Data store · SCIKIQ Datahub

Every number traces back to the file it came from

Why this matters to IIB. Every number in the demo traces back to the file it came from. This page shows the data as received, as checked and as ready to use, with every check and definition.

Data moves in three steps: as received, checked, and ready to use. Reference lists, public data and working tables sit alongside.

As received

23 tables

17,73,289 rows

Checked

21 tables

17,92,290 rows. Includes the result of every check, kept for audit.

Ready to use

28 tables

24,59,858 rows

Checks

22 must pass, 9 warnings

Built 7 Oct 2026, 09:37. Data as at 30 September 2026. Weekly claims from 7 October 2024 to the week starting 28 September 2026.

As received (23)

Files exactly as they arrived, with every value kept as text. Each policy and claim row notes its file and load time. The file receipt log keeps each file's fingerprint.

Claims files`raw_claims`	37,820
Claims since the September files`raw_claims_feed`	316
raw_health_claims	1,01,926
raw_health_hospitals	900
raw_health_policies	4,00,000
raw_health_procedures	25
raw_health_products	30
raw_health_weather	12,792
Insurer list`raw_insurers`	25
File issue log`raw_issue_log`	16
raw_life_agents	6,000
raw_life_death_claims	2,159
raw_life_policies	2,19,863
raw_life_products	12
File receipt log`raw_load_log`	1,202
Policy files`raw_policies`	6,02,123
raw_property_claims	1,821
raw_property_insurers	25
raw_property_policies	90,087
Opening policy book`raw_repository_policies`	2,94,685
RTO list`raw_rtos`	261
Sample design`raw_sample_design`	1
What each insurer said it sent`raw_submissions`	1,200

Checked (21)

Turned into dates and numbers and checked. Rows that fail a must-pass check are held back, not loaded. The result of every check is kept.

Claims, checked`cur_claims`	38,136
Results of file checks`cur_dq_file_results`	6,000
Results of row checks`cur_dq_results`	15,626
cur_health_claims	1,01,926
cur_health_hospitals	900
cur_health_policies	4,00,000
cur_health_procedures	25
cur_health_products	30
cur_health_quarantine	0
cur_health_weather	12,792
cur_life_agents	6,000
cur_life_death_claims	2,159
cur_life_policies	2,19,863
cur_life_products	12
cur_life_quarantine	0
Policies, checked`cur_policies`	8,94,724
cur_property_claims	1,821
cur_property_insurers	25
cur_property_policies	90,087
cur_property_quarantine	0
Rows held back, one line per check failed`cur_quarantine`	2,164

Ready to use (28)

The checked tables that people and the AI use. Ask the data can read only these.

Road accidents (MoRTH)`srv_accidents`	249
Claim frequency by month`srv_claim_frequency`	12,332
Claims by week`srv_claims_weekly`	36,131
Health claims by month`srv_health_claims_monthly`	80,351
Health claims by week`srv_health_claims_weekly`	97,528
Hospital claims and costs by month`srv_health_hospital_monthly`	17,249
Procedures billed by each hospital`srv_health_hospital_procedures`	10,379
Network hospitals`srv_health_hospitals`	900
Procedure packages`srv_health_procedures`	25
Health products and their terms`srv_health_products`	30
Weather and illness by week (mock weather)`srv_health_weather_weekly`	12,792
Life agents by month`srv_life_agent_monthly`	23,195
Life agent registry`srv_life_agents`	6,000
Life new business cohorts`srv_life_cohorts`	4,594
Life death claims by month`srv_life_death_claims_monthly`	2,078
Life death claims by week`srv_life_death_claims_weekly`	2,095
Life policies by month`srv_life_policies_monthly`	1,49,168
Life products`srv_life_products`	12
Policies by month`srv_policies_monthly`	7,65,830
Property claims by week`srv_property_claims_weekly`	1,784
Property exposure by district`srv_property_exposure_district`	860
Property exposure on a map grid`srv_property_exposure_grid`	4,047
Property exposure by week`srv_property_exposure_weekly`	11,84,260
Property premium by month`srv_property_premium_monthly`	39,785
Vehicle registrations`srv_registrations`	4,140
Theft claims with policyholder`srv_theft_claim_holders`	3,614
Motor vehicle theft (NCRB)`srv_theft_ncrb`	250
Uninsured estimate`srv_uninsured_estimate`	180

Work in progress (4)

Rows turned into dates and numbers, with a pass or fail for every check. Kept so every held-back row can be traced.

Check flags for claims`dq_flags_claims`	38,136
Check flags for policies`dq_flags_policies`	8,96,808
Claims as dates and numbers`stg_claims`	38,136
Policies as dates and numbers`stg_policies`	8,96,808

Lookup lists (3)

States, RTOs and insurers.

Insurers`ref_insurers`	25
RTOs (Regional Transport Offices)`ref_rtos`	261
States and union territories`ref_states`	36

Public data (6)

VAHAN, NCRB, MoRTH and IRDAI, checked and put in one shape, with where each came from.

IRDAI motor premium`open_irdai`	11
MoRTH road accidents`open_morth`	249
MoRTH registered vehicle stock`open_morth_stock`	1,680
NCRB motor vehicle theft`open_ncrb`	250
VAHAN vehicle registrations`open_vahan`	14,076
Where each source came from`provenance`	5

Ready to use: the only tables Ask the data can read

Policies by monthsrv_policies_monthly

One row per state, RTO, vehicle class, cover type, insurer and month. Source: synthetic insurer data. Policies in force at each month end, policies issued in the month and premium written. October 2024 to September 2026.

`month`	First day of the month, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code, for example TS for Telangana.
`state_name`	State or UT name, for example Telangana.
`rto_code`	Regional Transport Office (RTO) where the vehicle is registered, for example TS09 in Hyderabad.
`vehicle_class`	2W (two-wheeler), PCAR (private car), GCV (goods carrying), PCV (passenger carrying) or MISC.
`cover_type`	COMP (own damage and third party), TP (third party only) or SAOD (standalone own damage).
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`policies_in_force`	Policies in force at the month end.
`new_policies`	Policies issued in the month, new business and renewals.
`premium`	Total premium written in the month, rupees.

Claims by weeksrv_claims_weekly

One row per state, RTO, vehicle class, claim type, insurer, week and submission month. Source: synthetic insurer data. Claims by the week they were reported. Complete weeks only, Monday to Sunday, from 2024-10-07 to the week ending 2026-10-04. A week that spans two months is split by cycle.

