New brief — where your team's judgment goes, and how it compounds into decision memory that stays yours.Read the brief
A2go · Visual Identity · Reference · 2026

A visual identity for A2go, and the reasoning behind it.

A single reference for how A2go looks and behaves across its surfaces, with the reasoning behind each choice.

The aim of the brand is modest to state and hard to earn: to let a buyer recognize the judgment we sell before a word is exchanged. A2go · Visual Identity · Introduction
2026-07-23T07:59:48.696615 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ ORDER ATP coverage 96.2% NOMINAL SUPPLY supplier OTD 87.0% WATCH INVENTORY days of supply 41 d NOMINAL LOGISTICS tender acceptance 61% WATCH DEMAND rolling MAPE 31% BREACH EXTERNAL port dwell 3.1 d NOMINAL SIX SIGNAL FAMILIES · TRAILING 24 H ILLUSTRATIVE DATA · A2GO CHARTING STANDARD · MAPE BAND ≤ 22% PER THE PUBLISHED CASELET

Introduction

What this document is, and why it exists.

This document gathers the visual direction for A2go and the reasoning behind it. It covers how the brand looks and behaves across its surfaces, the color and type that carry it, the reference work that sets the standard, and the data-visualization practices that govern every chart we publish. Each recommendation is made with its reasoning attached, so the thinking is available and not only the conclusion.

The reason to take it seriously is that the brand is becoming the first place the market meets our judgment, and for a long stretch it is often the only place. A large share of enterprise software research now begins inside AI models, and a great deal more happens in quiet reading before anyone speaks with us, which means the website and the wider digital footprint now carry work that a salesperson once carried. A look that could belong to any enterprise vendor asks the market to take our differentiation on trust. A look that reads like the instrument we actually build allows that differentiation to show on its own, before a conversation begins.

It rests the brand on principle rather than taste, so that anyone building against it inherits the reasoning and not only the rules. Where a recommendation is made, the case for it sits alongside, including the options that were weighed and set aside, so the logic can be followed and checked.

§0 · The canonical sheet

Tokens at a glance.

Everything needed to put a document, deck, or page in A2go's identity, stated as values. These mirror the site's production token layer (tokens.css) and are the decision of record.

If anything disagrees, this section wins. The essays below carry the reasoning, and some rendered examples predate the final palette. When applying the A2go identity — by hand or with an AI assistant — use exactly the values on this sheet.

Naming & claims

RuleValue
CompanyA2go — domain a2go.ai. Always a lowercase g ("A2go", never "A2Go"). The agents2go.ai domain is the internal working site — don't cite it in produced material.
ProductA2go Decision Intelligence Platform, abbreviated ADIP after first use. Never "the A2go platform" as a name.
ArtemisThe name of the AI assistant persona only (e.g. the Ask Artemis widget). Never used as a product or company name.
SCORReference SCOR as ASCM's framework and link outbound to ASCM. Plain language before acronyms in front-page copy. Founder credential is stated as contributed to SCOR as a practitioner at PRTM — never "authored," "co-authored," or "founded."

Type — one super-family, three voices

RoleFaceUsed for
DisplayIBM Plex Serif · 500–600Headlines and section titles. Tight leading (~1.1), slight negative tracking.
BodyIBM Plex Sans · 400–600Running text, UI copy, forms. Fallback: system-ui.
Data / labelsIBM Plex Mono · 400–600Eyebrows, metadata, figures, code. Uppercase with wide tracking (.06–.14em) for labels.

Color — light surfaces

TokenHexRole
Paper#FFFFFFPage background.
Paper-2#F5F5F7Subtle bands and raised cards on light.
Ink#1D1D1FPrimary text.
Ink-2#3A3A3CSecondary text.
Muted#6E6E73Tertiary text, metadata, eyebrows.
Rule#D2D2D7Dividers and borders.

Color — the Signal teal accent

TokenHexRole
Signal#00707AThe one accent: links, CTAs, focal data point. 5.8:1 on white (AA). One focal use per view.
Text on Signal#FFFFFFButtons and fills carry white text (5.8:1, AA).
Signal hover#005860Hover / pressed state of Signal fills.
Signal tint#E0F0F1Accent tint background (selected cards, highlights).
Signal text#06545BAccent-colored text on tint or paper (8.6:1 on white, AAA).
On darkDeep teal recedes on navy (2.7:1) — on dark surfaces use Target #2CB6C0 for bright strokes/markers, or a Signal fill carrying white text.

