Stat Bands: Key Numbers With Sources and Periods

A stat band is a compact row of two to four headline figures, each paired with a label, source, and reporting period. Its job is to expose the few numbers that frame the page—not to compress an entire dataset into decorative tiles.

The values above are an illustrative support-team example, not claims about AmICited or a real company. The source and period remain visible because an example should model the production contract rather than teach readers to add evidence later.

Why this element matters

Readers notice large numbers before they read surrounding paragraphs. A good stat band uses that attention to establish scale, change, or outcome in seconds. Three strong figures create a simple mental model: response speed, service reliability, and customer reaction. Eight weaker figures create a scanning problem. They compete at the same visual weight, force the reader to decide which ones matter, and turn emphasis into noise.

The band works because it reduces selection effort, not because large typography makes a claim more persuasive. Each number must answer a different important question. If two tiles say essentially the same thing—for example, “94% resolved within SLA” and “6% missed SLA”—one of them wastes the limited space. If a figure is impressive but irrelevant to the page’s conclusion, its prominence misleads rather than helps.

Machine extractability is the ability of a crawler, search system, or AI answer engine to retain the relationship between a value and what it measures. “94%” alone is unusable. “94% of support tickets resolved within SLA, Support export, January–June 2026” is a bounded claim: value, metric, source, and time period remain connected. A stat band should expose those relationships as text in document order, not bake them into an image or assemble them only after JavaScript runs.

This is also why the typed element takes precedence over visually similar free text. Under the element writing rules , a passage whose purpose is to present a small set of headline figures must use the stat-band structure. Three bold paragraphs may look similar, but they provide no stable item boundaries, fields, or validation hooks for sources and periods.

When to use it

Use a stat band when two to four figures summarize the page’s central evidence and can be understood without reading a methodology section first. Suitable figures include a measured outcome, baseline-to-current change, market size, cohort count, median, rate, duration, or point-in-time operating fact. The band is especially useful when the figures answer complementary questions rather than repeat one conclusion.

The evidence must exist before the band is designed. A strong selection test is: if a reader copied only these figures, would each remain accurate, appropriately qualified, and representative of the page? If not, keep the number in prose or a detailed data display where its caveat can travel with it.

Near-miss cases are common:

• A row of product benefits: “Faster,” “simpler,” and “smarter” are claims, not statistics. Use benefit copy supported by evidence.
• A KPI dashboard: live operational monitoring needs timestamps, trends, filters, and status logic. A static editorial band cannot replace it.
• A statistical summary: mean, median, range, sample size, and confidence interval usually belong in a table or chart because the relationship among them matters.
• A list of eight numbers: the reader cannot distinguish headline findings from supporting data. Select the three that carry the conclusion and move the rest into the analysis.
• An unverified marketing claim: visual prominence increases the cost of ambiguity. Remove the claim until its source, denominator, and period are known.
• A single number: one figure normally belongs in a sentence, callout, or chart annotation. Use a band only when at least two values form a meaningful set.

Ready to Monitor Your AI Visibility?

Track how AI chatbots mention your brand across ChatGPT, Perplexity, and other platforms.

Get Started Contact

Where to place it

Position is part of the claim. On a research or benchmark page, place the band after the introduction has defined the subject, population, and period, and before the detailed findings. In a case study, place it after the situation and intervention are clear; otherwise readers may attribute an outcome to the wrong starting point or action. In a company profile, place it after identity and scope are established.

A band may appear near the top only when all items are independently understandable and the page immediately supplies methodology or context below. It must not replace the direct answer, research summary, or explanation of what changed. It is a preview of evidence, not the whole argument.

Do not place a stat band directly beside another stat band, a chart showing the same numbers, a price table, a testimonial carousel, or a high-emphasis call to action. Two equally loud modules compete for attention; duplicated figures also make readers wonder which version is current. Keep at least one explanatory paragraph between the band and another dense data display. Do not split the band from its shared methodology or source note with an advertisement, signup form, or unrelated image.

Anatomy

The labelled capture must identify seven functional regions:

1. Value: the observed number, including a sign when change direction matters.
2. Unit: percent, currency, duration, count, score, ratio, or another explicit measure.
3. Metric label: what was measured, including the denominator when ambiguity is possible.
4. Source: the dataset, system, survey, filing, or named publication behind that item.
5. Reporting period: when the underlying activity occurred, or an exact “as at” date for a snapshot.
6. Group context: an optional heading or sentence defining cohort, geography, plan, or scenario once for the band.
7. Item boundary: the semantic container that keeps each value attached to its own label and evidence.

The value should lead visually, but the source and period cannot be hidden behind hover, a tooltip, or an icon. They may use quieter typography, not absent typography.

