A single large number at the top of a dashboard - a KPI tile, a stat card, whatever the specific tool calls it - carries a disproportionate amount of communicative weight for something so visually simple. It's usually the first thing a reader's eye lands on, and for a genuinely busy reader skimming past everything else, it's sometimes the only thing that actually gets absorbed before they move on. That outsized influence makes the choice of which number earns that spot more consequential than it usually gets treated - more often decided by habit (whichever number the survey tool defaults to averaging) than by a deliberate judgment about what actually deserves that much visual authority.
Table of Contents¶
- What a KPI Tile Is Actually For
- The Aggregation Choice Matters More Than It Looks
- One Number, Not Five Competing Ones
- Give the Number Context, Not Just Precision
- A Worked Example
- FAQ
What a KPI Tile Is Actually For¶
A KPI tile's job is answering one specific question as fast as possible: "is this okay." It's not built to carry nuance, a breakdown, or a "why" - that's what the rest of the dashboard is for. A tile that tries to do more than that single job, by cramming in a secondary figure or an overly qualified label, undermines the one thing it's actually good at, which is delivering a single, unambiguous read in under a second. Choosing what goes in a KPI tile is really choosing which single question, out of everything a dashboard could answer, is important enough to answer before a reader has processed anything else on the page.
The Aggregation Choice Matters More Than It Looks¶
A KPI tile is typically built from an underlying aggregation - an average, a sum, a count, a minimum, or a maximum - and which one gets chosen changes what the number actually means in a way that's easy to gloss over. An average satisfaction score and a count of total responses answer completely different questions, and while that distinction is obvious stated plainly, it's less obvious at a glance when both are rendered as the same large, bold number in the same tile format. It's worth explicitly confirming, for any KPI tile, that the aggregation matches the actual question the tile is meant to answer - a sum is the right choice for "how many total responses have we collected," and the wrong choice entirely for "how satisfied are customers on average," even though both would technically render as a single confident-looking number.
One Number, Not Five Competing Ones¶
The most common mistake in KPI tile design isn't choosing the wrong number - it's choosing too many of them with equal visual weight, so that nothing on the dashboard actually reads as the headline. A row of six equally-sized tiles, each showing a different metric in the same font size and color, asks a reader to do the work of deciding which one matters most - work the dashboard should have already done on their behalf. Genuinely picking one or two numbers as the actual headline, and treating the rest as secondary, supporting figures (smaller, positioned lower, or grouped separately) forces the deliberate judgment call about what matters most, rather than quietly outsourcing that decision to whoever happens to be reading the dashboard that day.
Give the Number Context, Not Just Precision¶
A bare number - "74%" - states a fact without saying whether that fact is good, and a reader has to already know the relevant benchmark or prior value to interpret it correctly on their own. Pairing a KPI tile with a small amount of comparative context - a change indicator against the last period, a small trend sparkline, or a one-line note on what "good" looks like for this specific metric - turns a fact into a judgment the tile itself has already made most of the way, rather than leaving that interpretation entirely up to the reader. This is a small addition in visual space and a large addition in actual usefulness, since "74%, up 3 points from last quarter" tells a complete, self-contained story that "74%" alone simply doesn't.
A Worked Example¶
A customer success team's first dashboard draft opens with six KPI tiles in a single row: average satisfaction, total responses, response rate, average resolution time, NPS, and CSAT - all the same size, all the same font weight, none of them contextualized against a prior period. In a five-minute stakeholder review, half the meeting is spent on someone asking which of the six actually matters most this quarter. The revised version leads with a single, larger KPI tile - NPS, chosen because it's the metric the team's current initiative is specifically trying to move - paired with a small "up 4 points since last quarter" note beneath it, with the remaining five metrics moved into a smaller, secondary row underneath, still visible but visually subordinate. The next review opens with everyone immediately aligned on the one number the meeting is actually about, with the supporting metrics available for anyone who wants to dig further, rather than competing for the same first impression.
FAQ¶
How many KPI tiles should a dashboard have?
One or two genuinely headline-level tiles is a reasonable target for the top of a dashboard; additional supporting metrics are better placed as a visually secondary row rather than competing at the same size and prominence.
Should a KPI tile always show change over time?
It's not strictly necessary, but a comparison point - versus last period, versus a target - almost always makes the number more useful than showing it in isolation, since it turns a fact into an implicit judgment about whether things are improving.
What's the most common aggregation mistake in KPI tiles?
Using an average where a count (or vice versa) actually answers the intended question - it's worth explicitly checking that the chosen aggregation matches the specific question the tile is meant to answer, rather than defaulting to whichever one a tool suggests first.
Should every stakeholder see the same KPI tile as the headline?
Not necessarily - different audiences often care about different top-line numbers, which is a good reason to build audience-specific dashboard versions rather than assuming one universal headline metric works for everyone who might view the report.
For more on structuring the rest of a dashboard around a strong headline number, see Writing Survey Reports People Act On and Should You Show Statistical Uncertainty on a Dashboard?.