`week_start`	Monday of the week the claims were reported, a DATE. Use it for weekly questions.
`cycle`	Submission cycle, the month the claim was reported, as text YYYY-MM such as 2026-09. Use it for monthly, quarterly and per-cycle questions about claims by RTO or insurer.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`rto_code`	Regional Transport Office (RTO) where the vehicle is registered, for example TS09 in Hyderabad.
`vehicle_class`	2W, PCAR, GCV, PCV or MISC.
`claim_type`	OD (own damage), TP (third party) or THEFT.
`insurer_code`	Insurer code. Shown as a name or an alias, by role.
`claims`	Number of claims reported in the week.
`claim_amount`	Amount claimed, rupees.
`paid_amount`	Amount paid so far, rupees.

Claim frequency by monthsrv_claim_frequency

One row per state, vehicle class, claim type and month. Source: synthetic insurer data. Claims reported in each month against policies in force with eligible cover. OD and THEFT use COMP and SAOD policies. TP uses COMP and TP policies.

`month`	First day of the month, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`vehicle_class`	2W, PCAR, GCV, PCV or MISC.
`claim_type`	OD, TP or THEFT.
`claims`	Claims reported in the month.
`claim_amount`	Amount claimed in the month, rupees.
`policies_in_force`	Policies in force at the month end whose cover includes this claim type.
`frequency_per_1000`	Claims per 1,000 eligible policies in force, for the month.

Vehicle registrationssrv_registrations

One row per state, vehicle class and year. Source: VAHAN vehicle registrations. New vehicle registrations from VAHAN, mapped to the five vehicle classes.

`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`vehicle_class`	2W, PCAR, GCV, PCV or MISC.
`year`	Calendar year of registration.
`registrations`	Vehicles registered in the year.

Motor vehicle theft (NCRB)srv_theft_ncrb

One row per state and year. Source: NCRB motor vehicle theft. Motor vehicle theft cases reported to police, from NCRB Crime in India.

`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`year`	Calendar year.
`mv_theft_cases`	Motor vehicle theft cases.

Road accidents (MoRTH)srv_accidents

One row per state and year. Source: MoRTH road accidents. Road accidents, deaths and injuries, from MoRTH Road Accidents in India.

`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`year`	Calendar year.
`accidents`	Road accidents.
`killed`	People killed.
`injured`	People injured.

Uninsured estimatesrv_uninsured_estimate

One row per state and vehicle class. Source: VAHAN vehicle registrations and synthetic insurer data. Vehicles on the VAHAN register (registrations over the last 15 years, an assumed window) minus demo policies in force scaled up to the market, at 30 September 2026, for each state and vehicle class. An estimate, never below zero.

`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`vehicle_class`	2W, PCAR, GCV, PCV or MISC.
`registered_stock`	Registered stock. VAHAN registrations summed over the last N years.
`policies_in_force`	Policies in force at 2026-09-30, scaled to the market by the sampling weight.
`sample_policies_in_force`	Synthetic policies in force before scaling.
`insured_share`	Policies in force (scaled) divided by registered stock, between 0 and 1.
`uninsured_estimate`	Registered stock minus scaled policies in force. Never below zero.

Theft claims with policyholder. Data Operations only. Names always come back hidden.srv_theft_claim_holders

One row per theft claim. Source: synthetic insurer data. Row-level theft claims with the policyholder's name, mobile number and vehicle registration. Data Operations only. Personal data is always masked.

`claim_no`	Claim number.
`policy_no`	Policy number.
`insurer_code`	Insurer code. Shown as a name or an alias, by role.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`rto_code`	Regional Transport Office (RTO) where the vehicle is registered, for example TS09 in Hyderabad.
`vehicle_class`	2W, PCAR, GCV, PCV or MISC.
`loss_date`	Date of the theft, a DATE.
`report_date`	Date the claim was reported, a DATE.
`claim_amount`	Amount claimed, rupees.
`status`	Open, Settled or Repudiated.
`holder_name`	Policyholder name. Personal data, always masked.
`holder_mobile`	Policyholder mobile number. Personal data, always masked.
`reg_no`	Vehicle registration number. Personal data, always masked.

Health claims by monthsrv_health_claims_monthly

One row per month, state, district, hospital type, procedure group and insurer. Source: synthetic insurer data. Health insurance claims by the month of hospital admission, October 2024 to September 2026. A claim's place is the district of the treating hospital. Claims not yet settled on 5 October 2026 have status open. For a rate such as the rejection rate, divide by settled claims (claims minus claims_open). The claim volumes are synthetic, so do not work out claim frequency or loss ratio from them.

`month`	First day of the month of admission, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code of the treating hospital, for example MH for Maharashtra.
`state_name`	State or UT name, for example Maharashtra.
`district_code`	District code of the treating hospital, state code and three letters, for example MH-MUM for Mumbai.
`district`	District name, for example Mumbai.
`hospital_type`	multispeciality, single_speciality, nursing_home or day_care.
`procedure_group`	cardiac, orthopaedic, eye, maternity, general_surgery, infection, respiratory, renal, cancer or newborn.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`claims`	Number of claims (hospital admissions) in the month.
`claims_cashless`	Cashless claims, where the insurer pays the hospital directly, including open ones. The rest are reimbursement claims.
`billed_amount`	Total hospital bill, rupees. Patients still in hospital on 5 October 2026 have no bill yet and add nothing.
`approved_amount`	Total amount the insurer approved, rupees. Open claims add nothing. On top_up products the insurer pays only the part of the bill above the deductible.
`claims_settled_full`	Claims with status paid, settled in full.
`claims_settled_part`	Claims with status partly_paid, settled with a deduction.
`claims_rejected`	Claims with status rejected, nothing paid.
`claims_open`	Claims with status open, not yet settled on 5 October 2026.
`claims_discharged`	Claims whose patient has been discharged, so the bill and length of stay are known.
`avg_los_days`	Average length of stay in days, for discharged patients. Day care counts as 0, or 1 with an overnight stay. To combine rows, weight each by claims_discharged.

Health claims by weeksrv_health_claims_weekly

One row per week, state, district, procedure and insurer. Source: synthetic insurer data. Health claims by the week of hospital admission. Complete weeks only, Monday to Sunday, from 2024-10-07 to the week starting 2026-09-28 (ending 2026-10-04). A claim's place is the district of the treating hospital.

`week_start`	Monday of the week of admission, a DATE. 2024-10-07 to 2026-09-28.
`state_code`	Two-letter state or UT code of the treating hospital.
`state_name`	State or UT name.
`district_code`	District code of the treating hospital, for example DL-EAS for East Delhi.
`district`	District name.
`procedure_group`	cardiac, orthopaedic, eye, maternity, general_surgery, infection, respiratory, renal, cancer or newborn.
`procedure_code`	Procedure package code, for example MED-DENG for dengue fever or MED-MAL for malaria. See srv_health_procedures.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`claims`	Number of claims (hospital admissions) in the week.
`billed_amount`	Total hospital bill, rupees. Patients still in hospital have no bill yet.