Color — dark instrument surfaces

TokenHexRole
Navy#06243FHero, footer, instrument panels.
Navy raised#0A2F4FRaised cards within dark.
Grid#15579EGridlines and hairline borders within dark.
On-dark#FFFFFF · #BCD4E6 · #8A99A8Primary · secondary · tertiary text on navy.
Abyss#041627Marketing console ground (ratified 2026-07-15) — marketing dark and simulated-product staging only; product dark surfaces stay Navy.
Phosphor#7CE2CFLit numerals and the DECIDED stamp on Abyss (ratified 2026-07-15). Never body text, never on light surfaces, never status.

Color — the decision loop & status

TokenHexRole
Sense#1F77C4Signals in — data-in series, the Sense step.
Decide#0A4A92Reasoning — secondary series, the Decide step.
Target#2CB6C0Healthy / in-target status.
Watch#E8A13AAt-risk / watch ONLY — kept scarce, like Breach (adopted 2026-07-05, closing the "next conversation" from the status essay). 7.2:1 on navy; on light surfaces use #A66A00 for text/markers (4.5:1 on white). Amber family — never collides with Signal, Target, the data blues, or Breach.
Breach#E5484DCritical and breach only. Kept scarce so it always means something.

Voice, in one row

RuleValue
RegisterOperator to operator. Precision carries the authority; adjectives don't. Claims are specific and checkable, prices and figures are stated plainly, and hedged marketing filler ("world-class," "cutting-edge") is not used.

Surface registers — ratified 2026-07-15

SurfaceRegister
Marketing dark — site hero, deck dark slides, boothConsole: Abyss ground + dk-grid graph-paper texture + Phosphor numerals + the DECIDED stamp in Phosphor.
Marketing light — papers, memos, reading pagesLedger: this sheet unchanged, plus vocabulary — masthead double rules; ledger tables (mono-caps header under a double ink rule, hairline rows, serif amounts in Signal-text); the stamp in Signal, once per piece.
Product surfaces, realUntouched. The muted instrument law governs; nothing in the registers applies.
Product surfaces, simulated — demos, screenshots, hero mocksConsole — so the product in the ad and the product in use stay one instrument.
The stamp, everywhereRing + mono caps. Phosphor on dark, Signal on light. One per view. Never watch/breach; never encodes status — it marks a concluded decision.

§0b · Working with the identity

Putting the identity to work — by hand, by assistant, by template.

This page is written to be used, not only read. Anyone producing material in A2go's name — on staff or an external partner — can work from it three ways, and all three end at the same place: the §0 values, exactly.

With an AI assistant

Give your assistant the machine edition of this document — a2go.ai/visual-identity.md — and ask for what you need. It carries the canonical tokens, the naming and claim rules, the voice, and the chart standards as plain text that any current assistant (Claude, ChatGPT, Copilot, Gemini) can fetch by URL, or accept as a paste or attachment when it can't browse. The standard instruction — also on this page's cover card, with a copy button:

Fetch https://a2go.ai/visual-identity.md — the machine edition of the A2go visual identity (https://a2go.ai/visual-identity) — and apply it to everything you produce in this conversation. Hold exactly to its §0 tokens for color and type, its naming and claim rules, its voice, and its nine chart standards. For Word, PowerPoint, or Excel documents, start from the official A2go templates linked on that page rather than a blank file. If your output conflicts with the file, the file wins. If you can't fetch URLs, say so and I'll paste the file instead. Now produce: [describe the document you need].

Two habits keep the output honest. For Office documents, have the assistant start from the templates below rather than a blank page — the theme, logo, notices, and disclaimers are already correct there. And read the result against §0 before it ships: an assistant applies the rules; it doesn't replace the editorial pass.

The Office templates

The Word, PowerPoint, and Excel templates carry the identity already — theme colors and type, logo placement, dark- and light-mode sample pages, and the notices set; the Word and PowerPoint templates also carry the signature charts from Part IV as specimens. Install the free IBM Plex fonts first (github.com/IBM/plex), then add the templates to Office's user-templates folder.

