Methodology: how this was built, and where it stops

A polemic that cheats on its numbers hands the people it indicts an easy escape. One wrong figure, and the whole case goes with it. So the facts here are clean, the confidence in each is labelled, and every place the record complicates the thesis is stated up front — before a hostile reader finds it.

Where the facts come from

Council votes come from Edmonton's Open Data portal: dataset ekzq-b4qg for the 2021–2025 council, huhs-auh6 for 2017–2021 (datasets). Every vote attributed to Janz or Knack was pulled from those tables by councillor column, not from a reporter's summary. The recipe for repeating that pull lives in pipeline/edmonton-open-data-pipeline.md.

Everything else — encampment counts, utility dividends, food-bank numbers, rate filings, the two men's own words — comes from named sources archived under sources/. Each clipping carries the key passages, the original URL, and the retrieval date, so the citation outlives the page. Primary documents beat paraphrases: an EPCOR rate application over a story about it. Every figure is in Canadian dollars with its year noted. Nothing hides behind an exchange rate.

Two confidence tiers

Confirmed means the number is nailed to a primary or well-corroborated source and can be headlined as written. Contested does not mean doubtful — it means the honest reading is more careful than one clean number, because a measure was rebased, a source was secondary, or the point is an estimate between confirmed anchors. Contested claims carry the complication at the point of the claim, not in a footnote. If it says confirmed, we defend the exact number. If it says contested, we are showing you the edge before you find it.

The honesty ledger

Every complication below is load-bearing, and the argument survives all of them.

Where the record complicates the thesis

Janz does not support police. He moves the cuts, casts the dissenting vote, and loses: 5–8, 9–4, 12–1. The charge is the two-step that follows — he votes yes on the 2023 operating budget that funds the police he just failed to cut. Dissent that converts back into a signature is the thing being named. See the police formula.

Homicides did not rise on a trend. Edmonton's rate is volatile: it peaked around 2017 and 2021, then fell to 2.20 in 2024, among the decade's lowest. We claim no rising wave. The truer argument: violence spiked during the pandemic while the police budget grew on a locked formula regardless, and what climbed in 2024 was criminalized poverty — opioid possession up 152%, meth up 116%, shoplifting under $5,000 up 12%. See community and frontline.

Poverty is entrenched, not measured-to-have-risen. The 2015 and 2024 figures use different measures; the Market Basket Measure was rebased between them. Poverty stayed at roughly one in ten across a decade. We put no false decimal-point rise on it.

The electricity commodity is a provincial story. The UCP removed the 6.8¢ cap in 2019 and the regulated rate ran from about $47/MWh in 2020 to $107/MWh in 2021. Not city hall's doing, and Janz is right to point provincewide. Our argument stays on what council controls: the EPCOR dividend, the franchise fee, water and drainage filings, property tax. See energy and water.

Some EPCOR dividend years are estimates. The endpoints are solid — $166M in 2018, $206M in 2026, from EPCOR's own releases — but 2015, 2016, 2019 and 2020 are interpolated and marked EST.

Housing is the least damning part. They met the target: 2,700 units sought, 2,922 committed. Janz genuinely championed zoning and density reform. The critique is the ceiling — a target sized to close under 6% of the need — not that nothing was built. See community and frontline.

The community-fridge point is an absence, not a shutdown. No city order to remove a fridge is documented, and none is asserted. The charge is that the city funds this work at $0. See food.

Transit figures lean on secondary sources. The $13M shortfall and ~22% fare hike rest on secondary reporting (CBC returned a 403; corroborated via Taproot and Daily Hive) and are marked medium-confidence.

Knack's counter-claims get answered, not ignored. Crime is the lowest in a decade — partly true on homicides, and conceded. Edmonton is among the most affordable major cities — a claim that dies on the food-bank lines and the rent numbers here. We take his best arguments and lose them for him.

Quotes from video are auto-caption sourced. Several come from YouTube auto-captions with timestamps. Wording wants a human check against the audio before any is treated as verbatim.

Police violence is correlation and priority, not a trigger the budget pulled. The body count makes no causal claim. There is no clean per-year series; the anchors are 2022 (a record, ~30% of Canada's fatal police shootings), 2024 (nine shootings, four fatal), and thirty-nine deaths in EPS encounters since 2000. Three figures must not be conflated: the 221 "shootings" EPS reported in 2023 are citywide gun violence, not officer-involved; ASIRT charged-officer totals are Alberta-wide, not EPS-only; and the racial over-representation among the dead is qualitative in the sources, not a hard EPS percentage. The 91% mental-health-and-addictions figure is the solid one. EPS lost two officers to gun violence in March 2023 — a real loss, and a separate category.