Hospital claims and costs by monthsrv_health_hospital_monthly

One row per month and hospital. Source: synthetic insurer data. Claims, bills and length of stay for each network hospital by month of admission, October 2024 to September 2026, with a peer benchmark for the same procedures. Use it to find hospitals that bill far above similar hospitals.

`month`	First day of the month of admission, a DATE. 2024-10-01 to 2026-09-01.
`rohini_id`	The hospital's 13-digit registry id (ROHINI style, synthetic).
`hospital_name`	Hospital name (fictional).
`hospital_type`	multispeciality, single_speciality, nursing_home or day_care.
`tier`	City tier of the hospital's district. 1 for the largest metros, 2 for other cities, 3 for the rest.
`district_code`	District code of the hospital.
`district`	District name.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`claims`	Number of claims (admissions) in the month.
`claims_discharged`	Claims whose patient has been discharged, so the bill is known.
`billed_amount`	Total hospital bill for discharged patients, rupees.
`cost_per_admission`	Average bill per discharged patient, rupees.
`peer_billed_amount`	What peers would have billed for the same discharged patients, rupees. Peers are hospitals of the same type in the same city tier, and the benchmark is their median bill for the same procedure in the same quarter.
`cost_vs_peers`	billed_amount divided by peer_billed_amount. 1.0 means in line with peers, 2.0 means twice what peers bill for the same procedures. Over several months use SUM(billed_amount) / SUM(peer_billed_amount), not an average of this column.
`avg_los_days`	Average length of stay in days, for discharged patients. To combine months, weight each by claims_discharged.

Procedures billed by each hospitalsrv_health_hospital_procedures

One row per hospital and procedure. Source: synthetic insurer data. Which procedures each hospital billed, October 2024 to September 2026, against the capability the procedure needs and what the hospital declares in the registry. Use it to find hospitals billing procedures they are not equipped for.

`rohini_id`	The hospital's 13-digit registry id (ROHINI style, synthetic).
`hospital_name`	Hospital name (fictional).
`hospital_type`	multispeciality, single_speciality, nursing_home or day_care.
`district_code`	District code of the hospital.
`district`	District name.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`procedure_code`	Procedure package code, for example CARD-CABG for bypass surgery or CARD-PTCA for angioplasty.
`procedure_group`	cardiac, orthopaedic, eye, maternity, general_surgery, infection, respiratory, renal, cancer or newborn.
`required_capability`	The facility the procedure needs, for example cardiac_unit or cath_lab. none means any hospital can treat it.
`has_required_capability`	TRUE when the hospital declares the required capability in the registry (or none is needed). FALSE is a mismatch worth a closer look.
`first_month`	First month the hospital billed this procedure, a DATE.
`last_month`	Last month the hospital billed this procedure, a DATE.
`claims`	Number of claims for this procedure at this hospital.
`insurers`	Number of different insurers that received these claims.
`billed_amount`	Total hospital bill for these claims, rupees.

Network hospitalssrv_health_hospitals

One row per hospital. Source: synthetic insurer data. The registry of 900 network hospitals (ROHINI style, synthetic, fictional names) with type, size, accreditation and declared capabilities.

`rohini_id`	The hospital's 13-digit registry id (synthetic).
`hospital_name`	Hospital name (fictional).
`hospital_type`	multispeciality, single_speciality, nursing_home or day_care.
`speciality`	For single_speciality hospitals: cardiac, orthopaedic, eye, maternity, cancer, renal or paediatric. Empty text ('') for others.
`beds`	Number of beds.
`nabh`	TRUE when the hospital is NABH accredited.
`district_code`	District code of the hospital.
`district`	District name.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`tier`	City tier of the district. 1 for the largest metros, 2 for other cities, 3 for the rest.
`lat`	Latitude in degrees, within about 10 km of the district centre.
`lon`	Longitude in degrees.
`capabilities`	Declared facilities, separated by \|. From cardiac_unit, cath_lab, icu, dialysis, oncology, nicu, ortho_ot, general_ot, eye_ot, maternity. Test one with list_contains(string_split(capabilities, '\|'), 'cath_lab').

Procedure packagessrv_health_procedures

One row per procedure. Source: synthetic insurer data. The 25 procedure packages with base cost, typical stay, the capability each needs and how many claims each had, October 2024 to September 2026.

`procedure_code`	Procedure package code, for example OPH-CAT for cataract surgery.
`procedure_name`	Procedure name in plain words.
`procedure_group`	cardiac, orthopaedic, eye, maternity, general_surgery, infection, respiratory, renal, cancer or newborn.
`base_cost`	Typical package cost at an average hospital in a tier 2 city, rupees. Bills run higher in tier 1 cities and larger hospitals.
`typical_los_days`	Typical length of stay in days. 0 means day care.
`required_capability`	The facility the procedure needs, for example cardiac_unit. none means any hospital can treat it.
`seasonal`	TRUE for vector-borne and monsoon illnesses (dengue, malaria, typhoid, gastroenteritis, viral fever).
`season`	vector_borne (rises two to four weeks after rain), monsoon, winter (rises with cold and poor air) or none.
`claims`	Number of claims for the procedure, October 2024 to September 2026.
`avg_billed_amount`	Average bill per discharged patient, rupees.
`avg_los_days`	Average length of stay in days, for discharged patients.

Health products and their termssrv_health_products

One row per product. Source: synthetic insurer data. The 30 health products (fictional names) with their key policy terms, policies sold and how their claims ended, October 2024 to September 2026. Shares use settled claims only, so open claims are left out. top_up products have few claims, so their shares are unstable. Compare products with at least 300 settled claims.

`product_id`	Product id, HP001 to HP030.
`product_name`	Product name (fictional).
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`product_type`	individual, family_floater, senior_citizen or top_up.
`sum_insured_options`	Sums insured on offer, rupees, separated by \|.
`min_sum_insured`	Smallest sum insured on offer, rupees.
`max_sum_insured`	Largest sum insured on offer, rupees.
`deductible`	For top_up products, the part of a bill the policyholder or a base policy pays first, rupees. The top_up product pays only above it. 0 for other products.
`ped_wait_months`	Waiting period for pre-existing diseases (PED), months. 12, 24, 36 or 48. A claim for a pre-existing disease inside this period is rejected.
`specific_disease_wait_months`	Waiting period for named illnesses such as cataract or hernia, months.
`initial_wait_days`	Days from the start of a new policy before illness claims are covered. Accidents are covered from day one.
`room_rent_limit`	Limit on the hospital room: none, single private room, 2% of sum insured per day or 1% of sum insured per day. A tighter limit cuts the whole bill in proportion.
`room_rent_cap_pct`	The room rent cap as a percentage of the sum insured per day, 1 or 2. Empty when the limit is not a percentage.
`copay_pct`	Share of each claim the policyholder pays, percent. 0 means no co-pay.
`maternity_cover`	TRUE when childbirth is covered.
`maternity_wait_months`	Waiting period before childbirth is covered, months. Empty when not covered.
`restoration_benefit`	Refill of the sum insured once used up: none, once a year or unlimited.
`exclusions`	Named exclusions, separated by \|.
`policies`	Number of policies of this product in the synthetic book. The claim volumes are synthetic too, so do not divide claims by policies for a claim frequency.
`claims`	Number of claims on this product, October 2024 to September 2026.
`claims_settled`	Claims settled (paid, partly paid or rejected).
`claims_settled_full`	Claims with status paid, settled in full.
`claims_settled_part`	Claims with status partly_paid, settled with a deduction.
`claims_rejected`	Claims with status rejected.
`partly_paid_share`	claims_settled_part divided by claims_settled, between 0 and 1.
`rejected_share`	claims_rejected divided by claims_settled, between 0 and 1.
`cut_or_rejected_share`	Partly paid plus rejected claims divided by claims_settled, between 0 and 1.

