← Back to overview
CLASSLINK · ANALYTICS+ · STAKEHOLDER INPUTPRODUCT DESIGN STANDUP, 2026‑10‑28

Three PMs, three products, one theme: trust and relevance — not visuals

A GitHub-wide search across classlink-dev for customer complaints about charts and dashboards came back empty — that kind of feedback isn't tracked in Issues or PRs (real tickets live in Linear, and this feedback apparently never got ticketed at all). So we asked PMs directly in standup instead. Their answers reframe the reskin's problem statement.

01The feedback, verbatim
A
Analytics PM
Product Design standup, 2026-10-28

"From an Analytics [perspective] the pain has never been directly about the data viz side, it is more simplifying what we present on the pages so people have confidence in the numbers."

→Reads as: a data-trust / simplification problem, not a chart-visualization problem.
RS
Roster Server / OneData PM
Product Design standup, 2026-10-28

"We don't use a lot of graphs in RS. But the general consensus I hear is that the dashboard doesn't convey much use and they instantly hop to other task specific sections. Not to the point that they put in tickets about it though."

→Reads as: a perceived-value / relevance problem — the dashboard gets bypassed, below the threshold anyone files a ticket.
OS
PM, OneSync / Analytics / Roster Server
Product Design standup, 2026-10-28

"Here are the guiding principles we use in Analytics, drawing from Stephen Few's work, supported by work in the Big Book of Dashboards and Storytelling with Data: visualization is, at its core, about encoding aggregation, such as frequency, in order to gain insight."

→Reads as: a concrete, citable design standard — not taste — covering which visual attributes and color functions to reach for. Full framework below.
02What this changes
  1. The throughline isn't visual polish, it's trust and relevance. Two independent PMs, two different products, landed on the same non-aesthetic complaint — confirms the "charts show a number with no context" framing already in the measurement strategy rather than replacing it.
  2. None of this ever became a ticket. That's exactly why the GitHub/Linear search came up empty, and why going straight to PMs was the right move — this class of feedback has to be asked for, not searched for.
  3. Roster Server's problem may not be a dataviz problem at all. "Doesn't convey much use, they hop to task-specific sections" sounds like an information-architecture or relevance issue — worth a scoping conversation before assuming a visual reskin fixes it. Neither of our two in-flight pages is an RS page, so it doesn't block tomorrow, but it's a flag for what comes after.
  4. We now have a named standard to design against instead of a designer's-eye judgment call: preattentive attributes and five color functions (below), supplied by the team that already owns the Analytics viz principles.
03The framework: preattentive attributes & color functions

Stephen Few / The Big Book of Dashboards / Storytelling with Data, as relayed by the OneSync/Analytics/RS PM. Orientation and length were specifically called out as attributes to prefer.

Orientationuse this
Lengthuse this
Width
Size
Shape
Curvature
Added marks
Enclosure
Color value
Color hue
Position
Spatial grouping

Color should carry one of five jobs — never more than one at a time on the same mark:

1 · Sequential
Low → high, one hue ramp.
2 · Diverging
Two directions from a midpoint.
3 · Categorical
Distinct, unordered groups.
4 · Highlight
Stand out — not alert or alarm.
5 · Alarming
Something is actually wrong.
04Where this lands in the two measurement plans

Both plans' step 02 (finalist sketches) and step 05 (color evidence) now cite this framework directly.

Schools → Comparison →
Heatmap finalists scored against "position/length over color"; KPI row against "one form per measure."
Screen Time → Overview →
Hero/measures finalists scored against the same framework, plus the Analytics PM's "confidence in the numbers" framing.