How well do Unum11 reports actually score?
Six companies, from a local restaurant to a $3.9B contractor, graded on
the same 1000-point rubric across
9 categories. The best scored
887; the median is
848. Every score below is reproducible
from the delivered report.html, and the whole set is
available as machine-readable JSON.
- Companies graded
- 6
- Best score
- 887
- Median score
- 848
- Words written
- 35,237
- Outbound citations
- 245
- Measured
Full results by company
| Company | Entity scale | Score | Words | Sections | Citations | Graded |
|---|---|---|---|---|---|---|
| Digerati Technologies, Inc. | Small-cap public | 887 | 8,176 | 9 | 29 | |
| KORU Medical Systems, Inc. | Small-cap public | 886 | 6,630 | 14 | 82 | |
| Xenetic Biosciences, Inc. | Small-cap public | 849 | 3,969 | 13 | 9 | |
| Edible Garden AG Inc | Small-cap public | 846 | 9,367 | 19 | 71 | |
| Granite Construction Incorporated | Global enterprise | 817 | 4,296 | 13 | 17 | |
| Peppers Cantina | Local SMB | 750 | 2,799 | 15 | 37 |
Per-category scores
The rubric splits into a 584-point floor awarded for having no detectable defect and 416 earned points that start at zero. A flawless report demonstrating nothing scores 584, not 1000.
| Company | A | B | C | D | E | F | G | H | I | Total |
|---|---|---|---|---|---|---|---|---|---|---|
| Digerati Technologies, Inc. | 171 | 95 | 128 | 123 | 127 | 68 | 61 | 38 | 76 | 887 |
| KORU Medical Systems, Inc. | 191 | 95 | 143 | 103 | 111 | 68 | 61 | 38 | 76 | 886 |
| Xenetic Biosciences, Inc. | 171 | 85 | 143 | 103 | 111 | 68 | 61 | 31 | 76 | 849 |
| Edible Garden AG Inc | 171 | 95 | 143 | 103 | 111 | 55 | 61 | 31 | 76 | 846 |
| Granite Construction Incorporated | 171 | 85 | 143 | 103 | 95 | 68 | 45 | 31 | 76 | 817 |
| Peppers Cantina | 175 | 95 | 127 | 123 | 20 | 82 | 30 | 31 | 67 | 750 |
| Maximum | 191 | 95 | 143 | 143 | 143 | 95 | 76 | 38 | 76 | 1000 |
- A Zero fabrication
- B Entity scale
- C Financial rigour
- D Simulation insight
- E Recommendation substance
- F Research completeness
- G Narrative coherence
- H Honesty
- I Visual and technical
What the numbers say, including the unflattering part
Recommendations are the weakest category, by a wide margin
Category E ranges from 127/143 down to 20/143. The floor case is the local restaurant, where the pipeline researched well (F: 82/95, the highest in the set) and then had almost nothing prescriptive to say. Research quality and recommendation quality are not the same capability, and this benchmark shows the gap rather than averaging it away.
Report length does not predict score
The 849-point Xenetic report is 3,969 words with 9 citations. The 846-point Edible Garden report is 9,367 words with 71 citations — more than twice as long, eight times the sources, three points lower. Length is what a thin public record does not force, and the rubric is largely indifferent to it.
Entity scale held across a 1,000× size range
Category B stayed at 85–95 out of 95 from a single-location restaurant to a $3.9B national contractor. Wrong-scale advice is one of the six zero-tolerance rules, so this is the category most likely to veto a report outright, and it did not.
Nobody has scored full marks on the same category twice
KORU took 191/191 on zero-fabrication and 143/143 on financial rigour; no other run matched either. Digerati leads on recommendations at 127/143 and gives up 15 points on financial rigour to do it. The ceiling is real, and no single run is near it everywhere.
How to read these numbers
This benchmark is self-administered. The rubric was written by Unum11 and run against Unum11's own output. The honest framing for anyone citing it is "Unum11's own benchmark". A real McKinsey Global Institute report scores 409/1000 on this same rubric with 3 categories at zero — a measurement of our house bias, not a verdict on McKinsey.
Companies named above are public companies analysed from public filings,
or a local business analysed from public reviews. None is a Unum11
client; nothing here was produced for or reviewed by them. Scores are
reproducible by running grade_report() against the
report.html in each run directory. Dataset published under
CC BY 4.0 — quote it, with attribution. See also
what a report contains and
the stated limitations.