Weather and illness by week (mock weather)srv_health_weather_weekly

One row per week and district. Source: mock and synthetic insurer data. Weekly rainfall, top temperature and air quality for every district, beside hospital admissions for dengue, malaria and breathing illness. The weather is MOCK data standing in for IMD and CPCB. Complete weeks from 2024-10-07 to the week starting 2026-09-28. Dengue tends to rise two to three weeks after heavy rain.

`week_start`	Monday of the week, a DATE. 2024-10-07 to 2026-09-28.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`district_code`	District code, for example WB-KOL for Kolkata.
`district`	District name.
`rainfall_mm`	Total rainfall in the week, millimetres (mock).
`temp_max_c`	Average daily top temperature in the week, degrees Celsius (mock).
`aqi`	Average air quality index in the week (mock). Above 200 is poor, above 300 very poor.
`dengue_claims`	Hospital admissions for dengue (MED-DENG) at hospitals in the district that week, by admission date.
`malaria_claims`	Hospital admissions for malaria (MED-MAL) in the district that week.
`respiratory_claims`	Hospital admissions for pneumonia, asthma or COPD in the district that week.
`all_claims`	All health claims (admissions) in the district that week.

Life policies by monthsrv_life_policies_monthly

One row per district, channel, product type, insurer and month. Source: synthetic insurer data. Life policies in force at each month end, new policies started in the month with their sum assured and premium, and lapses and surrenders in the month. October 2024 to September 2026. Only combinations with at least one policy have a row. policies_in_force and sum_assured_in_force are month-end counts, so take one month and never add them up across months. The other counts are for the month and can be added up.

`month`	First day of the month, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code, for example BR for Bihar.
`state_name`	State or UT name, for example Bihar.
`district_code`	District code, state code and three letters, for example BR-GAY for Gaya.
`district_name`	District name, for example Gaya.
`channel`	How the policy was sold. agency, bancassurance (through a bank), pos (Point of Sales person), online, broker or direct.
`product_type`	term, endowment, ulip (unit linked), whole_life or money_back.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role. Only some of the 25 insurers write life business, so an insurer with no rows wrote none.
`new_policies`	Policies that started in the month.
`policies_in_force`	Policies in force at the month end. A policy leaves the book when it lapses, is surrendered or the holder dies. A month-end count, so do not add it up across months.
`sum_assured`	Sum assured of the new policies that started in the month, rupees.
`sum_assured_in_force`	Sum assured of the policies in force at the month end, rupees. A month-end amount, so do not add it up across months.
`annual_premium`	Annual premium of the new policies that started in the month, rupees.
`lapses`	Policies that lapsed in the month (premium not paid).
`surrenders`	Policies surrendered in the month (the holder took the surrender value).

Life death claims by monthsrv_life_death_claims_monthly

One row per district, channel, product type, insurer, duration band, cause group and month. Source: synthetic insurer data. Death claims by the month the insurer was told of the death (intimation date). October 2024 to September 2026. An early claim is a death within 365 days of the policy start. Only combinations with at least one claim have a row. Every count can be added up across rows.

`month`	First day of the month the death was intimated, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`district_code`	District code of the policy, for example BR-NAL for Nalanda.
`district_name`	District name.
`channel`	agency, bancassurance, pos, online, broker or direct.
`product_type`	term, endowment, ulip, whole_life or money_back.
`insurer_code`	Insurer code. Shown as a name or an alias, by role.
`duration_band`	Time from the policy start to the death. under 1 year, 1 to 3 years or over 3 years (3 years or more).
`cause_group`	Cause of death as reported. illness, accident, natural, suicide or unknown.
`claims`	Death claims intimated in the month.
`early_claims`	Claims where the death came within 365 days of the policy start. Equal to claims when duration_band is under 1 year, else 0.
`amount`	Amount claimed, rupees. Sum assured plus any bonus, or the fund value for a ULIP if higher. Counted whatever the decision.
`paid`	Claims paid, as known at 2026-10-04.
`repudiated`	Claims rejected by the insurer, as known at 2026-10-04. The repudiation rate is repudiated divided by paid plus repudiated.
`under_investigation`	Claims not yet decided at 2026-10-04.
`late_intimations`	Claims intimated more than 30 days after the death.
`non_medical_claims`	Claims on policies issued without a medical examination (non-medical underwriting).

Life death claims by weeksrv_life_death_claims_weekly

One row per district, channel, insurer and week. Source: synthetic insurer data. Death claims by the week the insurer was told of the death (intimation date). Complete weeks only, Monday to Sunday, from the week starting 2024-10-07 to the week ending 2026-10-04. Only weeks with at least one claim have a row. Use it for weekly questions and the early death claims alert.

`week_start`	Monday of the week the death was intimated, a DATE. 2024-10-07 to 2026-09-28.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`district_code`	District code of the policy.
`district_name`	District name.
`channel`	agency, bancassurance, pos, online, broker or direct.
`insurer_code`	Insurer code. Shown as a name or an alias, by role.
`claims`	Death claims intimated in the week.
`early_claims`	Claims where the death came within 365 days of the policy start.
`amount`	Amount claimed, rupees.
`late_intimations`	Claims intimated more than 30 days after the death.
`non_medical_claims`	Claims on policies issued without a medical examination (non-medical underwriting).

Life agents by monthsrv_life_agent_monthly

One row per agent and month. Source: synthetic insurer data. What each agent sold in the month and what happened to their policies. October 2024 to September 2026. Agents are identified by registry number only. Only months with some activity have a row. Online and direct policies have no agent.

