Introduction
This file is a different kind of dataset from a typical single-table
CSV: it’s a sample extract from the U.S. General Services
Administration’s Natural Gas Acquisition Program (NGAP)
database, and it packs 11 separate relational tables
into one Excel worksheet, stacked vertically and separated by
blank rows and "Table Name: X" marker rows. Each table
holds only a handful of illustrative sample rows (4-5 typically) rather
than a full population — this looks like a schema/format sample rather
than production data, so the right analysis here is different from a
standard descriptive report: understanding what each table
represents, how they relate to one another, and what data-quality issues
are visible even in this small sample, rather than computing
distributions across a large population.
The NGAP program is how GSA procures and manages natural gas
contracts and delivery for federal facilities nationwide — the tables
below cover contracts, contractors, local distribution companies (LDCs),
facilities, points of contact, invoices, and monthly cost/usage
records.
raw <- read_excel("natural_gas.xlsx", sheet = "NGAP Sample Data", col_names = FALSE, col_types = "text")
# The sheet stacks 11 tables vertically. Each table starts with a single-cell
# "Table Name: X" row, is followed by a header row, then data rows, and ends
# at the next fully-blank row. This loop slices the sheet into a named list
# of tidy data frames, one per table.
col1 <- raw[[1]]
table_starts <- which(str_detect(col1, "^Table Name:"))
table_names <- str_trim(str_remove(col1[table_starts], "^Table Name:\\s*"))
is_blank_row <- function(row) all(is.na(row) | row == "")
extract_table <- function(start_row) {
header_row <- start_row + 1
header <- raw[header_row, ] %>% unlist() %>% unname()
n_cols <- sum(!is.na(header))
header <- header[1:n_cols]
# find the first fully-blank row after the header to mark the table's end
r <- header_row + 1
while (r <= nrow(raw) && !is_blank_row(raw[r, ])) r <- r + 1
end_row <- r - 1
raw[(header_row + 1):end_row, 1:n_cols] %>%
set_names(header) %>%
mutate(across(everything(), ~ na_if(str_trim(.), "NULL")))
}
ngap <- table_starts %>%
map(extract_table) %>%
set_names(table_names)
About the data
tibble(
Table = names(ngap),
Columns = map_int(ngap, ncol),
`Sample Rows` = map_int(ngap, nrow)
) %>%
kable(caption = "Tables parsed out of the single worksheet") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)
Tables parsed out of the single worksheet
|
Table
|
Columns
|
Sample Rows
|
|
Contract
|
5
|
5
|
|
ContractFacilityLDC
|
6
|
5
|
|
Contractor
|
15
|
5
|
|
Facility
|
15
|
5
|
|
FacilityLDC
|
4
|
5
|
|
LDC
|
8
|
5
|
|
Master_Facility
|
2
|
5
|
|
POC
|
18
|
5
|
|
POCEntitySubContact
|
2
|
5
|
|
tblCostUsageTest
|
26
|
5
|
|
tblInvoiceMaster
|
7
|
5
|
|
tblInvoiceStatus
|
2
|
5
|
Every table here has 4 to 15 sample rows — far too
few to support the kind of distribution or correlation analysis
appropriate for a real population sample. That’s expected: this file
exists to illustrate the shape of the NGAP database schema, not
to summarize the full program. The analysis below treats it
accordingly.