TemplateFileCarries
WordA2go-v14.dotxCovers (dark + light), letterhead, memo, policy, one-pager, notices, style specimens, the §0 canon reference, the chart gallery page.
PowerPointA2go-v14.potxDark and light slide pairs, agenda and progress, the Sense · Decide · Act diagrams, the signature pack, notices, and the chart specimens.
ExcelA2go-v14.xltxBranded theme, light and dark worksheets, header/footer furniture, assets sheet.

Every template ships the full notices set — confidentiality, general disclaimer, forward-looking statements, illustrative results, copyright and trademarks — and the Word notices page carries per-block keep-or-delete guidance. Version markers live in filenames only; the set above is the current one, and this page updates when it revs.

Logos

Two marks, chosen by the surface — the rule and the logo blues are the 2025 logo sheet's, reproduced here (the sheet itself rides along as provenance). The files are named by the background they sit on, so the choice makes itself. Place them as delivered: no recoloring, no stretching, no effects — and never a ® with the wordmark.

Use onFileWhich mark
Light or white backgroundsa2go-logo-for-light-backgrounds.svg · pngThe DARK logo — logo blue #1C75BB.
Dark backgroundsa2go-logo-for-dark-backgrounds.svg · pngThe LIGHT logo — logo blue #4E9FD3. On this identity's Navy and Abyss surfaces, always this one.

The sheet also records the mark's own background indigo, #061544 — a logo-sheet color, distinct from §0's Navy. SVG is the format of record (true vector, scales to any size, ratified 2026-07-23); the PNG rides along for tools that can't place SVG, such as Office pastes.

Fonts & tokens

IBM Plex carries all three voices. For Office work, install the desktop fonts from IBM's releases (free, open-licensed) before opening the templates. For web work, the hosted set and the §0 values are one link each — reference them directly or copy them into the project:

ResourceFileCarries
Webfonts stylesheeta2go-fonts.cssThe @font-face set — IBM Plex Sans (variable), Serif, and Mono, latin subset, absolute URLs into this site's /fonts/ (served with CORS open, so it works from any origin).
Design tokensa2go-tokens.css§0 as CSS custom properties (--a2go-*): color, type stacks, type scale, spacing — for hand-built pages and AI-generated ones alike.

Contents

How the document is organized.

Four parts, moving from why the brand matters, to the system itself, to the reference work behind it, to the standard that governs the charts the product puts on screen.

Part I · The case

Why the brand should read like an instrument.

Enterprise buyers rarely grant trust to a claim. They grant it to evidence that judgment is present and working, and they look for that evidence in whatever they can see before a conversation begins. The brand should behave the way the product does. It senses, reasons, and acts, and it shows its work, so that the qualities we describe in words are already visible in the way the brand carries itself.

It helps to name the two broad postures a company in this category can take, because they send very different signals to a reader who is deciding whether to put us on a shortlist. Neither is wrong in the abstract. The question is which one lets a differentiated product be recognized as differentiated.

The conventional posture

  • A look that could belong to almost any enterprise vendor, close to what a model produces when asked for "enterprise software"
  • An accent drawn from the same family as the base, so there is no single point of recall the market carries away
  • Deep hierarchical menus that ask the visitor to learn an org chart before finding what they came for
  • An implicit message of "adopt our platform," which positions the company as one more system to take on
  • Little for a model, an agent, or a developer to reach for

The A2go posture

  • A decision console assembled from the company's own supply-chain telemetry, recognizable as itself
  • One disciplined accent that gives the market a single thing to remember
  • Flat, intent-based entry, so a visitor arrives by their pain, their stack, or by reaching A2go inside an AI model
  • A message of "the layer your stack reaches for," which matches how the product is actually positioned
  • Structure that is legible to machines as well as people, for the surface where much enterprise research now begins

Part II · The system

One identity, living on three surfaces.

A brand for A2go has to do more than dress a single marketing website, because the company meets its audience in three quite different places. The identity is built as three registers that share one ink, one accent, and one type system, and that differ only in density and mode. Naming the registers is what keeps the identity coherent across surfaces that would otherwise drift apart.

Document · light

Editorial and AI-citable

Long-form research, explainers, and case studies, written to be read by people and parsed by models in equal measure. Calibrated instrument-paper, a serif for the longer reading, and a generous measure.

A ten-month buying cycle is won by being remembered, not merely found. Structured, sourced, and quotable.

Console · dark

Live product and demo

The diagnostic, the dynamic demos, the agent surface. Signal traces, status telemetry, and write-back, where decision intelligence stops being a claim and becomes something a visitor can watch happen.

