Why Knessetil activity metrics are measurements, not a quality score
A product decision behind Knessetil’s civic interface: show discrete activity counts, separate source data from calculated metrics, and never collapse several measurements into one quality score.
The temptation
Once you have open Knesset data, the easy product move is to collapse several metrics into one leaderboard: one composite number per MK, sorted descending, framed as who “works harder.” It looks decisive. It also quietly invents a judgment the source never made.
Knessetil pulls from the Knesset’s public OData API and surfaces bills, queries, agenda initiatives, tenure, and faction averages. Sorting and statistical ranking by a single named metric is intentional. The risk was not missing a metric — it was turning a bundle of measurements into one quality verdict.
The decision
Treat every activity figure as a specific measurement, not a quality score.
On the home surface that means comparisons are labeled as counts by Knesset term — “it is not a quality score.” On About, the same rule is explicit: no manual good/bad grades; show activity as it appears in official protocols and databases. Methodology goes further: raw fields stay raw; calculated fields (cumulative days in office, bill-initiator counts where IsInitiator is true, query totals, agenda totals, faction averages) are named as processing steps, with sync cadence and missing-data behavior documented.
Votes are still withheld until person-to-vote ID matching is verified — and even then, a vote-record count is not attendance. Missing source data shows as missing, not guessed.
What it changes in the product
- No single weighted / composite quality score. Sorting by bills or queries is allowed; collapsing those into “best MK” in one number is not.
- Source vs calculated stays visible. Visitors can tell what the Knesset published versus what Knessetil derived.
- Independence stays legible. The site is a civic interface, not an official Knesset product, and metrics stay informational rather than electoral advice.
What we learned
Civic data products fail first on framing. The hard part was not the sync — four times daily from ParliamentInfo — but refusing the composite scoreboard UX that would make the same numbers feel authoritative in the wrong way. A count without its unit and limits is already halfway to a quality claim.
What’s next
Keep methodology and correction paths as first-class surfaces as the 26th-election context evolves. Any new metric has to earn the same split: named measurement, documented calculation, no implied quality grade.
Follow on Google Search
Add Bersaglio as a preferred source to prioritize our notes, research, and product insights in Search and AI Overviews.