How the tables relate
tibble(
Table = c("Contract", "ContractFacilityLDC", "Contractor", "Facility", "FacilityLDC", "LDC",
"Master_Facility", "POC", "POCEntitySubContact", "tblCostUsageTest",
"tblInvoiceMaster", "tblInvoiceStatus"),
Purpose = c(
"One row per natural gas supply contract, linked to a contractor",
"Links a contract to a facility/LDC combination, with contract terms (cost, months, type)",
"The natural gas suppliers/marketers GSA contracts with",
"A federal facility (building) receiving natural gas service",
"Links a facility to its Local Distribution Company, with estimated annual volume",
"The Local Distribution Company (LDC) — the utility that physically delivers gas to a facility",
"A master list of facility/region codes",
"Point of Contact — the person at a facility responsible for energy/utility matters",
"Sub-classification of a point of contact's role (e.g. Primary, Secondary)",
"Monthly cost and usage records per facility, tracking both LDC and supplier-side cost/volume",
"Invoice files received per facility, by year and month",
"Status history for uploaded invoices"
)
) %>%
kable(caption = "Table purposes within the NGAP schema") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)
Table purposes within the NGAP schema
|
Table
|
Purpose
|
|
Contract
|
One row per natural gas supply contract, linked to a contractor
|
|
ContractFacilityLDC
|
Links a contract to a facility/LDC combination, with contract terms
(cost, months, type)
|
|
Contractor
|
The natural gas suppliers/marketers GSA contracts with
|
|
Facility
|
A federal facility (building) receiving natural gas service
|
|
FacilityLDC
|
Links a facility to its Local Distribution Company, with estimated
annual volume
|
|
LDC
|
The Local Distribution Company (LDC) — the utility that physically
delivers gas to a facility
|
|
Master_Facility
|
A master list of facility/region codes
|
|
POC
|
Point of Contact — the person at a facility responsible for
energy/utility matters
|
|
POCEntitySubContact
|
Sub-classification of a point of contact’s role (e.g. Primary,
Secondary)
|
|
tblCostUsageTest
|
Monthly cost and usage records per facility, tracking both LDC and
supplier-side cost/volume
|
|
tblInvoiceMaster
|
Invoice files received per facility, by year and month
|
|
tblInvoiceStatus
|
Status history for uploaded invoices
|
The natural key that threads through most of these tables is
FacilityCode
(e.g. "VA0003ZZ") — it links Facility,
FacilityLDC, tblCostUsageTest, and
tblInvoiceMaster together, so that a single federal
building’s gas contract, delivery utility, monthly costs, and invoices
can all be traced from one code. Contract and
Contractor are linked by ContractorId, and
LDC/Contractor/POC all carry a
StateAbbrev field that ties them loosely to geography.
Table-by-table contents
for (nm in names(ngap)) {
cat("### ", nm, "\n\n")
print(
ngap[[nm]] %>%
kable() %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = TRUE, font_size = 12)
)
cat("\n\n")