OTIF · jeopardy · 2d

Developer · MCP

Callable layer and API

The MCP server, the API documentation, the authenticated tier. Mono-led, terminal-grade, and precise, for the moment a model or an agent reaches for A2go in the course of its own work.

$ mcp call a2go.ground \
  --metric "OTIF" --domain "deliver"

Each register corresponds to a real product surface: the Layer 1 diagnostic, the Layer 2 earned-access tier behind login, and the Layer 0 callable substrate. Defined this way, the system holds all three without ever looking like three different companies.

II.2 · Reading status by color

Why a status screen needs more than one or two colors.

The need for more than one or two colors is not a question of branding taste; it follows from what the product does. For a logo or a marketing page, one or two colors is often the disciplined choice. But A2go's screens show live operational status, and the whole job of a status screen is to let an operator find the one item about to miss its promise, out of thousands, in about a second. For the eye to do that, on-track has to look different from at-risk, and at-risk different again from breached. That is the function, not decoration, and it is the reason a status product needs a small set of distinct colors, each with a fixed meaning.

One colorcolor says nothing
Find the three breaches. Every square is the same color, so the only way is to check each one.
With one color, the board can tell you nothing at a glance.
Color carrying the statecolor does the work
Same squares, same data. Now each state has its own color.
The three red squares find you, in about a second, without reading anything.
On track At risk Breached One color, used for everything

This board has 126 squares. An operator has thousands.

With one color, the effort to find a problem grows with every row, because each one has to be read. With color carrying the state, the effort stays flat however large the board becomes, because the eye goes straight to what is wrong. That is the whole argument for a small, fixed set of status colors.

The same thing, on the live decision panel

One colorread every row
OrderCustomerPromise inStatus
SO-4471Continental Foods6 daysIn transit
SO-4472Harbor Retail9 daysIn transit
SO-4480Meridian Mfg2 daysIn transit
SO-4488Delphi Group11 daysIn transit
SO-4491Castellan Colate by 1 dayIn transit
SO-4495Orbu Logistics7 daysIn transit
SO-4499Volt Industrial8 daysIn transit
SO-4503Nyx Brands12 daysIn transit
Color carrying the statethe late one pops
OrderCustomerPromise inStatus
SO-4471Continental Foods6 daysOn track
SO-4472Harbor Retail9 daysOn track
SO-4480Meridian Mfg2 daysAt risk
SO-4488Delphi Group11 daysOn track
SO-4491Castellan Colate by 1 dayBreached
SO-4495Orbu Logistics7 daysOn track
SO-4499Volt Industrial8 daysOn track
SO-4503Nyx Brands12 daysOn track

And why the brand color has to stay out of the way

Brand color is also a status colorthey blur
A2go
Location 1 — on track
Location 2 — on track
Location 3 — breached
Is that teal a button or a live status? You cannot tell at a glance.
Brand color kept separateboth read cleanly
A2go
Location 1 — on track
Location 2 — on track
Location 3 — breached
Whatever the accent ends up being, it should be a color the status set isn't using.

The colors here are the universal traffic-light convention, used only to make the point, and are not a proposed brand palette. The argument is functional: a screen that shows status needs a small set of distinct colors, each with a fixed meaning, and the brand's own color has to stay clear of them. How many, and which, is the next conversation. That the answer is more than one or two is settled by what the product has to do.

II.3 · Color

A palette where color carries meaning.

In this palette, color does more than decorate. Each hue maps to a step in the A2go decision loop, and the status colors borrow directly from the language of supply-chain telemetry. A useful test for any palette is whether it would still make sense on a product in another category. This one would not, which is the point: it is specific to what A2go does.

The decision loop · color follows the step

Sense
#1F77C4
signals in: events, demand, supplier status
Decide
#0A4A92
agents reason against the constraints
Act
#00707A
write-back, the accent, the call to action
Judgment
#06243F
the substrate: encoded agent memory

Telemetry status · the product's own language

IN TARGET WATCH / AT-RISK BREACH / CRITICAL

Core tokens

Navy / Console
#06243F
Canvas of the live register — hero bands, footer, instrument panels.
Paper
#FFFFFF
The page. #F5F5F7 for subtle bands and raised cards; #1D1D1F for ink.
Signal Teal
#00707A
The accent. The reasoning is below.
Target
#2CB6C0
Healthy and in-target. A calm teal rather than an acid green.
Sense
#1F77C4
Neutral signal and data-in series. The one bright blue, kept because it carries a job.
Decide
#0A4A92
Reasoning and secondary data series. A restrained deep blue.
Breach
#E5484D
Critical and breach only. Kept scarce so that it always means something.
Grid
#15579E
Instrument hairlines and panel borders. The structure of the console.