Stay Updated on AI Visibility Trends

Get the latest insights on AI mentions, brand monitoring, and optimization strategies.

Email address

Subscribe

Design examples

The design system supports four content variants. Each keeps the same semantic fields and evidence rules; only item count and value shape change.

Standard three-item band: the default. Use three distinct figures that together establish the page’s main result. Equal visual weight is appropriate when no single figure is the explicit primary outcome.

Two-item pair: use for a meaningful comparison such as baseline and current, or organization and benchmark. Labels must state the relationship; physical proximity alone must not imply which period or cohort each tile represents.

Four-item maximum: use only when all four figures answer separate decision-relevant questions. Labels need to be shorter because the layout is denser. A fifth item moves the content to a table, chart, or findings section.

Mixed-unit responsive band: percentages, durations, counts, and scores may coexist when their labels are explicit. On narrow screens, tiles stack in document order; the unit stays with its value and the evidence stays with its item.

There is no “unsourced minimal” variant. Removing sources or periods is not a design choice because it changes a bounded measurement into an ambiguous claim.

Parameters

The contract separates the visible value from its meaning and provenance so a renderer can preserve each relationship across platforms.

Context becomes required when the items share a cohort, geography, currency basis, plan, or methodology that cannot fit accurately in each label. A shared source or period may be rendered once only when it applies identically to every item and the markup still associates it with the group. The canonical content model should nevertheless retain source and period on each item so reuse cannot detach a value from its evidence.

Syntax and code examples

All three implementations must preserve item order and the four required item fields: value, label, source, and period. These examples are illustrative content rather than a claim about an organization.

Portable Markdown directive

:::stat-band{heading="Support performance" context="All priority levels"}
- value: "12 min"
  label: "Median first response"
  source: "Support export"
  period: "2026-01-01/2026-06-30"
- value: "94%"
  label: "Tickets resolved within SLA"
  source: "Support export"
  period: "2026-01-01/2026-06-30"
- value: "4.7/5"
  label: "Customer satisfaction"
  source: "Post-resolution survey"
  period: "2026-01-01/2026-06-30"
:::

Hugo shortcode

The existing statgrid helper accepts one pipe-delimited item per line: value, label, and optional source URL. Until a dedicated renderer exposes period and source as separate fields, include both visibly in the label and do not omit either:

{{< statgrid >}}
12 min | Median first response · Support export · Jan–Jun 2026
94% | Tickets resolved within SLA · Support export · Jan–Jun 2026
4.7/5 | Customer satisfaction · Post-resolution survey · Jan–Jun 2026
{{< /statgrid >}}

This mapping is acceptable for display but not ideal for machine validation because three meanings share one label string. A future typed renderer should implement the canonical fields above without changing authored meaning.

WordPress block or shortcode

[stat_band heading="Support performance" context="All priority levels"]
[stat value="12 min" label="Median first response" source="Support export" period="2026-01-01/2026-06-30"]
[stat value="94%" label="Tickets resolved within SLA" source="Support export" period="2026-01-01/2026-06-30"]
[stat value="4.7/5" label="Customer satisfaction" source="Post-resolution survey" period="2026-01-01/2026-06-30"]
[/stat_band]

A custom WordPress block may present the same fields as form controls. It must render them as real text, preserve source links, and keep the reading order sensible when styles are disabled.

Examples

Good: three complementary, bounded figures

12 min — median first response; Support export; January–June 2026
94% — tickets resolved within SLA; Support export; January–June 2026
4.7/5 — customer satisfaction; Post-resolution survey; January–June 2026

This works because each value answers a different operational question, includes its unit, identifies its evidence, and states when the activity occurred. The same period makes comparison easier, while the separate survey source honestly distinguishes sentiment data from ticket-system data.

Bad: impressive numbers without boundaries

12 — Response time
94% — Success rate
4.7 — Customer score
2× — Faster
#1 — Best service

This fails even if the values were copied from a valid internal report. “12” has no unit. “Success rate” has no numerator or denominator. The score has no scale. “2× faster” lacks a baseline, and “#1” lacks a category and comparison set. None names a source or period. Five equal-weight items also conceal which result matters most. The repair is not to add smaller footnotes to all five; it is to select the three decision-relevant measures, restore their definitions, and attach evidence to each.

Schema markup and accessibility

A stat band has no dedicated Schema.org type. It normally remains visible content within an Article, Report, Dataset, Organization, or another valid page-level entity. A research page may map verified fields such as a dataset’s temporal coverage or measured variables into Dataset markup, but the visual presence of a number does not create eligibility. Do not invent a StatBand, Statistic, rating, award, or performance property.

