Every setting of the Calendar Heatmap

What this visual is really for

Most business metrics live on a daily heartbeat — orders, support tickets, website sessions, machine downtime, hospital admissions, gym check-ins. A line chart shows you the trend of that heartbeat, but it hides the rhythm: which weekdays carry the load, whether weekends matter, where the dead weeks are, how December really compares to July.

A calendar heatmap lays a whole year on the table at once, one colored square per day. Patterns you would need five slicers to find in a table simply appear: the Monday surge, the summer lull, the campaign spike, the outage gap. That's the promise of this visual — and every setting below exists to sharpen one of those stories.

The two layouts: a year-at-a-glance vs. a monthly review

Continuous is the “GitHub contributions” view: weeks flow left to right in one long band per year. Use it when the question is about momentum and seasonality — “how did the year develop?”, “are we busier than last quarter?”, “where are the streaks and the gaps?”. Time reads like a sentence, left to right.

Month blocks chops the same days into twelve little calendars, the way people actually plan their lives. Use it when the question is month-shaped — “how did March go?”, “compare the first week of each month”, “was the month-end crunch worse in Q3?”. Nobody schedules anything by ISO week 37; everyone knows what “mid-March” means.

Nothing about your data changes when you switch — only the arrangement. Colors, badges, legend and every other setting apply identically on top of either layout.

Week start: a small toggle with a cultural opinion

Should the week begin on Sunday or Monday? US audiences and classic wall calendars say Sunday; European businesses, retail ops and ISO standards say Monday. It sounds cosmetic, but it changes what “the weekend” looks like: with Monday first, Saturday and Sunday sit together at the bottom of each column as one visual block — which is exactly what you want when weekend behavior is the story (restaurants, gyms, streaming).

Aggregating duplicate days: what does “Tuesday” even mean?

Real datasets rarely arrive as one tidy row per day. A sales table has one row per transaction; a ticketing system one row per ticket. When forty rows share the same date, the visual has to collapse them into a single square — and how it collapses them changes the question the calendar answers:

• Sum — “how much business did we do that day?” The default, right for revenue and volume.
• Average — “what was a typical deal that day?” Right for rates, scores, response times.
• Min / Max — “what was the worst / best moment that day?” Right for SLAs and peaks.
• Count — “how many things happened?” Ignores the values entirely; pure activity.

A day that looks dark under Sum can look pale under Average — lots of small transactions. That's not a bug; it's two different business questions about the same Tuesday.

Fiscal years: for companies whose year doesn't start in January

Plenty of organizations live April-to-March or July-to-June. For them, a January-cut calendar quietly lies: “this year vs last year” compares the wrong months, and the year-end crunch appears mid-band instead of at the edge where a finance audience expects it.

Fiscal year starts tells the visual where the business year begins. On its own it only fixes the math behind year-over-year comparisons (used by the insights). Turn on Fiscal year layout and the calendar itself re-bands: each row of the continuous layout becomes one fiscal year, labeled “FY 2026” — the year the period ends in, matching how finance teams talk.

Small multiples: comparing teams without lying to anyone

Drop a category field — region, team, product line — into Split by and the calendar becomes a grid of small calendars, one per group. This is where a subtle but important honesty question appears: Shared color scale.

Shared ON answers “who is actually bigger?” One color scale covers everyone, so a dark square means the same thing in every panel. The cost: a small region looks washed out next to a big one — because it is smaller. Shared OFF answers “what is each team's own rhythm?” Every panel gets the full color range fit to its own numbers, so patterns pop everywhere — but you can no longer compare colors across panels. Neither is wrong; they answer different questions, and mixing them up is how dashboards mislead.

Color: the biggest lever in the whole visual

Color is the data in a heatmap, so this card deserves the most care — and the most testing.

Palette type — what story is the color telling?

• Ramp — six curated multi-stop palettes, each tuned separately for light and dark reports. The safe, pretty default.
• Mono — one brand hue fading into the background. Quietest option; great when the heatmap supports a page rather than starring on it.
• Duotone — your own two colors, low to high. Brand control without designing a full ramp.
• Split (diverging) — three colors: below / middle / above. The right choice when the data has a meaningful midpoint (variance vs plan, temperature vs normal, sentiment).
• Report theme — borrows the report's own theme color, so the visual matches whatever report it lands in without configuration.

