Method: The issue library — backgrounders, briefs, cards, discussion guides

Most of what a reader meets on this site isn't the full book-length treatment of an issue — it's a shorter piece built for a specific purpose: a neutral backgrounder that lays out what's known, a localized brief, a single recommendation you could hand to a candidate, or a discussion guide you could take into a room. This page explains how each of those gets built and checked, as a lighter, faster instance of the same discipline behind our full research process.

What it produces

Four kinds of pieces, each with its own job:

• Backgrounders — the neutral summary of what's known about one issue, with no "for" or "against" framing built in. Every load-bearing statement in a backgrounder traces to a specific checked source.
• Briefs — a backgrounder applied to a specific place, citing back to the backgrounder section it draws from.
• Recommendation cards — a single, concrete ask, built strictly inside a fixed template rather than as free-form writing, so that every card states its position the same structured way and none smuggles in a claim the template didn't check for. Cost estimates in a card use rounded, order-of-magnitude ranges and named comparison points — never fake-precise numbers dressed up to look exact.
• Discussion guides — neutral questions built from already-checked material, meant to be taken into an actual conversation: a dinner table, a community meeting, a candidates' debate. A guide never argues a position; it opens one up for discussion.

How it works, step by step

Before any new research goes into an issue, we first check what we already have on file for it and credit that existing coverage — so we never spend effort re-researching ground already covered.

Backgrounders come first: the neutral synthesis for one issue, written so that every meaningful statement in it is backed by a specific, checked source, and written to keep out any hint of "this side is right." A fixed template exists specifically to enforce that neutrality — it's easier to stay neutral when the structure itself resists slipping into advocacy.

Briefs are built one layer down from a backgrounder, applying it to a specific place and citing back to the exact backgrounder section a given line comes from.

Recommendation cards are the single, atomic unit — one clear, adoptable ask. A card is built only from already-checked material: a backgrounder, a checked claim, a named comparison point from elsewhere. It is never written as freehand prose making an argument; it lives entirely inside the card template's fixed structure, which is what keeps every card checkable the same way. We also insist that a card's ask names something a public office-holder can actually do — not an internal task of ours dressed up as if it were addressed to a decision-maker. Any card that fails that test gets pulled from anything a reader or candidate would see.

Discussion guides sit apart from the other three: they add no new research and check no new claims of their own. A guide is built entirely from material that's already been through verification elsewhere, and it inherits that material's checked status rather than being independently re-verified itself. The governing question for a guide isn't "is this true" — that was settled upstream — it's "can a citizen take this question to a real conversation and have it land as genuinely open," never as a leading question dressed up as neutral.

Before anything in a batch goes out, an independent reviewer — someone with no part in drafting that batch — checks it over on their own before any fixes are made, so problems get caught by a second, disinterested set of eyes rather than only by the people who wrote it.

Rendering the pieces into their public shape runs as its own mechanical step, kept deliberately separate from the research and drafting stages.

How it's checked

The bar here is deliberately concrete, so an outsider has something specific to check us against:

• Backgrounders, briefs, and cards get the same independent, source-by-source re-check described on our flagship research method page — a separate checker re-fetches the original source for every claim that matters, rather than trusting the draft.
• The evidence rule has two tiers, and we say which one a claim is using. When we say a document states or found something, we mean we read that document itself — a summary of it, or someone else's account of what it says, never substitutes for the original. When we're instead stating a fact about the world, and the original source is genuinely unreachable after a real effort to find it — paywalled, offline, never digitized — we still require at least two independent accounts that agree before it goes in, and we say plainly, on the claim's own face, that it rests on that weaker basis rather than a source we read directly. A reachable original always wins over any number of secondary accounts.
• Discussion guides are checked for accurate inheritance, not re-verified from scratch. Because a guide adds nothing new, what we check is narrower but still real: does the guide's stated status honestly match the status of the material it's built from? A guide is never allowed to claim a stronger, more-verified status than its own source material actually holds.
• Every card is uniquely identifiable. A recommendation can't be tracked, referenced, or held against a candidate's answer if its own identifier isn't unique — we treat that as a hard, mechanical requirement, not a nice-to-have, and we caught and fixed a real batch of duplicate identifiers precisely because we check for this.
• Numbers are read from the record, not restated from memory. Any count or tally we report is pulled fresh from the underlying record at the time of reporting, because we've caught our own summary write-ups disagreeing with their own underlying numbers before — and the fix was to stop trusting a prior summary and always recompute from source.
• A bad claim's damage is traced everywhere it spread. If a claim is later found wrong or disputed, every other place that cited it gets swept and corrected — a systematic, checkable pass across the whole corpus, not a best guess at what might reference it. See How we verify for how that sweep runs across the site as a whole.

What it honestly costs

Estimates below, drawn from one real batch of work we've actually run — not extrapolated beyond it.

Backgrounder, brief, and card drafting is steady, methodical work rather than the kind of judgment call that needs our most careful review. The heavier judgment calls — auditing a large batch of cards for quality, or resolving a genuinely disputed question about what standard of evidence a claim needs to meet — get our most careful attention specifically because those are the calls that are hard to get right and expensive to get wrong.

On real scale: one audit pass verdicted every card in a batch of roughly three hundred across well over a hundred files, and hand-read a meaningful fraction of them directly rather than relying only on automated checks. A separate independent review re-derived every summary number across more than twenty files from scratch in one sitting, including re-checking a document and two initially-blocked sources live, with the reviewer having no prior involvement in the work under review.

It also costs operator time: a short list of ranked decisions after each batch, plus any question serious enough that it needs a person, not a process, to rule on it.

Known limits

Named plainly, not smoothed over.

• There's no single frozen list of every issue we've committed to cover. Our internal tracking is a live, current status board, not a fixed roster — so any claim about how much of the corpus is complete has to point at that current state, dated, rather than being stated as a settled, permanent count.
• The independent-review step isn't a standing rule yet for every future batch. We've run it as a one-time, dedicated check for specific batches so far, and it worked well enough that we think it should become a standing practice — but as of this writing that commitment isn't yet formalized.
• How deep a check goes into a backgrounder's connecting prose (as opposed to just its explicit sourced claims) isn't written down as a fixed rule. In practice we've focused checking on the explicit sourced claims, but we haven't yet stated that scope as a formal rule either way.

What stays internal, and why

A backgrounder, brief, card, or discussion guide is a working document until a deliberate, separate decision publishes it — the fact that something has been checked and sealed internally is not by itself a decision to put it in front of the public. Some material, particularly anything that would name or grade specific organizations or candidates in a way that could carry legal weight, sits behind an additional layer of outside legal review before it can go public. Nothing in this process authorizes publication on its own; publishing is always its own separate, intentional step.

Every load-bearing claim on this site carries a receipt regardless of who drafted it, including where AI assistance was used to draft the underlying research. Run this in your municipality if you want to build the same kind of library for your own community, or read How we verify for how this checking discipline runs across the whole site.

— The Unknown Soldier · uniteTOlove · Toronto