}
Contract
|
ContractNumber
|
ContractorId
|
ExpiryDate
|
CFileName
|
ReactivationDt
|
|
GS-00P-02-BSC-0201
|
23
|
2004-09-30 00:00:00
|
NA
|
2004-09-30 00:00:00
|
|
GS-00P-02-BSC-0204
|
5
|
2003-10-31 00:00:00
|
NA
|
NA
|
|
GS-00P-02-BSC-0206
|
6
|
2004-10-31 00:00:00
|
NA
|
2004-11-02 00:00:00
|
|
GS-00P-02-BSC-0207
|
4
|
2006-10-31 00:00:00
|
NA
|
2004-11-01 00:00:00
|
|
GS-00P-02-BSC-0209
|
7
|
2004-10-31 00:00:00
|
NA
|
2004-11-01 00:00:00
|
ContractFacilityLDC
|
ConFacCode
|
ContractCost
|
FlowDate
|
ContractMonths
|
ContractType
|
Checked
|
|
316
|
4.1500000000000004
|
2002-11-01 00:00:00
|
24
|
Firm Fixed
|
1
|
|
340
|
4.5999999999999996
|
2002-11-01 00:00:00
|
24
|
Firm Fixed
|
1
|
|
274
|
4.2298999999999998
|
2002-11-01 00:00:00
|
24
|
NA
|
1
|
|
193
|
7.3879999999999999
|
2005-02-01 00:00:00
|
24
|
Fixed
|
1
|
|
610
|
1
|
2005-11-01 00:00:00
|
12
|
Variable
|
1
|
Contractor
|
StateAbbrev
|
CName
|
Address1
|
Address2
|
Address3
|
City
|
Zip
|
CuCurrent
|
PaCurrent
|
PoCurrent
|
Contact
|
Ext
|
Fax
|
email
|
Status
|
|
OK
|
Tiger Natural Gas
|
1422 E 71st Suite J
|
NA
|
NA
|
Tulsa
|
74136
|
0
|
1
|
0
|
918-491-6998
|
0
|
918-491-6659
|
NA
|
1
|
|
GA
|
PS Energy
|
2987 Clairmont Rd
|
Suite 500
|
NA
|
Atlanta
|
30329
|
1
|
0
|
0
|
404-321-5711
|
NA
|
NA
|
NA
|
1
|
|
OK
|
Geary Energy
|
7712 S Yale Ave
|
Suite 201
|
NA
|
Tulsa
|
74136
|
1
|
0
|
0
|
918-523-2516
|
0
|
918-523-2522
|
NA
|
1
|
|
CO
|
Utility Resource Solutions
|
1700 Lincoln St.
|
Suite 2530
|
NA
|
Denver
|
80203
|
0
|
1
|
0
|
303-864-1919
|
0
|
NA
|
NA
|
1
|
|
CO
|
Select Natural Gas LLC
|
8122 Southpark Lane
|
Suite 204
|
NA
|
Littleton
|
80120
|
1
|
0
|
0
|
345-098-8890
|
456
|
NA
|
NA
|
1
|
Facility
|
FacilityCode
|
SCity
|
SZip
|
BName
|
BCity
|
BState
|
BZip
|
Mtelemetry
|
Cucustomer
|
Pacustomer
|
PoCustomer
|
InvoicesNeeded
|
InvoiceMonths
|
GSA
|
MajorTypeId
|
|
AL030004
|
Talladega
|
35160
|
NA
|
NA
|
NA
|
NA
|
0
|
0
|
0
|
1
|
0
|
NA
|
0
|
3
|
|
AR030004
|
Forrest City
|
72336-7000
|
NA
|
Forrest City
|
AR
|
72336-7000
|
0
|
0
|
0
|
1
|
0
|
NA
|
0
|
3
|
|
AR030104
|
na
|
75501
|
NA
|
na
|
AR
|
75501
|
0
|
0
|
0
|
1
|
0
|
NA
|
0
|
3
|
|
AR150232
|
Stuttgart
|
72160
|
NA
|
Stuttgart
|
AR
|
72160
|
0
|
1
|
0
|
0
|
NA
|
NA
|
0
|
15
|
|
AZ010001
|
Mesa
|
NA
|
NA
|
NA
|
NA
|
NA
|
0
|
1
|
0
|
0
|
0
|
NA
|
0
|
1
|
FacilityLDC
|
LDCId
|
FacilityCode
|
EstVolume
|
Unit
|
|
39
|
NY0270ZZ
|
9923
|
Mmbtu
|
|
25
|
PA010501
|
52681
|
Mmbtu
|
|
17
|
MI010201
|
NA
|
Mmbtu
|
|
20
|
NY0026ZZ
|
5621
|
Mmbtu
|
|
1
|
MA080502
|
14947
|
Mmbtu
|
LDC
|
StateAbbrev
|
Lname
|
Address1
|
Address2
|
Address3
|
City
|
Zip
|
Emergency
|
|
MA
|
Columbia Gas of Massachusetts
|
2025 Roosevelt Ave
|
NA
|
NA
|
Springfield
|
1104
|
800-688-6160
|
|
WA
|
Cascade Natural Gas
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
NA
|
Central Hudson Gas
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
OH
|
Cincinnati Gas & Electric
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
|
AZ
|
Citizens Utilities
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
Master_Facility
|
BldgNum
|
RegnNum
|
|
AK0001ZZ
|
NA
|
|
AK0005AK
|
NA
|
|
AK0009ZZ
|
NA
|
|
AK0013ZZ
|
NA
|
|
AK0020ZZ
|
NA
|
POC
|
StateAbbrev
|
Title
|
POCPosition
|
FirstName
|
LastName
|
NickName
|
Address1
|
Address2
|
Address3
|
City
|
Zip
|
Phone
|
Ext
|
Fax
|
Emergency
|
Email
|
CellPhone
|
Comments
|
|
NC
|
NA
|
Chief Financial Officer
|
Albert
|
Brese
|
NA
|
300 Morgan St.
|
Suite 1402
|
NA
|
Durham
|
27701
|
919-956-5541
|
NA
|
919- 956-7152
|
NA
|
al.brese@med.va.gov
|
NA
|
NA
|
|
AZ
|
NA
|
NA
|
Albert
|
Iamiceli
|
NA
|
500 Highway 89 North
|
NA
|
NA
|
Prescott
|
86313
|
928-717-7472
|
NA
|
716- 344-3303
|
NA
|
albert.iamiceli@med.va.gov
|
NA
|
NA
|
|
KS
|
Mr.