The single disciplined accent: Signal Teal.

The decision of record is one accent, a deep signal teal, #00707A — the "Brand 3a" palette, adopted June 2026 as the live identity. An earlier draft of this document argued for a warm amber; that case, and why teal carried the decision, is preserved below so the reasoning stays checkable.

Signal Teal#00707A
Live
Signal Sense Decide Target Breach

The green cast keeps Signal distinct from the Sense and Decide blues, and its depth keeps it distinct from the brighter Target teal, so the accent reads as its own voice rather than another data hue.

HarmonyTeal is the palette's own family concentrated rather than an import from outside it. The identity reads as one temperature — calibrated and cool — with the accent as its most saturated note, which suits an instrument better than a contrast-for-contrast's-sake pop.
MeaningDeep petrol teal is the color of the Act step, so the accent literally marks action: the call to action, the write-back, the focal signal. It reads technical and deliberate rather than promotional.
DistinctnessIt avoids the collisions that disqualify most accents in this palette: it is not the hyperlink blue of the data series, not the green of health, and not the red of breach. Nothing else in view competes for its meaning.
AccessibilityOn white it reaches 5.8:1 (AA) for text and lines, and as a fill it carries white ink at 5.8:1 (AA). The darker text variant #06545B reaches 8.6:1 (AAA) on paper. On navy the deep teal recedes (2.7:1), so dark surfaces hand the bright-stroke job to Target #2CB6C0 at 6.4:1.

How the accent is used

AFills carry white ink

Primary buttons and accent fills use Signal Teal with white text. Dark ink on teal falls below contrast and is not used.

bg #00707A · text #FFFFFF · 5.8:1
/Lines, markers, live states

On dark surfaces the deep teal recedes, so the bright stroke for the signal-trace, the data marker, and the live indicator is Target teal; Signal fills with white ink still carry the call to action.

stroke #2CB6C0 on #06243F · 6.4:1
aAccent text on light

Links and accent text on paper use Signal directly, or the darker variant where longer passages need more headroom.

#00707A · 5.8:1 or #06545B · 8.6:1 on white
·Used sparingly

One accent, applied with restraint, marking the single most important action or signal in view, so that it never loses its meaning.

one focal point per view

Alternatives that were weighed

The strongest was the warm accent this document originally recommended; the others were set aside for reasons of craft rather than preference.

Signal amber #F5A623
The original draft recommendation, overtaken by the decision. A single warm point opposite a cool base is the textbook accent move, and amber carried the instrument connotation well. Teal won on cohesion — one temperature across the whole identity — and on the working preference of the brand review. Amber remains the documented fallback if a warm register is ever wanted.
Antique gold / brass #C99A2E
Navy and gold is among the most trusted pairings in serious enterprise design, with a heavier heritage note. Set aside with amber for the same cohesion reasons.
Acid green / lime
Sits too close to the Target health hue, and reads as a familiar AI and crypto default rather than a considered choice.
Magenta / hot pink
Striking on screen, but it reads as consumer and playful, and pinks tend to shift and muddy in print, which is the wrong register for enterprise trust.
Coral / bright red
Occupies the Breach slot. An accent that also signals a critical state cannot comfortably serve as the everyday call to action.

The decision of record: Signal Teal #00707A anchors the palette. Amber and brass are documented alternatives, not active options.

II.4 · Typography

A recommended type system: one super-family, three voices.

The type system is the IBM Plex super-family: its serif for editorial authority, its sans for product and interface, and its mono for data, code, and the callable register. Drawing all three from one source keeps the registers in tune with one another. It is open-licensed and self-hostable, which suits a no-CDN posture, and it reads as a deliberate system rather than a default.

font-display
IBM Plex Serif
Judgment, made callable.
font-body
IBM Plex Sans
Detect the disruption early, then decide and act before margin leaks out.
font-mono
IBM Plex Mono
a2go.ground("OTIF") → Fulfill · deterministic

Display is set in the serif at medium weight with tight tracking, reserved for the cover and section openers, with body text never falling below 16px. Mono appears wherever the content is a metric, a token, an agent name, or a call, the small typographic signal that the brand is instrument-grade.

