2026 · Lead Designer · Case Study No. 3

Sentiment Monitor Experience

B2BData heavyMobile & WebFintech

Fitting three data stories into one chart — without losing the reader.

+27%
More users served
+34%
Better served
three axes. one chart. pray for me ✎
The Sentiment Monitor experience — sentiment signal, score, subjectivity and confidence, the sentiment chart, and peer comparison.
Where it started

The original Sentiment History was a plain chart with exactly one job: show how sentiment around a stock was growing. Useful, but shallow — and the analyst team requested something far more powerful: a MACD-style sentiment chart.

That request tripled the data. The new chart had to carry:

  1. Sentiment Trend — the MACD line itself.
  2. Price — the market line to read sentiment against.
  3. Sentiment Score — a histogram of the difference between the lines.

Three series. Three value axes, plus time. One chart that mobile and desktop users both need to read at a glance. This is an information-design problem before it's a UI problem.

The original Sentiment History — a single plain line chart of sentiment over time.
The layout decision

Cramming three axes into one plot was never going to read. So the chart became a stacked composition:

  • Upper chart: Sentiment Trend and Price together — trend anchored to the left Y-axis, price to the right. The two lines users compare most live in the same space.
  • Lower chart: the Sentiment Score histogram on its own baseline, colour coded bullish/bearish, out of the lines' way but sharing the same timeline.

One X-axis of time ties the whole story together vertically.

The redesigned Sentiment Monitor chart — an upper plot with Sentiment Trend (left axis) and Price (right axis), a lower Sentiment Score histogram colour-coded bullish/bearish, and a hover crosshair reading all three values on a shared time axis.
The hard part

Hovering with three truths

Traders don't read charts approximately — they want the exact value at the exact moment. The hover interaction had to answer for all three series at once: touch any line, and a vertical crosshair reveals the trend value, the price, and the histogram value at that timestamp. Hover the price line, still see everything. One gesture, the complete picture.

Then the edge cases started biting.

Values colliding with the axes

Hover near the chart's edges and the value tags crash into the Y-axis labels. The fix was side-anchoring: each tag appears on the side of the crosshair that matches its axis — sentiment trend tags to the left of the line, price tags to the right. At the extreme left, the trend tag may overlap its own axis — an accepted imperfection, because the exact value stays legible, which is the entire point. The price tag, living on the right, never collides at all.

Colour alone couldn't carry the mapping

For colour-blind users, "the blue line goes with the left numbers" is meaningless. So the legend spells it out in words: Sentiment Trend (left) and Price (right). Position language instead of colour language — redundant coding that costs nothing and excludes no one.

left goes left, right goes right ✎
Problem → decision → outcome
01

Three series fought for one space

A MACD chart demands trend, price, and histogram — three axes' worth of data.

Decision

Stacked composition — lines above with split left/right axes, histogram below on its own baseline, one shared timeline.

Outcome

Each series stays legible at a glance; the chart reads as one story, not three overlapping ones.

02

Precision reading required three hovers

Users needed exact values, but per-line tooltips meant hunting.

Decision

A single vertical crosshair that surfaces all three values at that timestamp, whichever line you touch.

Outcome

One gesture delivers the complete picture — no hover hunting, no mental cross-referencing.

03

The axis mapping excluded people

Colour-blind users couldn't associate lines with axes, and edge hovers collided with axis labels.

Decision

Legend labels written as "(left)" and "(right)", plus side-anchored value tags that mirror their axis position.

Outcome

The mapping works in words and position, not just colour — and edge collisions resolved everywhere it matters.

Overall: the new chart serves 27% more users, 34% better.

The Sentiment Monitor chart on mobile — the stacked composition holding at small size.
The composition holds at 392px

Dense data doesn't need a simpler chart. It needs a chart where every pixel knows which question it's answering.

Anna ✎