`month`	First day of the month, a DATE. 2024-10-01 to 2026-09-01.
`agent_id`	Agent registry number. ENV- for the Envoy style registry (agency, bancassurance and broker), POS- for Point of Sales persons.
`registry`	envoy or pos.
`channel`	agency, bancassurance, pos or broker.
`insurer_code`	Insurer the agent sells for. Shown as a name or an alias, by role.
`state_code`	Two-letter state or UT code of the agent's district.
`state_name`	State or UT name.
`district_code`	District where the agent is registered.
`district_name`	District name.
`policies_sold`	Policies this agent sold that started in the month.
`sum_assured_sold`	Sum assured of those policies, rupees.
`early_lapses`	This agent's policies that lapsed in the month within 365 days of their start.
`death_claims`	Death claims intimated in the month on this agent's policies.
`early_death_claims`	Of those, claims where the death came within 365 days of the policy start.

Life agent registrysrv_life_agents

One row per agent. Source: synthetic insurer data. The agent registry with each agent's sales and claims from October 2024 to September 2026. Agents are identified by registry number only. No agent names.

`agent_id`	Agent registry number, ENV- or POS- and six digits.
`registry`	envoy (agency, bancassurance and broker staff) or pos (Point of Sales person).
`channel`	agency, bancassurance, pos or broker.
`insurer_code`	Insurer the agent sells for. Shown as a name or an alias, by role.
`state_code`	Two-letter state or UT code of the agent's district.
`state_name`	State or UT name.
`district_code`	District where the agent is registered.
`district_name`	District name.
`licence_date`	Date the agent was licensed, a DATE. PoS licences start in April 2016.
`status`	Registry status at 2026-10-04. active, suspended, lapsed (licence not renewed) or terminated.
`first_sale_date`	Start date of the agent's earliest policy in the data (the book in force at October 2024 plus later sales), a DATE. Empty if there is none.
`last_sale_date`	Start date of the agent's latest policy, a DATE. Empty if there is none.
`policies_sold`	Policies the agent sold that started from 2024-10-01 to 2026-09-30.
`early_lapses`	The agent's policies that lapsed within 365 days of their start, lapsing from 2024-10-01 to 2026-09-30.
`death_claims`	Death claims on the agent's policies intimated from 2024-10-01 to 2026-09-30.
`early_death_claims`	Of those, claims where the death came within 365 days of the policy start.

Life productssrv_life_products

One row per product. Source: synthetic insurer data. The 12 fictional life products and their terms. Each product belongs to one insurer.

`product_id`	Product code, LP01 to LP12.
`insurer_code`	Insurer that sells the product. Shown as a name or an alias, by role.
`product_name`	Fictional product name.
`product_type`	term, endowment, ulip, whole_life or money_back.
`payment_term`	How long premiums are paid. regular (for the whole policy term) or limited to a number of years, for example limited 10 years.
`policy_term_min_years`	Shortest policy term offered, years.
`policy_term_max_years`	Longest policy term offered, years. Whole life cover runs to age 100.
`min_entry_age`	Youngest age at entry, years.
`max_entry_age`	Oldest age at entry, years.
`min_sum_assured`	Smallest sum assured, rupees.
`max_sum_assured`	Largest sum assured, rupees.
`launch_date`	Date the product was launched, a DATE.

Life new business cohortssrv_life_cohorts

One row per start month, channel, product type, insurer and age band. Source: synthetic insurer data. Policies that started from October 2024 to September 2026, grouped by start month, and how many lapsed or were surrendered early. Use it for persistency and for the profile of ULIPs sold through banks. Exits are counted up to 2026-10-04, so recent months have had less time to lapse.

`start_month`	First day of the month the policies started, a DATE. 2024-10-01 to 2026-09-01.
`channel`	agency, bancassurance, pos, online, broker or direct.
`product_type`	term, endowment, ulip, whole_life or money_back.
`insurer_code`	Insurer code. Shown as a name or an alias, by role.
`age_band`	Age of the holder at entry, years. 18 to 34, 35 to 44, 45 to 54, 55 to 64 or 65 and over. Buyers aged 55 or over are the last two bands.
`policies`	Policies that started in the month.
`sum_assured`	Sum assured of those policies, rupees.
`annual_premium`	Annual premium of those policies, rupees.
`lapsed_within_1_year`	Of those, policies that lapsed within 365 days of their start.
`exits_within_2_years`	Of those, policies that lapsed or were surrendered within 730 days of their start, so far.
`full_year_observed`	TRUE when the start month is 2025-09-01 or earlier, so every policy has had a full first year by 2026-10-04. Use only these rows for lapse within 1 year rates.

Property exposure by weeksrv_property_exposure_weekly

One row per week, state, district, property type and insurer. Source: synthetic insurer data. Fire, property and marine cargo policies in force at the end of each week, with their total sum insured, and the new business that started in the week. Complete weeks only, Monday to Sunday, from the week starting 2024-10-07 to the week ending 2026-10-04. A policy counts as in force when the Sunday that ends the week falls between its start and end dates. Only combinations with at least one policy in force appear. policies_in_force and sum_insured are counts on one day, so never add them up across weeks. Filter to one week_start, or take an average over weeks. new_policies and new_sum_insured can be added across weeks. Use it for exposure, accumulation and build-up questions, for example the latest week against the average of earlier weeks.

`week_start`	Monday of the week, a DATE. 2024-10-07 to 2026-09-28. The latest week is 2026-09-28 to 2026-10-04.
`state_code`	Two-letter state or UT code, for example MH for Maharashtra or AS for Assam.
`state_name`	State or UT name, for example Maharashtra.
`district_code`	District code, the state code then a short district code, for example MH-THN for Thane or AS-BAR for Barpeta.
`district`	District name, for example Thane.
`property_type`	What is insured: house, retail, office, warehouse, factory, hotel or hospital.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`policies_in_force`	Policies in force on the Sunday that ends the week. A count on that day. Do not add it up across weeks.
`sum_insured`	Total sum insured of the policies in force on the Sunday that ends the week, rupees. A total on that day. Do not add it up across weeks. 1 crore is 100 lakh, which is 10,000,000 rupees.
`new_policies`	New business policies (not renewals) that started in the week. They are also counted in policies_in_force.
`new_sum_insured`	Sum insured of the new business policies that started in the week, rupees.

Property claims by weeksrv_property_claims_weekly

One row per week, state, district, cause, line, property type and insurer. Source: synthetic insurer data. Fire, property and marine cargo claims by the week they were reported. Complete weeks only, Monday to Sunday, from the week starting 2024-10-07 to the week ending 2026-10-04. Floods and storms peak in the monsoon. Cyclones hit the east coast in October to December and in May. Claims reported late for a loss still show in the week they were reported. Claims can be added across weeks. For a loss ratio, divide claim_amount here by premium in srv_property_premium_monthly over the same months.

