these fields must match to detect a duplicate summary record:
    • Total Paid
    • netTotal Paid
    • netGross Total Incurred
    • Gross Total Incurred



How does Groundspeed identify loss run claim occurrences?

Overview

The Groundspeed claim occurrence identification logic aims to identify a singular occurrence from which claims originated.

Examples include an auto accident, a hail storm, or an explosion at a factory which resulted in multiple claims.

Rules

  • All fields must match in the record to identify an occurrence:
    • Account ID - the account to which the claim belongs
    • Line of Business - the normalized line of business of the claim
    • Carrier - the carrier responsible for the claim
    • Policy Number - the cleaned policy number of the claim
    • Policy Year - the policy year of the claim
    • Loss State - the loss state in which the claim occurred
    • Date of Loss - the date on which the claim occurred
  • At least 1 of these fields must match in the record to identify an occurrence:
    • Claim ID - identifier of the claim
    • Occurrence ID - identifier of the occurrence
    • Driver Name - the name of the driver
    • Claimant Name - the name of the claimant
    • Loss Description - the description of the loss
    • Claim Number - the cleaned claim number



What insurance documents does Groundspeed support?

Product Type Support Document Type
Setup and Clearance/Submission
    Email
    ACORD Forms
Loss Runs
    PDF Loss report
    Excel/CSV loss report
Exposures
    Statements of Value
    Schedule of Hazards
    Driver List
    Vehicle Schedule
    Payroll Schedule
    Location Schedule
    Revenue Schedule
    Aircraft Schedule
    NCCI Emod
    CA WCRIB Emod
    ACORD Forms
Emod
    NCCI Emod
    CA WCRIB Emod

Supported Acord Forms


Note: All versions and variants of each form will be supported. This includes state-specific variants, old form versions, and ACORD forms generated from an agency management system. For example: "Applied 130API," "INS 127," and "137 TX."


Form Number Notes
27
45
101
125 Including: "ANI" and "CIS" supplements
126
127
129
130 Including: "ANI" supplement
131
137 Including: State-specific variants, ex. "137 FL"
139 Including: CISGEM C540 SOV supplement
140
141
143
146
147
148
152
160 Including: "ANI" supplement
163
211
810
823
98
OFCLADCOV Additional Endorsement/Supplement
OFGARVEH Additional Endorsement/Supplement
OFSOICOV Additional Endorsement/Supplement
OFAPPINF Additional Endorsement/Supplement
OFREMARK Additional Endorsement/Supplement
OFCONINF Additional Endorsement/Supplement
OFAPPINFCNI Additional Endorsement/Supplement
OFBAADCV Additional Endorsement/Supplement
OFSUPPNAME Supplemental names of other named insureds
APDWCLOC Applied WC form associated with the ACORD 130
OFWCRATE AMS WC form associated with the ACORD 130



What documents doesn't Groundspeed support?

Out-Of-Scope Type Description Examples
Unsupported Document File extensions or file types that are not machine-readable, OR are not relevant to commercial insurance.
These documents will likely never be on our product roadmap
    PPTX
    TXT
    Signature Block Images
    Audio, Video files
    Executables, scripts
    Password protected
    Low Image Quality
    No data - empty files
    Incorrect MIME/Content type
    Files with handwriting
    Files that do not pass virus scanning checks
    Company Brochure
    White Paper
    Marketing Materials
    Any Personal Lines Submission



Why do some files fail to ingest?

Error Code Error Description
100 unsupported_file_type The archive file type sent to our intake process was not on the list of supported file types
101 failed_virus_scan The file, or the file container sent failed the virus scan step, and was quarantined
102 bad_zip_file The file extension was a zip, however the file itself could not be opened
106 unsupported_document_type No files within the zip are supported



What ACORD forms doesn't Groundspeed support?

This list only covers commercial lines ACORD forms that we consider out-of-scope. All personal lines ACORD forms are out-of-scope at the moment.

  • Form ID: 143, 88, 89, 145, 138 TX, 128, 125PCIS, 125LHS, 127COV, 130PCS, 131CCCS, 133, 133FL, 133TN, 144, 155BM, 188, 194, 25, 39NY, 60US, 61FL, 62FL, 63, 75, 815, 829, 829CLAP, 829LOB, 832.



I sent a ZIP file and then later sent another, but they're not coming back under the same account?

  • An account is created for each individual ZIP file. If 2 ZIP files were sent, they would each get their own unique accountId and be returned as separate outputs.



How is data populated in extracts for fields that can be drawn from multiple documents?

Groundspeed uses doctype precedence to determine which document a field is pulled from. If that field is not present in any of the doctypes listed then that field will appear blank on the extract.


Example: “insuredName” in the submissionData section of standard submission extracts.


insuredName has a doctype preference order of slip, email, acord, policyform, brokerspec, exposuredoc, sov, application, and locationschedule.


Scenario: An insured name of “Acme Corp Inc.” is found on a slip doctype and an insured name of “Acme Corp” is found on a brokerspec in the same submission. Because the slip doctype precedes the brokerspec in the doctype precedence rules set for this field, the insured name presented on the extract would be “Acme Corp Inc.”