From an AGS4 file to a laboratory summary: what the groups carry

Updated: 2026-08-13 · ISKRIT

An AGS4 file from a laboratory is not a report. It is a set of linked tables, and turning it into a summary means walking those links correctly. Most of the work is in the links, not in the numbers.

The chain

Every laboratory result hangs off a chain: the location group holds the borehole, the sample group holds the sample taken from it at a depth, and each test group holds results referenced to that sample. Break any link and the result is orphaned: a plasticity index with no depth, or a grading curve belonging to a borehole that is not in the file.

Where summaries go wrong

The first failure is unit assumptions. A field carrying moisture content as a percentage in one file and as a fraction in another produces a summary that is out by a factor of a hundred and looks entirely plausible on a chart with auto-scaled axes.

The second is duplicate sample references. Laboratories reissue files, and a merged set can hold the same sample reference twice with different results. Silently taking the last one is how a superseded result reaches a report.

The third is partial tests. A grading with no hydrometer stage stops at 0.063 mm, and a summary that computes d10 from it is extrapolating past the data without saying so.

What a summary should carry alongside the number

The test method, the sample type, and the count of results behind each statistic. A mean plasticity index over eleven samples and over two samples are both a number in a cell, and only one of them supports a design decision.

Grouping by unit, not by borehole

Results are collected per borehole because that is how they were sampled, but they are used per geotechnical unit. The regrouping step is where a mismatch surfaces: a unit whose samples come from a single borehole is not characterised, however many tests were run on them.

The AGS4 laboratory tool walks the location, sample and test chain, reports orphaned and duplicated references before it computes anything, and shows the count behind every statistic in the summary.