Evidence

Do not confuse where the failure appeared with where the failure was designed.

These are patterns, recorded in the same structure every time. We state what we found. Where a claim would require asserting cause, we mark it and hold it for human review instead of printing it.

How we record evidence

Every record moves through five fields.

What we were shown
The account as it arrived: the decision, the document, the stated reason.
What we found underneath
The structure producing that account — rules, evidence standards, hand-offs, authority.
What we architected
The work built in response: costing, evidence structure, capital route, sequence.
What changed
Not stated. No outcome is published here until a human has verified it.
What remains to be proven
The open question, named plainly rather than filled in.

The records below are patterns recorded in this format. They are not client case studies, and they carry no outcome figures.

What follows is not proof of outcomes. It is the discipline itself: three patterns DSG encounters repeatedly, each recorded in the exact structure every DSG record uses. Where a line would require asserting that one thing caused another, it stays unstated and goes to human review. Verified engagement records will be published here as clients clear them and the documentation is complete.

Community organization · capital access

The organization was read as not-yet-ready.

Recorded pattern · not a client case

What we were shown

A declined proposal and an invitation to a readiness cohort. The stated reason was organizational capacity.

What we found underneath

The capacity language was describing a documentation gap, not an operating gap. The work being financed was already running; the evidence for it lived in relationships and staff memory rather than in a form the reviewer was permitted to credit.

What we architected

A documentation and evidence structure that carried what was already true, plus a corrected cost basis and a capital route matched to the actual decision-maker.

What changed

Not stated. No outcome is published here until a human has verified it.

What remains to be proven

Whether the correction moves the decision, and whether the reviewer's standard changes on the next cycle.

Youth and family systems

The value was called priceless and funded at zero.

Recorded pattern · not a client case

What we were shown

A program budget that counted staff time and supplies and nothing else.

What we found underneath

Trust, translation, transportation navigation, and after-hours availability were part of the mechanism producing participation and retention. All of it was uncosted, and most of it was being absorbed personally.

What we architected

A costing model that names the mechanism, an evidence set for the mechanism where the record supports it, and a capital proposition addressed to the parties already benefiting from the result.

What changed

Not stated. No outcome is published here until a human has verified it.

What remains to be proven

Causal weight of each mechanism component. Attribution here requires human review before it is stated as a finding.

Public and institutional capital

The money left the room and did not arrive.

Recorded pattern · not a client case

What we were shown

An announced allocation with a published total and a defined service area.

What we found underneath

The allocation moved through intermediaries with their own compliance costs, risk postures, timelines, and eligibility interpretations. Each pass narrowed who could reach it.

What we architected

A map of the middle — every hand-off, rule, and gate between the allocation and the consequence — used to reposition where the ask actually needed to land.

What changed

Not stated. No outcome is published here until a human has verified it.

What remains to be proven

Amount lost at each hand-off. Reported as an open question rather than a figure.

DSG Evidence Standard

When a conclusion depends on saying one thing caused another, DSG requires human review before that conclusion is included.

We do not invent statistics, ROI figures, funder names, or motives. Community intelligence — what your community already knows and tells you, even before it's written down anywhere — is first-class evidence. Where the record is thin, the gap gets named.

Read the claim ledger — every public claim, its evidence class, and its sources

How evidence actually works

The part most applications skip: the mechanism

A mechanism is the step-by-step reason a change happens. Not what you did, and not what changed — the part in between that explains how one led to the other.

Funders rarely reject a number because it is small. They reject it because nothing explains it. When you can name the mechanism, a modest number becomes evidence. When you cannot, a large number becomes a coincidence.

The five links in the chain, what proves each one, and what would break it. This table carries the same content as the diagram above.
Link	In plain words	Evidence you can show	What would break it
What you put in	The people, hours, money, food, space or transport you actually spend.	Staff and volunteer hours, budget lines, donated goods, site records.	You cannot show the input was ever delivered at the level you describe.
What you do	The activity itself, described concretely enough that someone else could repeat it.	Session logs, attendance, delivery counts, schedules.	The activity varies so much that two participants received different programs.
What changes first	The nearest, most immediate change — the one you can see within weeks.	Before-and-after counts, short surveys, staff observation records.	Nothing measurable moves early, which usually means the chain has a missing link.
Why that leads to the next thing	The reason the first change makes the later change more likely. This is the mechanism.	Published research on the same link, plus your own pattern over time.	The link only holds under conditions your participants do not live in.
What changes later	The outcome you want to claim.	Longer-term records, partner data, follow-up contact.	Something else in the same period explains the change at least as well.

How strong is the link, and what may you say about it?

Each strength of evidence and the language it permits.
Strength	What it means	Language you may use
Counted	You have the record and can show it.	"We delivered 412 meals in March. Here is the log."
Observed pattern	You see it repeat, but you have not ruled out other explanations.	"Attendance rose in the months we ran evening sessions."
Supported by outside research	Someone else studied this link in a comparable setting.	"Research on this link reports X. Our setting matches it in these ways."
Community intelligence	People who live the conditions know why it works. This is real evidence, not a lesser kind.	"Families told us the barrier was transport, not interest, and the schedule change tested that."
Caused	You can show the change would not have happened otherwise. This needs a comparison.	Held for human review. DSG does not publish a causal claim without it.