|
Boiler Plant Operator
|
Alex
|
Strouhal
|
NA
|
4101 S. 4th Street, Traff
|
NA
|
NA
|
Leavenwoth
|
66048
|
913-682-2000
|
x2951
|
913- 758-4230
|
NA
|
NA
|
NA
|
NA
|
|
DC
|
NA
|
Energy Coordinator
|
Amy
|
Hudson
|
NA
|
General Svcs Admin (WPY)
|
301 7th St. SW
|
Room 7512
|
Washington
|
20024
|
202-708-4995
|
NA
|
202-401-3722
|
202-497-6164
|
amy.hudson@gsa.gov
|
202-497-6164
|
NA
|
|
VA
|
NA
|
NA
|
Angel
|
Perry
|
NA
|
1234 I don’t know Ave.
|
NA
|
NA
|
Petersburg
|
0
|
804-504-7200
|
1067
|
804-504-7227
|
804-504-7200
|
aperry@bop.gov
|
NA
|
NA
|
tblCostUsageTest
|
Flag_DC
|
OLD_Penalty_Adj
|
OLD_Supplier_Adj
|
CostUsageId
|
LDC_Vol
|
LDC_Vol_Adj
|
LDC_Org_Cost
|
LDC_Cost_Adj
|
LDC_Misc_Fee
|
LDC_FinalVolume
|
LDC_FinalCost
|
LDC_UnitPrice
|
LDC_UnitPrice_Original
|
Supp_Vol
|
Supp_Vol_Adj
|
Supp_Org_Cost
|
Supp_Cost_Adj
|
Supp_Misc_fee
|
Supp_FinalVolume
|
Supp_FinalCost
|
Supp_UnitPrice
|
CU_Year
|
CU_Month
|
FacilityCode
|
Total_FinalCost
|
BT_Cost
|
|
NA
|
NA
|
NA
|
1
|
50.6
|
NA
|
139.97999999999999
|
NA
|
NA
|
50.6
|
139.97999999999999
|
2.7664031620553402
|
NA
|
50.6
|
NA
|
350.15199999999999
|
0
|
NA
|
50.6
|
350.15199999999999
|
6.92
|
NA
|
1
|
VA0003ZZ
|
490.13
|
9.69
|
|
DC
|
NA
|
NA
|
2
|
100.98
|
NA
|
312.95
|
NA
|
NA
|
100.98
|
312.95
|
3.0991285403050099
|
NA
|
100.98
|
NA
|
757.35
|
0
|
NA
|
100.98
|
757.35
|
7.5
|
NA
|
2
|
VA0003ZZ
|
1070.3
|
10.6
|
|
DC
|
NA
|
NA
|
3
|
237.02
|
NA
|
644.04
|
NA
|
NA
|
237.02
|
644.04
|
2.7172390515568301
|
NA
|
237.02
|
NA
|
1777.65
|
0
|
NA
|
237.02
|
1777.65
|
7.5
|
NA
|
3
|
VA0003ZZ
|
2421.69
|
10.220000000000001
|
|
DC
|
NA
|
NA
|
4
|
335.04
|
NA
|
877.87
|
NA
|
NA
|
335.04
|
877.87
|
2.6201946036294199
|
NA
|
335.04
|
NA
|
2512.8000000000002
|
0
|
NA
|
335.04
|
2512.8000000000002
|
7.5
|
NA
|
4
|
VA0003ZZ
|
3390.67
|
10.119999999999999
|
|
DC
|
NA
|
NA
|
5
|
242.33
|
NA
|
656.73
|
NA
|
NA
|
242.33
|
656.73
|
2.7100647876862101
|
NA
|
242.33
|
NA
|
1817.4749999999999
|
0
|
NA
|
242.33
|
1817.4749999999999
|
7.5
|
NA
|
5
|
VA0003ZZ
|
2474.1999999999998
|
10.210000000000001
|
tblInvoiceMaster
|
FacilityCode
|
InvoiceTypeID
|
InvYear
|
InvMonth
|
InvFileName
|
InvFileName1
|
InvFileName2
|
|
VA0422ZZ
|
1
|
2007
|
7
|
05-01-2007_11_33_22_3096668631.307.pdf
|
NA
|
NA
|
|
VA0422ZZ
|
2
|
2007
|
7
|
05-01-2007_11_38_47_3096668631.307.pdf
|
NA
|
NA
|
|
VA0422ZZ-01
|
1
|
2007
|
7
|
05-01-2007_11_51_41_3096915545.407.pdf
|
NA
|
NA
|
|
VA0422ZZ-01
|
2
|
2007
|
7
|
05-01-2007_11_52_24_3096915545.407.pdf
|
NA
|
NA
|
|
KY010101
|
2
|
2007
|
7
|
05-01-2007_01_31_18_saglexcoop407.pdf
|
NA
|
NA
|
tblInvoiceStatus
|
StatusDate
|
Status
|
|
2007-05-01 00:00:00.000
|
Uploaded/Verified
|
|
2007-05-01 00:00:00.000
|
Uploaded/Verified
|
|
2007-05-01 00:00:00.000
|
Uploaded/Verified
|
|
2007-05-01 00:00:00.000
|
Uploaded/Verified