Color scaling — the honesty dial

Imagine one Black-Friday day that's 40× a normal day. On a Linear scale — an honest ruler — that one day is dark and the other 364 look nearly identical. Truthful about magnitude, useless for patterns. Quantile grades on a curve instead: colors show each day's rank, so the calendar stays richly textured no matter how wild the outliers — that's why it's the default. Log is the compromise for data spanning orders of magnitude, where 10 → 100 should look like the same step as 100 → 1,000.

Buckets — gradient elegance vs. executive readability

A continuous gradient shows the finest differences, but nobody can tell “62%” from “68%” by eye. Bucketing into 3, 5 or 7 shades turns the calendar into named classes — quiet / normal / busy — that a leadership audience reads instantly and consistently. Rule of thumb: gradients for exploration, buckets for reporting.

The legend: small strip, big trust

The legend is what lets a viewer trust the colors. “Less → More” with the actual swatches (or the actual gradient) is the contract between the visual and its audience. The no-data swatch matters more than it looks: in sparse datasets it's the difference between “that day was quiet” and “we have no record of that day” — two very different business statements. Custom low/high labels let the strip speak the data's language: “Fewer tickets → More tickets” beats “Less → More”.

KPI header & insights: the visual that reads itself

The KPI header turns the tile into a self-contained report block: a title plus stat chips (total, daily average, peak) so a viewer gets the headline number before decoding any color. The accent rule under the title is the brand signature — recolor or hide it per report.

Insights go a step further: one-line, auto-computed takeaways — peak day, strongest weekday, month-over-month direction. The subtle setting here is polarity: is a higher number good news? For sales, yes — an upward trend colors positively. For support tickets, complaints or downtime, more is bad — flip polarity and the same upward trend colors as a warning. And Neutral exists because some metrics genuinely have no good or bad direction, and coloring them would editorialize.

Day badges & rules: turning a picture into a monitor

Everything so far shows patterns. Badges mark events. Mark peak day stamps your emoji on the single best day — the “what was our record?” question answered permanently. The three rule slots are the operational layer: each rule tests every day against a condition and marks the hits with a colored ring, an optional emoji, and — critically — an optional hatch pattern.

Business examples that map straight onto rules: “ring every day we breached 200 tickets” (Value ≥ 200), “flag days we missed plan” (Value − Target < 0), “highlight the healthy band” (between 80 and 120). The hatch pattern is the difference between a pretty chart and a compliant one: rings and colors vanish in grayscale print and for colorblind readers; the hatch survives both.

Annotations: institutional memory, pinned to a date

Six months from now, someone will look at the March spike and ask “what happened there?” Annotations are the answer, written down at the moment anyone still remembers: a note pinned to a specific day — “Campaign launch”, “Price change”, “Outage” — stored in the report itself, not in the data. Markers can be numbered chips, dots, or an emoji; each note can show as just a marker, floating text, or text with a pointer arrow.

Summary table: the escape hatch for numbers people

Some stakeholders will never trust a color. The summary table is the same data as an honest per-month (or per-group) table — totals, daily averages, peaks — one click away via a floating toggle in the corner. It's also an accessibility feature: a screen-reader-friendly, precise alternative to color perception. The reader's flip is session-local; it never rewrites the report author's choice.

Day detail panel: click a square, get the receipt

Hovering gives a glance; clicking opens the receipt: full date, exact value, target comparison, how many rows were aggregated — and with split data, which group contributed most to that day. Position auto docks it right on wide tiles and bottom on tall ones, so it covers as little of the calendar as possible.

Accessibility: not a checkbox, a parallel channel

Roughly 1 in 12 men can't reliably distinguish the hues most heatmaps rely on. Pattern on threshold gives high values a texture channel alongside the color channel — the calendar stays readable in grayscale print, on bad projectors, and for colorblind readers. The focus ring makes keyboard navigation visible. These aren't optional polish: Microsoft certification reviews them, and the defaults (Okabe-Ito safe colors for the rule slots, a colorblind-safe ramp in the palette list) mean the visual is presentable to any audience without extra work.

Testing-day cheat sheet

The one-table version — what a correct visual does when you flip each thing:

FAQ — questions real users ask

Click a question to expand it.