`week_start`	Monday of the week the claims were reported, a DATE. 2024-10-07 to 2026-09-28.
`state_code`	Two-letter state or UT code where the insured property is.
`state_name`	State or UT name.
`district_code`	District code where the insured property is, for example AS-KMR for Kamrup.
`district`	District name.
`cause`	Cause of loss: fire, flood, cyclone, storm, burglary, earthquake or other.
`line`	Line of business of the policy that claimed: fire, property (householder and shopkeeper packages and similar) or marine_cargo.
`property_type`	What is insured: house, retail, office, warehouse, factory, hotel or hospital.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`claims`	Number of claims reported in the week.
`claim_amount`	Amount claimed, rupees.
`paid_amount`	Amount paid so far, rupees. Recent claims are mostly still open, so paid is often zero.

Property exposure by districtsrv_property_exposure_district

One row per district and property type. Source: synthetic insurer data. Fire, property and marine cargo policies in force on 2026-10-04, the end of the latest complete week, with their total sum insured, by district and property type, all insurers together. It matches the latest week of srv_property_exposure_weekly. Latitude and longitude are the approximate district centre, for maps.

`as_at`	The date the policies are counted in force, a DATE. Always 2026-10-04.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`district_code`	District code, for example MH-THN for Thane.
`district`	District name.
`lat`	Latitude of the approximate district centre, decimal degrees.
`lon`	Longitude of the approximate district centre, decimal degrees.
`property_type`	What is insured: house, retail, office, warehouse, factory, hotel or hospital.
`policies_in_force`	Policies in force on 2026-10-04.
`sum_insured`	Total sum insured of those policies, rupees.

Property premium by monthsrv_property_premium_monthly

One row per state, line, property type, insurer and month. Source: synthetic insurer data. Fire, property and marine cargo policies that started in each month, new business and renewals, with premium and sum insured written. October 2024 to September 2026. Each policy is an annual term, so a location that renews shows again a year later. All columns can be added across months. For a loss ratio, divide claim_amount in srv_property_claims_weekly by premium here over the same months.

`month`	First day of the month the policies started, a DATE. 2024-10-01 to 2026-09-01.
`state_code`	Two-letter state or UT code.
`state_name`	State or UT name.
`line`	Line of business: fire, property (householder and shopkeeper packages and similar) or marine_cargo.
`property_type`	What is insured: house, retail, office, warehouse, factory, hotel or hospital.
`insurer_code`	Insurer code, INS01 to INS25. Shown as a name or an alias, by role.
`policies_written`	Policies that started in the month, new business and renewals.
`new_policies`	New business policies (not renewals) that started in the month.
`premium`	Premium written for the policies that started in the month, rupees, before taxes.
`sum_insured_written`	Total sum insured of the policies that started in the month, rupees.

Property exposure on a map gridsrv_property_exposure_grid

One row per map cell of 0.1 degree and property type. Source: synthetic insurer data. Fire, property and marine cargo policies in force on 2026-10-04, counted in map cells of 0.1 degree of latitude and longitude, about 11 km across, by property type. Use it to see where insured value sits, for example near a flood or a cyclone track.

`as_at`	The date the policies are counted in force, a DATE. Always 2026-10-04.
`cell_lat`	Latitude of the centre of the map cell, decimal degrees.
`cell_lon`	Longitude of the centre of the map cell, decimal degrees.
`property_type`	What is insured: house, retail, office, warehouse, factory, hotel or hospital.
`policies_in_force`	Policies in force on 2026-10-04 whose location falls in the cell.
`sum_insured`	Total sum insured of those policies, rupees.

Data quality checks

Check	Runs on	Field	Type	What it checks	Failed
`DQ-P01`	Each policy row	`policy_no`	must pass	Policy number is missing.	0
`DQ-P02`	Each policy row	`insurer_code`	must pass	Insurer code is missing or not a registered insurer.	0
`DQ-P03`	Each policy row	`vehicle_class`	must pass	Vehicle class is missing or not a recognised class: two-wheeler, private car, goods vehicle, passenger vehicle or other.	522
`DQ-P04`	Each policy row	`cover_type`	must pass	Cover type is missing or not comprehensive, third-party only or standalone own damage.	0
`DQ-P05`	Each policy row	`total_premium`	must pass	Premium is outside the plausible range for this vehicle class. Check for a paise-for-rupees unit error.	1,124
`DQ-P06`	Each policy row	`end_date`	must pass	Policy end date is missing or not after the start date.	0
`DQ-P07`	Each policy row	`reg_no`	warning	Registration number is missing or not in the standard format of state code, RTO number, series and number, as in TS 09 AB 1234.	0
`DQ-P08`	Each policy row	`policy_no`	must pass	Duplicate policy number. The first copy is kept and later copies are held back.	118
`DQ-P09`	Each policy row	`state_code`	must pass	State code is missing or not a valid state or union territory.	400
`DQ-P10`	Each policy row	`rto_code`	must pass	RTO code is missing, unknown or does not belong to the state on the row.	0
`DQ-P11`	Each policy row	`total_premium`	warning	Total premium does not equal OD premium plus TP premium.	0
`DQ-P12`	Each policy row	`od_premium`	warning	A third-party-only policy carries own-damage premium.	0
`DQ-P13`	Each policy row	`issue_date`	must pass	Issue date is missing or not a valid date.	0
`DQ-P14`	Each policy row	`vehicle_age`	warning	Vehicle age is missing or outside 0 to 25 years.	0
`DQ-P15`	Each policy row	`holder_mobile`	warning	Policyholder mobile number is missing or not a valid 10-digit Indian mobile number.	0
`DQ-P16`	Each policy row	`holder_email`	warning	Policyholder email is missing or not a valid email address.	1,169
`DQ-C01`	Each claim row	`claim_no`	must pass	Claim number is missing.	0
`DQ-C02`	Each claim row	`policy_no`	must pass	Claim has no policy number.	0
`DQ-C03`	Each claim row	`claim_type`	must pass	Claim type is missing or not own damage, third party or theft.	0
`DQ-C04`	Each claim row	`loss_date`	must pass	Loss date is missing or falls outside the policy period.	0
`DQ-C05`	Each claim row	`claim_amount`	must pass	Own-damage or theft claim amount is more than the vehicle's insured declared value (IDV).	0
`DQ-C06`	Each claim row	`report_date`	must pass	Report date is missing or earlier than the loss date.	0
`DQ-C07`	Each claim row	`claim_no`	must pass	Duplicate claim number. The first copy is kept and later copies are held back.	0
`DQ-C08`	Each claim row	`rto_code`	must pass	RTO code is missing, unknown or does not belong to the state on the row.	0
`DQ-C09`	Each claim row	`policy_no`	warning	Claim refers to a policy number that is not in the data store.	9
`DQ-C10`	Each claim row	`claim_amount`	must pass	Claim amount is missing, negative or above 5 crore rupees.	0
`DQ-F01`	Each file	`received_date`	warning	File arrived after the due date.	73
`DQ-F02`	Each file	`received_date`	must pass	File arrived more than 3 days after the due date.	6
`DQ-F03`	Each file	`row_count`	must pass	File arrived with no data rows.	0
`DQ-F04`	Each file	`row_count`	warning	Rows received do not match the count the insurer said it sent.	0
`DQ-F05`	Each file	`sha256`	must pass	File contents do not match the fingerprint the insurer registered.	0

Row checks run on every row received and count rows. File checks run once per file and count files. A row that fails a must-pass check is held back, so it never reaches the ready-to-use tables. A file that fails a must-pass file check is flagged on Insurer file checks, but its rows are not held back.

