One worksheet row

The counting unit is a row of the original DIR workbook, including repeated case identifiers. The retained set has 84,093 rows and 81,783 distinct nonblank case identifiers. Neither number establishes a count of employers, lawsuits, employees, violations, or adjudicated outcomes.

Filing dates come from column F. The drawing includes January 1, 2014 through December 31, 2025, divided into three inclusive periods: 2014–2017, 2018–2021, and 2022–2025. The 76,757 window rows consist of 50,478 geographically eligible rows and 26,279 unmapped rows. The complete record browser also includes 7,322 dated rows outside the window and 14 undated rows.

The record view preserves columns B and D–K exactly as extracted, with original worksheet cell addresses. Original city, state, and postal-code values remain visible alongside the derived geographic result. City names are not used as a fallback for missing or inconsistent state and ZIP fields.

The geographic rule

1. Trim outer whitespace in the recorded employer state; accept CA or California, ignoring letter case. A missing state or any other label stops geographic assignment.
2. Trim outer whitespace in the recorded ZIP. Accept five ASCII digits, or five digits followed by a hyphen and four digits. Use the first five digits of ZIP+4. No zero padding, punctuation repair, nine-digit splitting, or inference from city names is performed.
3. Match that five-digit value exactly to a GEOID in the 2020 national ZCTA Gazetteer. Use its unrounded INTPTLAT and INTPTLONG values.
4. Require that point to fall inside latitude 32.5–42.1 and longitude −124.5–−114.0, including the edges. This authored rectangle is a coarse plausibility screen only. It includes places outside California and does not validate the state boundary.
5. Assign the point to the nearest of the eight Census place internal points below, using great-circle distance on a sphere of radius 6,371.0088 km. An exact tie follows the table’s order.

The Census describes ZCTAs as geographic representations of ZIP delivery areas, constructed from census blocks. Some valid ZIP codes have no ZCTA. The reference is fixed to 2020 for all filing dates; it is not historical address geocoding or a claim about current ZIP validity. See the Census explanation and record layouts.

The state label and ZIP-area point describe fields in the employer record. They do not identify a workplace, headquarters, filing court, street address, or county. In particular, the group names below are names for authored nearest-anchor groups, not official regional boundaries.

Eight anchors

Coordinates and one-based physical line numbers come from the retained California places Gazetteer; the header occupies line 1. The manifest also retains each complete anchor source line.

Oakland city supplies the Bay anchor. The same source places San Francisco city’s internal point offshore at 37.727239, −123.032229 (line 1267); that published value was neither moved nor treated as a downtown location. Oakland provides a mainland source point without altering the Census coordinates.

For every eligible row, the source browser supplies the matched ZCTA, both reference coordinates, source lines, nearest anchor, and calculated distance. That distance is between reference points; it does not measure travel or distance from an employer’s street address.

What the marks carry

The source distribution contains 24 cells: eight groups × three filing periods. For a drawing budget N, each cell first receives the floor of N × cell count ÷ 50,478. Remaining whole marks go to the largest fractional remainders, with ties ordered by the anchor table and then filing period. The allocations sum exactly to the selected drawing budget.

Every mark keeps its group and filing period when it respawns. Its represented weight is the cell’s source-row count divided by its allocated marks. A cell receiving no mark still remains in the counts. A mark does not correspond to an individual worksheet row. Trails and repeated motion add no source records.

The seed changes placement. Cluster pull, flow, noise, and fading change motion and rendering. Color distinguishes filing periods. These controls do not change the source totals or geographic assignments. Selecting a group identifies its center. Group and period selections change the displayed counts and linked row set, while the complete field remains visible.

The original 1,100 × 760 drawing, eight anchor positions and spreads, motion equations, paper colors, typefaces, and desktop composition are retained. Canvas positions are illustrative and not projected geographic coordinates. The original hand-set regional weights and unsupported theory colors have been replaced by the measured group-and-period distribution.

Reading and retaining the result

Use the group and period selectors, tap a center, or focus the canvas and use arrow keys to inspect a group. Escape clears the group. The source link follows the selection. The phone layout fits the field to the available width; “Enlarge drawing” exposes its full logical width in a scrollable panel.

“Download PNG” preserves the full 2,200 × 1,520 drawing and adds a source caption below it. The caption records snapshot, window, counted rows, mark budget, selection, seed, and source-library location. “Download source counts” retains all 24 cells, original counts, allocated marks, weights, and linked source-row scopes. The row browser exports its complete filtered set, rather than only the displayed page.

Reduced-motion preferences start the field paused. The pause control also freezes the current composition. Source loading and validation failures show a retry control instead of a fabricated distribution.