Structured data must never be more specific than the visible claim. If the tile says “4.7/5 customer satisfaction,” markup cannot silently reinterpret it as a product aggregateRating unless the page also supplies the required rating population, method, and eligible subject. The safest default is no element-level schema.

For accessibility, render each item as a coherent text group in DOM order. The value should be read with its label before the source and period. Do not use CSS-generated text for units, because assistive technology may miss it. Do not encode positive or negative performance only through green and red. If an arrow shows direction, include text such as “up 8 percentage points.” Source links need descriptive accessible names, and a repeated “Source” link should expose the actual publication or dataset name.

The band must reflow rather than shrink. At narrow widths, items stack in the same editorial order. Large numerals still require sufficient contrast, but supporting evidence cannot become illegibly small. The container is not an alert, live region, list of buttons, or keyboard target unless it includes genuine interactive controls—which the editorial form should not.

Writing rules

The reader grants a large number more authority than a sentence in ordinary body copy. That makes restraint a content requirement. Use three figures by default, two for a deliberate pair, and four only when each adds a distinct dimension. Never publish five or more as a stat band.

Write the value as compactly as accuracy allows: normally 1–12 characters including its unit. Use numerals, preserve meaningful decimal precision, and keep the unit attached: “12 min,” “94%,” “€2.4m,” or “4.7/5.” Do not round a value merely to make the tile look cleaner if the rounding changes its interpretation. Do not mix percentage change with percentage-point change; write the correct one in the label or qualifier.

Labels should be two to ten words and name the measured thing, not praise it. “Tickets resolved within SLA” is testable. “Outstanding service performance” is promotional fog. State the denominator where a reasonable reader could misunderstand it: “32% of audited URLs,” not “32% compliant.” Define unfamiliar abbreviations before the band or write them out in the label.

Every item needs a source and period because numbers decay at different rates and may come from different systems. Use the period of observation—“January–June 2026”—rather than the publication date. For inventory, employee count, price, or another snapshot, use “as at 30 June 2026.” For a forecast, put “projected” in the label and identify the model or plan as the source.

The band must never contain an unsourced superlative, a testimonial fragment, a button, a long methodology note, a chart, a second nested band, or a number whose qualification reverses its apparent meaning. Do not use animated count-up effects: they delay comprehension, can distract readers, and expose intermediate values that are not claims. Put detailed methods and citations immediately below the band or in the page’s sources block , while retaining a visible source name and period on every tile.

Post types that use it

The postTypes frontmatter is the machine-readable relationship. The table below defines when and where each registered post type uses the element.

QA checklist

• The band contains two to four figures, with three used unless the content justifies another count.
• Every figure is central to the page’s conclusion and adds a distinct dimension.
• Every value includes its unit, scale, sign, or ratio denominator where applicable.
• Every label names a measurable fact rather than a benefit, superlative, or vague “success” metric.
• Every item identifies its source, even when a shared source note also appears below the band.
• Every item states the observation period or an exact point-in-time date.
• Cohort, geography, currency basis, plan, and comparison set are visible wherever needed for interpretation.
• Forecasts and targets are explicitly labelled and cannot be mistaken for observed results.
• The band follows enough context to make the figures understandable and precedes detailed analysis.
• It does not sit beside another high-emphasis data, testimonial, price, or conversion module.
• It does not duplicate a nearby chart or table without adding a distinct summary purpose.
• Values, labels, sources, and periods are real text and remain connected in document order.
• Units are not supplied only through CSS, icons, color, hover, or animation.
• The narrow-screen layout preserves editorial order, readable type, and item boundaries.
• Structured data uses only valid page-level types and claims no rating, award, or property that the evidence does not support.
• Portable Markdown, Hugo, and WordPress versions preserve the same values and provenance.
• All screenshot comments remain non-rendering capture instructions until the named assets exist.

FAQ

How many figures should a stat band contain?

Use three by default. Two work when they form a meaningful pair, and four are the maximum. A fifth figure is evidence that the content needs prioritization or a different format.

Does every figure need its own source?

Yes. A shared visible note may cover all items when they use one dataset, but the content model should still associate that source with each value so extraction or reuse cannot separate them.

What counts as a reporting period?

Use the interval in which the measured activity occurred, such as “January–June 2026” or “Q2 2026.” For a snapshot, write an exact “as at” date. A page publication date only tells readers when the article appeared.

Can the band contain projections or targets?

Yes, when the label explicitly says “projected,” “forecast,” or “target” and the source identifies the model, plan, or accountable owner. Keep observed results and future estimates in separate bands or label the distinction on every item.

Does a stat band need Schema.org markup?

Usually not. Keep it within the enclosing page schema. Map values to structured properties only when the subject, property, methodology, and evidence independently qualify; visual emphasis alone creates no schema entitlement.