Rows held back, by month and check

A row can break more than one check. In September 2026, 1,463 rows broke checks 1,543 times.

Month	Check	Rows
September 2026	`DQ-P03` Vehicle class is missing or not a recognised class: two-wheeler, private car, goods vehicle, passenger vehicle or other.	409
September 2026	`DQ-P05` Premium is outside the plausible range for this vehicle class.	1,074
September 2026	`DQ-P08` Duplicate policy number.	60
August 2026	`DQ-P03` Vehicle class is missing or not a recognised class: two-wheeler, private car, goods vehicle, passenger vehicle or other.	113
July 2026	`DQ-P05` Premium is outside the plausible range for this vehicle class.	17
June 2026	`DQ-P05` Premium is outside the plausible range for this vehicle class.	33
November 2025	`DQ-P08` Duplicate policy number.	38
November 2025	`DQ-P09` State code is missing or not a valid state or union territory.	400
October 2025	`DQ-P08` Duplicate policy number.	20

Glossary

Term	Definition	Formula	Owner	Key data item
Policy in force	A motor policy whose cover period includes the date in question. The Datahub counts policies in force at each month end.	`Count of policies where start date <= month end <= end date`	Motor Vertical	Key
Claim frequency	Claims reported per 1,000 policies in force whose cover includes that claim type. Own damage and theft use COMP and SAOD policies. Third party uses COMP and TP policies.	`Claims reported in the period x 1,000 / eligible policies in force. For a quarter, divide by the average of its month-end counts.`	Actuarial	Key
Claim severity	The average amount claimed per claim.	`Claim amount / number of claims`	Actuarial	Key
Own damage (OD)	Cover for loss of or damage to the insured vehicle itself, for example from an accident, fire or flood. In this demo, theft is recorded as its own claim type, THEFT, and is not counted in OD.	`claim_type = 'OD'`	Motor Vertical
Third party (TP)	Compulsory liability cover for death, injury or property damage caused to other people. A TP-only policy has no own-damage cover.	`claim_type = 'TP' for claims. cover_type = 'TP' for third-party-only policies.`	Motor Vertical
Standalone OD (SAOD)	Own-damage cover bought on its own, when the vehicle's third-party cover is held under a separate policy.	`cover_type = 'SAOD'`	Motor Vertical
Comprehensive cover (COMP)	A package policy with both own-damage and third-party cover.	`cover_type = 'COMP'`	Motor Vertical
IDV	Insured declared value. The sum insured for the vehicle, close to its market value after depreciation. It is the most an own-damage or theft claim can pay.	`Manufacturer's listed price x (1 - depreciation for vehicle age)`	Motor Vertical	Key
Loss ratio	Claims as a share of premium. The serving views hold written premium, not earned premium, so a loss ratio built from them is an approximation.	`Claim amount / premium`	Actuarial	Key
Submission cycle	One calendar month of insurer data. Policy and claim files for cycle M are due on the 5th of month M+1.	`cycle = YYYY-MM of the issue date (policies) or report date (claims)`	Data Operations	Key
Insurer code	A stable code for each insurer, INS01 to INS25. Depending on role, the screen shows the insurer's name or an alias such as Insurer K.	`INS + two digits`	Data Operations	Key
Vehicle class	The five classes used across the demo. 2W is two-wheeler, PCAR is private car, GCV is goods carrying vehicle, PCV is passenger carrying vehicle and MISC is everything else. VAHAN categories are mapped to these classes.	`One of 2W, PCAR, GCV, PCV, MISC`	Data Governance	Key
RTO	Regional Transport Office, the local office that registers vehicles. Each has a code made of the state code and a number, for example TS09 in Hyderabad. Policies and claims are counted by the RTO where the vehicle is registered.	`State code + two digits, for example TS09`	Motor Vertical	Key
Registered stock	Vehicles on the register, estimated as VAHAN registrations summed over the last N years. N is 15, an assumption. Older vehicles are assumed to be off the road.	`Sum of VAHAN registrations for the last N calendar years (N = 15, assumed)`	Data Governance	Key
Uninsured estimate	Registered stock minus policies in force scaled to the market. It is an estimate of the insurance gap, not a list of named vehicles.	`Registered stock - (synthetic policies in force x sampling weight)`	Motor Vertical	Key
Insured share	The share of registered stock with a policy in force.	`Policies in force (scaled to the market) / registered stock`	Motor Vertical	Key
New policies	Policies issued in the period, both new business and renewals.	`Count of policies with issue date in the period`	Motor Vertical
Premium written	Total premium, own damage plus third party, on policies issued in the period. Rupees.	`Sum of total premium for policies issued in the period`	Actuarial	Key
Theft claim rate	Theft claims reported per 1,000 policies in force with own-damage cover (COMP or SAOD).	`THEFT claims x 1,000 / COMP and SAOD policies in force`	Actuarial
Sampling weight	The synthetic book is a sample of the market. Each synthetic policy in force stands for this many insured vehicles. It is an assumption of the demo.	`Assumed insured vehicles in the market / synthetic policies in force`	Data Governance
Lapse	A policy that is not renewed within 30 days of its expiry date. The vehicle may then be uninsured.	`No renewal policy issued from 30 days before to 30 days after expiry`	Motor Vertical
Cashless claim	A health claim where the insurer pays the hospital directly, so the patient does not pay the bill first. The other kind is a reimbursement claim.	`Cashless share = cashless claims / all claims`	Health Vertical	Key
Reimbursement claim	A health claim where the patient pays the hospital and the insurer pays the patient back afterwards.	`Reimbursement claims = claims - cashless claims`	Health Vertical
ROHINI id	The 13-digit registry number of a network hospital. ROHINI is the national registry of hospitals in the insurance network. In this demo the ids and hospital names are synthetic.	`13 digits, one per hospital`	Health Vertical	Key
Length of stay	Days from admission to discharge, for patients who have left hospital. Day care counts as 0 days, or 1 with an overnight stay. Averages over several rows are weighted by discharged claims.	`SUM(avg_los_days x claims_discharged) / SUM(claims_discharged)`	Health Vertical	Key
Cost per admission	The average hospital bill for each patient who has been discharged. Patients still in hospital have no bill yet, so they are left out.	`Billed amount / discharged claims`	Actuarial	Key
Cost against peers	What a hospital billed divided by what similar hospitals bill for the same procedures. Peers are hospitals of the same type in the same city tier. 1.0 is in line, 2.0 is twice what peers bill.	`SUM(billed_amount) / SUM(peer_billed_amount)`	Health Vertical	Key
Procedure package	A treatment with an agreed package price, such as cataract surgery or angioplasty. Each has a code, for example OPH-CAT, and belongs to a procedure group such as eye or cardiac.	`One of 25 procedure codes`	Health Vertical