II.5 · Components and motif

Patterns drawn from the instrument.

A handful of patterns carry most of the interface, and one recurring device gives the brand something to be remembered by: the signal-trace and the instrument grid, with everything around them kept quiet.

Metric card

18h → 15m

MPS cycle, multi-site manufacturer

Status callout

WATCH

Supplier OTIF drifting, with four days of lead-time slip detected.

Agent tag

FORECAST · INVENTORY · OTIF

Agent-collection tags travel in mono, a quiet mark of credibility.

The discipline that makes this work is restraint. With the signal-trace and the grid doing the recurring work, the system avoids accent stripes, rules beneath headings, and gradient-on-white heroes, so that the one signature stays the one signature.

II.6 · The telemetry language

The product's own data, dressed as brand.

The visuals that carry A2go are its supply-chain telemetry, so the brand defines how that data looks, so that any chart, whether in a demo, a case study, or the live console, is recognizably A2go. Four representative patterns appear here; the full craft standard for charts is set out in Part IV.

Signal trace · time-seriesforecast vs actual
MAPE, OTIF, and lead-time drift, with a target band, an actual line, and the at-risk marker.
Status heatmap · matrixSKU × node
Inventory health and service level by location, with the status triad read as a field of cells.
Flow · sankeysource → transform → fulfill
Material and order flow, with bottleneck width, and channels colored by loop step.
Target · bullet KPIvs. plan
Forecast accuracy · 85% (target 92%)Supplier OTIF · 95% (target 95%)
One KPI against its target and threshold, the CFO-grade number that has to be exact.

II.8 · Voice

Operator to operator.

The voice carries its authority through precision rather than adjectives. It is plain, specific, and a little dry, and it trusts the reader to recognize competence without being told it is there. The examples below calibrate the register.

Closer to the intended voice
Further from it
Detect the slip two days out, weigh the re-promise, and write the change back with its rationale logged.
Revolutionary AI-powered synergy for next-generation supply chains.
A generic model can only approximate, and supply chain needs a number that comes out identical every time.
Leverage cutting-edge solutions to unlock transformative value.
Start with one agent, on the systems you already run, and add the next once it has earned the place.
A holistic, end-to-end platform for digital transformation.

The agent collection is the credibility signal, used with restraint, with the deeper methodology kept for technical conversations rather than the front page.

Part III · The reference standard

Reference work that sets the bar.

The clearest way to show what this direction is aiming at is to point to work that already meets the standard. The sites below come in two groups, those in or near the category and those drawn from other fields, and each is worth opening, since juried, well-crafted work reads better than any description of it. Where an award is claimed it should be confirmed on the gallery itself; the value here is the direction each one demonstrates.

In and near the category

Palantir
Ontology · decision platform
The closest archetype to this direction. Its ontology is made into a brand through cinematic dark surfaces, severe restraint, and very little marketing gloss, and it stands as proof that the opposite of cheerful enterprise-blue can read as the most serious company in the room.
Altana
Supply-chain intelligence · AI-native
The strongest design in the category itself, with a genuinely modern, data-network aesthetic and a restrained palette. The best evidence that a supply-chain company can look entirely current, and worth studying for its data visuals and its measured use of color.
o9 Solutions
SCM platform
Useful for the way its "Digital Brain" gives the brand a concept to draw, beyond a feature list. The caution worth noting is that the graph motif can read as decoration rather than as something alive, which is the gap the A2go signal-trace closes.
Celonis
Process intelligence
A clean execution of the idea that a model can be derived from how the business actually runs, expressed in process-graph visuals that are directly relevant to the orchestration story.
project44
Visibility · control tower
A good study in making real-time legible, with motion and map telemetry. The closest peer to the live Console register.
Across the strongest work the pattern is consistent: one owned concept, a dark or restrained base, and color used sparingly. None of it leads with a generic blue gradient, which is the bar this direction is built to meet.