|
|
2007-05-01 00:00:00.000
|
Uploaded/Verified
|
The one table with enough data to analyze: monthly cost &
usage
tblCostUsageTest is the only table with more than a few
rows of numeric data, and it’s the closest thing in this file to a real
analytic table: 5 consecutive monthly cost/usage
records for a single facility (VA0003ZZ), tracking
gas volume and cost separately on the LDC (delivery) side and the
supplier (commodity) side.
cu <- ngap$tblCostUsageTest %>%
mutate(across(c(LDC_Vol, LDC_Org_Cost, LDC_FinalVolume, LDC_FinalCost, LDC_UnitPrice,
Supp_Vol, Supp_Org_Cost, Supp_FinalVolume, Supp_FinalCost, Supp_UnitPrice,
CU_Month, Total_FinalCost, BT_Cost), as.numeric))
cu %>%
select(CU_Month, FacilityCode, LDC_FinalVolume, LDC_FinalCost, LDC_UnitPrice,
Supp_FinalVolume, Supp_FinalCost, Supp_UnitPrice, Total_FinalCost) %>%
kable(caption = "Monthly cost and usage for facility VA0003ZZ (5-month sample)") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)
Monthly cost and usage for facility VA0003ZZ (5-month sample)
|
CU_Month
|
FacilityCode
|
LDC_FinalVolume
|
LDC_FinalCost
|
LDC_UnitPrice
|
Supp_FinalVolume
|
Supp_FinalCost
|
Supp_UnitPrice
|
Total_FinalCost
|
|
1
|
VA0003ZZ
|
50.60
|
139.98
|
2.766403
|
50.60
|
350.152
|
6.92
|
490.13
|
|
2
|
VA0003ZZ
|
100.98
|
312.95
|
3.099128
|
100.98
|
757.350
|
7.50
|
1070.30
|
|
3
|
VA0003ZZ
|
237.02
|
644.04
|
2.717239
|
237.02
|
1777.650
|
7.50
|
2421.69
|
|
4
|
VA0003ZZ
|
335.04
|
877.87
|
2.620195
|
335.04
|
2512.800
|
7.50
|
3390.67
|
|
5
|
VA0003ZZ
|
242.33
|
656.73
|
2.710065
|
242.33
|
1817.475
|
7.50
|
2474.20
|
cu_long <- cu %>%
select(CU_Month, LDC_UnitPrice, Supp_UnitPrice) %>%
pivot_longer(-CU_Month, names_to = "Cost Component", values_to = "Unit Price") %>%
mutate(`Cost Component` = recode(`Cost Component`,
LDC_UnitPrice = "LDC (delivery) unit price",
Supp_UnitPrice = "Supplier (commodity) unit price"))
ggplot(cu_long, aes(x = CU_Month, y = `Unit Price`, color = `Cost Component`)) +
geom_line(linewidth = 1) +
geom_point(size = 2.5) +
scale_x_continuous(breaks = 1:5) +
scale_y_continuous(labels = dollar) +
labs(title = "Unit price by cost component, Facility VA0003ZZ",
x = "Month (sample sequence)", y = "Unit price ($/unit)", color = NULL) +
theme_minimal(base_size = 12) +
theme(legend.position = "bottom")
Two things stand out even in this small sample: the supplier
(commodity) unit price is consistently more than double the LDC
(delivery) unit price for every month shown — expected, since
the commodity itself typically costs more than the utility’s delivery
charge — and the supplier price jumps from $6.92 to $7.50
between month 1 and month 2 and then holds flat, while the LDC
delivery price fluctuates modestly month to month. Total combined final
cost across the 5-month sample sums to $9,846.99.