Required capability	The facility a procedure needs, for example a cath lab for angioplasty. A hospital billing a procedure without declaring the facility in the registry is worth a closer look.	`has_required_capability = FALSE marks a mismatch`	Health Vertical
Pre-existing disease waiting period	Months a new policy must run before illnesses the person already had are covered. A claim for a pre-existing disease (PED) inside this period is rejected. 12 to 48 months in this demo.	`ped_wait_months on the product`	Health Vertical	Key
Room rent limit	A cap on the hospital room a policy pays for, such as 1% of the sum insured a day. A patient in a dearer room has the whole bill cut in proportion.	`room_rent_limit on the product, with room_rent_cap_pct when it is a percentage`	Health Vertical
Co-pay	The share of each claim the policyholder pays, as a percentage. 0 means no co-pay.	`copay_pct on the product`	Health Vertical
Health sum insured	The most a health policy pays in a policy year. Each product offers a few sums insured to choose from.	`One of sum_insured_options on the product`	Health Vertical	Key
Claim rejection rate	Rejected claims as a share of settled claims. Claims still open are left out, because they have no decision yet.	`claims_rejected / (claims - claims_open)`	Health Vertical	Key
Cut or rejected share	Claims paid with a deduction or rejected, as a share of settled claims. A high share can point to policy terms that surprise customers.	`(claims_settled_part + claims_rejected) / claims_settled`	Health Vertical
City tier	A rough size band for the hospital's city. 1 for the largest metros, 2 for other cities, 3 for the rest. Bills run higher in tier 1.	`1, 2 or 3`	Health Vertical
NABH accreditation	Accreditation of a hospital by the National Accreditation Board for Hospitals. In this demo the flag is synthetic.	`nabh = TRUE`	Health Vertical
Vector-borne illness	Illness spread by mosquitoes, such as dengue and malaria. Admissions tend to rise two to three weeks after heavy rain. The weather beside them is mock data.	`season = 'vector_borne' (MED-DENG, MED-MAL)`	Health Vertical
Sum assured	The amount a life policy pays on death. For a ULIP the claim is the fund value if that is higher.	`Sum assured of the policy, rupees`	Life Vertical	Key
Life policy in force	A life policy that has not lapsed, been surrendered or paid out on death at the month end. A month-end count, so it is never added up across months.	`Policies in force at the month end`	Life Vertical	Key
Death claim	A claim made when the life assured dies. It is dated by the day the insurer was told of the death, the intimation date.	`Count of claims by intimation date`	Life Vertical	Key
Early death claim	A death within 365 days of the policy start. A cluster of early claims sold by a few agents can point to policies bought on people already seriously ill.	`Claims with duration_band = 'under 1 year'`	Life Vertical	Key
Intimation date	The day the insurer was told of the death. A late intimation is one more than 30 days after the death.	`late_intimations counts claims intimated more than 30 days after the death`	Life Vertical
Repudiation rate	Death claims the insurer rejected, as a share of claims decided. Claims still under investigation are left out.	`repudiated / (paid + repudiated)`	Life Vertical	Key
Non-medical underwriting	A policy issued without a medical examination. Insurers allow it below a sum assured limit.	`non_medical_claims counts claims on such policies`	Life Vertical
PoS agent	A Point of Sales person, licensed to sell simple products. Their registry numbers start POS-. Other agents are in the Envoy style registry, ENV-.	`registry = 'pos' or channel = 'pos'`	Life Vertical
Bancassurance	Insurance sold through a bank's branches and staff. One of the sales channels, beside agency, PoS, online, broker and direct.	`channel = 'bancassurance'`	Life Vertical
Sales channel	How a life policy was sold: agency, bancassurance, pos, online, broker or direct.	`One of six channel values`	Life Vertical
ULIP	A unit linked insurance plan. Part of the premium is invested in funds, so the value goes up and down. Selling one to a buyer who wanted a deposit is a common mis-selling complaint.	`product_type = 'ulip'`	Life Vertical
Lapse (life)	A life policy that ends because a premium was not paid. A surrender is different: the holder ends the policy and takes the surrender value.	`lapses in srv_life_policies_monthly`	Life Vertical
First-year lapse rate	Policies that lapsed within 365 days of their start, as a share of policies started. Only start months with a full first year are used, up to September 2025.	`SUM(lapsed_within_1_year) / SUM(policies) where full_year_observed`	Actuarial	Key
Annual premium	The premium a new life policy pays each year, rupees.	`Sum of annual premium of policies started in the period`	Actuarial	Key
Sum insured (property)	The most a fire, property or marine cargo policy pays, usually the value of what is insured. Added up over policies in force, it is the exposure.	`Sum of sum insured of policies in force on a day`	Other Lines Vertical	Key
Exposure	The total sum insured of policies in force on one day, in one place. It is a count on a day, so it is never added up across weeks.	`SUM(sum_insured) for one week_start`	Other Lines Vertical	Key
Accumulation	Insured value building up in one place, so one flood, fire or cyclone could hit many policies at once. Watched as the latest week against earlier weeks.	`Latest week's sum insured / average of earlier weeks, for one district and property type`	Other Lines Vertical
Property type	What is insured: house, retail, office, warehouse, factory, hotel or hospital.	`One of seven property types`	Other Lines Vertical
Line of business	Fire, property (householder and shopkeeper packages and similar) or marine cargo, for goods in transit.	`line = 'fire', 'property' or 'marine_cargo'`	Other Lines Vertical
Cause of loss	What caused a fire or property claim: fire, flood, cyclone, storm, burglary, earthquake or other.	`cause on the claim`	Other Lines Vertical
Loss ratio (Other Lines)	Amount claimed as a share of premium written over the same period. It is a rough guide, since claims are by the week reported and premium is written, not earned.	`SUM(claim_amount) / SUM(premium) over the same months`	Actuarial	Key
New business (Other Lines)	Policies taken out for the first time, not renewals, with the sum insured they add.	`new_policies and new_sum_insured in the period`	Other Lines Vertical
Map cell	A square of 0.1 degree of latitude and longitude, about 11 km across, used to show where insured value sits.	`Round latitude and longitude to the nearest 0.1 degree centre`	Other Lines Vertical

Key marks a critical data element, one the glossary flags as most important to get right. Every term has one agreed definition and an owner. Owners are example teams for the demo, not named people.