# Reporting fields in 28 ChemVerify sample records

A frozen ChemVerify export contains **28 sample records, 104 lab report records and 216 result rows**. Numbers and units appear together in 82 result rows. Another 108 rows carry the categorical verdict `not_detected`, 22 carry `pass`, and four have neither a numeric value nor a verdict. This case study measures which fields the public export contains. [ChemVerify dataset](https://www.chemverify.com/tests) · [Public JSON export](https://www.chemverify.com/tests/data.json)

Lab Data Field Guide is a ChemVerify-associated educational publication. ChemVerify published the export and says it commissioned the laboratory testing. This analysis has not received an independent expert review.

## Count samples, reports and results separately

ChemVerify purchased the samples and commissioned testing. The snapshot covers a nonrandom collection associated with ChemVerify. It contains records for five vendors, with 20 of 28 samples from one vendor, and lists one laboratory, Janoshik Analytical. The analysis includes every record in the snapshot. It counts each top-level sample once, each nested report once and each nested result once. Sample protocol URLs and report references are unique in this file. A report can have several result rows, so 216 rows represent 216 entries at the result level. [ChemVerify dataset and collection description](https://www.chemverify.com/tests)

The sample field `analysed_at` runs from September 2 to November 11, 2026. The report field `tested_at` runs from August 28 to November 11, 2026, matching the date range declared at the top of the export. In 65 report records, the test date differs from the parent sample's analysis date. Use the field that corresponds to the event you are counting. [Public JSON export](https://www.chemverify.com/tests/data.json)

## Which fields were populated?

A field was counted as populated when its key held a non-null, nonempty value. Explicit numeric zero would count as populated. Every observed key existed on every record at its own level in this snapshot. Unpopulated entries in the table were all JSON `null`.

| Level | Field | Populated | JSON null |
| --- | --- | ---: | ---: |
| Sample | Label batch number | 5 of 28 | 23 |
| Sample | Label quantity in milligrams | 22 of 28 | 6 |
| Sample | Received date | 13 of 28 | 15 |
| Sample | Date sent to the lab | 13 of 28 | 15 |
| Report | Report URL | 104 of 104 | 0 |
| Report | Report method label | 104 of 104 | 0 |
| Result | Numeric value | 82 of 216 | 134 |
| Result | Unit | 82 of 216 | 134 |
| Result | Result method label | 109 of 216 | 107 |
| Result | Uncertainty | 0 of 216 | 216 |

The report and result method labels answer different coverage questions. All 104 reports had a method label. Only 109 result rows had one, and nine of those labels differed from the corresponding report label. Keep the fields separate if you need to trace how the export describes each measurement. A null uncertainty field records what the export provides. It gives no basis for assigning zero uncertainty or determining what appears in the underlying report.

## Keep numbers, bounds and verdicts distinct

| Result representation | Count out of 216 |
| --- | ---: |
| Number with `eq` | 68 |
| Number with `lt` | 13 |
| Number with `gt` | 1 |
| `not_detected` verdict with null numeric value | 108 |
| `pass` verdict with null numeric value | 22 |
| Null numeric value and null verdict | 4 |

All 82 numeric rows have a unit and none contains explicit zero. They comprise 31 content rows in `mg`, 26 endotoxin rows in `EU/Vial` and 25 purity rows in `%`; these are the original unit strings, without inferred denominators. All 14 inequality values belong to endotoxin entries. A number with `lt` or `gt` preserves a bound, so an analysis of measured magnitudes needs to keep those qualifiers. The 108 `not_detected` rows are heavy-metal entries, four per report across 27 reports. Each of these rows has null numeric value and unit. Encoding their values as zero would introduce 108 zeroes absent from the export. The 22 `pass` rows are LC–MS entries without a numeric value. The four rows without a number or verdict are labeled Endotoxin, Benzyl Alcohol, pH and Comment. These are entries in nested results arrays, not necessarily 216 analytical measurements. The comparator is present on all 216 rows, including 134 rows without a number, so a comparator by itself does not identify a numeric observation.

## Reproduce the field counts

Download the frozen JSON snapshot and the Python analysis. Run `python3 analyze.py` alongside the snapshot metadata to regenerate the three aggregate CSV files and the summary. The field counts describe this public export and cannot establish the completeness of the original lab reports. The collection supports descriptive examples of reporting structure. Its sampling and single-laboratory coverage limit comparisons and generalization.

Source credit goes to ChemVerify (2026), *Independent peptide lab tests* [data set], licensed under CC BY 4.0. Lab Data Field Guide counted and grouped fields without changing or imputing source values. The unchanged bytes were verified by HTTPS retrieval at **2026-11-12 11:24:41 UTC**. SHA-256 `451bee625402c1f15b7cf29ea6cbdf46d7bd915631aec9aef0753bca6c470ad8`. [ChemVerify dataset](https://www.chemverify.com/tests) · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)