Data quality observations
Even at this small scale, several patterns are worth flagging for
anyone building a pipeline on top of the full production NGAP
database:
tibble(
Issue = c(
"Literal text \"NULL\" instead of a true empty/NA value",
"Trailing whitespace padding on fixed-width codes",
"Inconsistent missing-value conventions across tables",
"Placeholder/junk data in free-text fields",
"A ZIP code of \"0\" on one POC record"
),
Example = c(
"`CFileName`, `Address2`, many `Contractor`/`Facility`/`POC` fields",
"`Master_Facility.BldgNum` values like `\"AK0001ZZ \"`",
"`LDC` uses literal `\"N/A\"` strings while other tables use `\"NULL\"`",
"`POC.Address1` = \"1234 I don't know Ave.\" for contact Angel Perry",
"`POC` record for Angel Perry (Petersburg, VA) has `Zip = 0`"
)
) %>%
kable(caption = "Data quality issues visible in this sample") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)
Data quality issues visible in this sample
|
Issue
|
Example
|
|
Literal text “NULL” instead of a true empty/NA value
|
CFileName, Address2, many
Contractor/Facility/POC fields
|
|
Trailing whitespace padding on fixed-width codes
|
Master_Facility.BldgNum values like
"AK0001ZZ "
|
|
Inconsistent missing-value conventions across tables
|
LDC uses literal "N/A" strings while other
tables use "NULL"
|
|
Placeholder/junk data in free-text fields
|
POC.Address1 = “1234 I don’t know Ave.” for contact Angel
Perry
|
|
A ZIP code of “0” on one POC record
|
POC record for Angel Perry (Petersburg, VA) has
Zip = 0
|
The parsing step above already converts literal "NULL"
strings to proper NA and trims whitespace from every field,
which is a necessary first step before any of these tables could
be loaded into a real database or joined reliably on their key
fields — a code like "AK0001ZZ" with 14 trailing
spaces will fail to match a clean "AK0001ZZ" in a join even
though they look identical when printed.
Explore the data
cu %>%
select(CU_Month, FacilityCode, LDC_FinalVolume, LDC_FinalCost, LDC_UnitPrice,
Supp_FinalVolume, Supp_FinalCost, Supp_UnitPrice, Total_FinalCost, BT_Cost) %>%
datatable(options = list(pageLength = 10, scrollX = TRUE), rownames = FALSE)
Key takeaways
- This is a schema sample, not a population dataset:
11 relational tables with 4-15 rows each, meant to illustrate the
structure of GSA’s NGAP database rather than support statistical
analysis.
FacilityCode is the connective key
across contracts, LDC assignments, cost/usage records, and invoices — a
real analysis of the full database would join on this field.- The only table with enough rows to chart,
tblCostUsageTest, shows supplier (commodity) gas
consistently costing more than double the LDC (delivery) charge
per unit, with the supplier price jumping once early in the sample
window and then holding flat.
- The sample itself demonstrates real-world data-quality
issues worth fixing before production use: literal
"NULL" text instead of true nulls, inconsistent
missing-value conventions between tables, trailing whitespace on key
fields, and at least one clearly invalid value (a 0 ZIP
code).
Limitations
With only a handful of rows per table, no finding here should
be read as representative of the full NGAP program — this file
is best understood as a structural and data-quality reference. A
meaningful analysis of natural gas costs, contract terms, or facility
coverage would require the full production database behind this schema,
most likely queried directly rather than distributed as a stacked-table
spreadsheet extract.