On the cold, we count amputations cleanly, not deaths. The link between sweeps and frostbite amputations is correlation supported by clinical testimony, not proven cause. The two amputation counts — the fiscal Edmonton-zone series (91/71/110/113) and a separate calendar count — use different definitions and do not stack. Part of the rise is better detection after FOIs, though the steep shape and majority-unhoused share survive that. Calgary's lower numbers are over-determined by its use of iloprost, so the argument rests on Edmonton's own curve. On deaths: the rise from 37 (2019) to 302 (2023) is HIGH confidence, but 302 is all homeless deaths, mostly drug poisoning, with exposure an unquantified subset. We never write "302 froze."

We never call Janz anti-union. He is union-endorsed three straight elections, opposed the UCP's 2020 school support-staff layoffs, and votes the pro-labour line. The labour exhibit is the wage outcome under his council, not his loyalty. The "Janz signed a letter against the union" claim is unconfirmed and not asserted; the documented episode is the inverse — the Edmonton Police Association's 24 complaints against him (Jan 2022), all dismissed. The wage restraint (CSU 52's 0/1.25/2/3, the firefighters' six years to arbitration) was a closed-session employer decision with no attributable roll-call. We invent no vote. Knack's P3 openness is for new capital only.

On Palestine, we never brand Knack a sympathizer of the Israeli state or complicit in genocide. The charge circulating about the mayor is unsourced, and the single on-record data point contradicts it: in March 2026 he publicly rebuked his own police chief for joining a police delegation to Israel. We publish only the sourceable version on the closing page — establishment silence (no ceasefire motion in any of the 8,412 recorded 2021–2025 motions) and structural complicity through the police he funds. Silence is council-wide, and silence is not sympathy for a genocide. We name it as silence only.

All criticism here is of a state and its policies, never of ethnicity or religion. "Zionism" is treated strictly as a political ideology about a state. The University of Alberta encampment clearing (May 2024) was the university president's decision, executed by EPS; Knack's link is structural, not a personal vote, and is attributed that way. The injuries remain a documented dispute (police said none serious; faculty and protesters reported four hurt, one hospitalized). We do not conflate the January 2024 homeless-encampment criticism with the May 2024 Palestine encampment. The Treaty 6 throughline is sourced to the movement's own Statement of Treaty Principles and labelled as argument.

"Come down on these institutions" means elections and organizing, not violence. The closing argument is about political power: confrontation, organizing, material solidarity, and defeating the complicit at the ballot box. It never means force against persons, and the page says so.

What this method cannot tell you

Two limits. The police escalator's ~5%/year rate is not derivable from the votes — the roll-calls confirm the formula's adoption (C604A, C604B), but the percentage belongs to the Financial and Corporate Services reports. And the encampment sweeps left almost no roll-call to pull, because they were administrative actions; that pillar rests on Administration records, the court file, and reporting. Its evidentiary base is different in kind from the budget votes, and saying so is part of the honesty.

The council scorecard, and its honest edges

The whole-council scorecard is computed in code from the on-disk ekzq-b4qg JSON, cross-checked against huhs-auh6. Its central number, the dissent index, counts one thing: across nine named carceral and order motions, how often a member voted the restraint way. The nine are the police levy cut, the $407M floor, transit-enforcement bylaw 19983, the Healthy Streets EPS top-up, formulas C604A and C604B, and shelter clauses 20657/20842/21193. Each was located by meeting date, agenda number, and motion text, then all thirteen member columns were read and normalized.

Four edges travel with that number. The council was not a monolith — a real four-to-five-member dissent cluster orbited Janz; the claim is only that it never assembled a majority. The 2025 election was not a wipeout — progressives held their seats and independents kept a supermajority; what died is the ceiling. Knack brands "independent," not "progressive" — a strategist called him the only progressive candidate and he let it ride, but the label is not put in his mouth. And absences count as non-dissent, the conservative choice, which changes no ranking. The 2023 operating budget is excluded by design: it is the counter-exhibit where the pattern inverts.

Byline: Independent research. No author is named anywhere, by design. To check the work rather than trust it, the sources index maps every claim to its clipping, and every clipping to its original URL.