Adjacent, from other fields

Linear
The console benchmark
The clearest reference for the Console register, dark and precise, with restraint treated as a strength. Widely held as the standard for product-led design craft.
Anthropic
A counter to the blue default
A frontier-AI company that reads as trustworthy precisely by declining the generic palette, with warm paper and a serif-forward, editorial calm. Direct support for the Document register.
Stripe
Density done well · the docs and API register
The reference for making dense, technical information feel effortless, and for treating documentation and the API as designed product surfaces. The model for the Developer register.
Watershed
Data as the hero
A platform that makes charts and numbers the brand, with legible visualization at the center. The closest model for treating telemetry as the product, dressed as brand.
Vercel · Mistral · Ramp
Technical authority · distinctive AI · efficiency
Three more reference points: infrastructure-grade technical confidence, an AI brand with a real point of view rather than the template, and a tight palette that makes efficiency a felt quality.
A single comparison settles much of the conversation about color. Placed side by side, anthropic.com and palantir.com share the same seriousness through opposite palettes, and both are unmistakably themselves. Distinctiveness and trust tend to reinforce one another rather than trade off.

Part IV · A craft standard for charts

Charts that carry the weight of a decision.

Because A2go sells judgment, its charts deserve a publication standard rather than a dashboard one. This part sets out the body of knowledge behind rigorous statistical graphics, the working rules that follow from it, a gallery built to those rules in the A2go palette, and the tooling to produce charts once and render them everywhere. The reader decodes quantities through visual channels, and those channels are not equal: position on a common scale is read most accurately, then length, then angle and area, then color and shade. A careful chart spends its most accurate channels on its most important quantities, shows the uncertainty around them, and leaves out whatever is neither data nor essential reference.

IV.1 · The canon

The body of knowledge worth standing on.

A short reading list that settles most disagreements before they begin, drawn from primary sources rather than summaries of them.

Edward Tufte
The Visual Display of Quantitative Information

The data-ink ratio, small multiples, sparklines, and the lie factor, and the argument for removing anything that does not carry information.

William Cleveland
The Elements of Graphing Data

Graphical perception established by experiment, including the accuracy hierarchy of position, length, angle, and area, and why dot plots read more accurately than bars.

Leland Wilkinson
The Grammar of Graphics

The formal system beneath modern tools such as ggplot2, Vega-Lite, and Observable Plot, and the reason a chart is best thought of as a set of mappings rather than a fixed picture type.

ft.com/vocabulary · the chart-for-question doctrine

More than forty chart types organized by the relationship in the data, covering deviation, correlation, ranking, distribution, change over time, part-to-whole, magnitude, spatial, and flow. The single most useful guide for choosing a form.

Jonathan Schwabish
Better Data Visualizations · policy & economics

Written by a government economist for analytical communication, and the closest match to a statistical-agency standard, drawing on the same lineage the FT vocabulary builds on.

IBCS & agency style guides
IBCS · Eurostat · ONS · Urban Institute

IBCS for consistent business and financial reporting, alongside the statistical-agency guides for the institutional rulebook on units, scales, and provenance.

IV.2 · The operating rules

Nine standards every chart passes before it ships.

What the canon implies in daily practice, written as the checks a chart should clear.

01
Pick the chart from the question. Identify the data relationship first, whether deviation, correlation, ranking, distribution, change, part-to-whole, magnitude, flow, or spatial, and let that choose the form. Habit is not a reason.
02
Index to a common base when comparing series of different units or scales, rebasing to a shared period set to 100 so the comparison is of change rather than of magnitude.
03
Reach for a log axis only when the story is multiplicative. Constant percentage growth then reads as a straight line, and an additive comparison never belongs on a log scale.
04
Show the uncertainty. Confidence and prediction intervals are part of the honest picture, since a forecast drawn as a single line overstates the confidence behind it.
05
Keep the bar-chart baseline at zero, because length encodes magnitude there. Truncation is acceptable on a line chart, where position carries the value instead.
06
Prefer dot plots to bars for ranking, and small multiples to a single overplotted chart, so the form matches how the eye actually reads.
07
Direct-label, and keep the data-ink high, placing labels at the line rather than in a legend and removing gridlines, borders, and fills that carry no information.
08
Carry the provenance, stating units and any seasonal adjustment and footnoting the source and the revision date, since the footnote is part of the chart.
09
Hold to one idea per chart, letting the title state the finding rather than the variables. A chart that needs two sentences to defend is usually two charts.

IV.3 · The gallery

Built to the rules, in the A2go palette.

Each figure names the data relationship it serves and the standard it demonstrates. All are computed from representative supply-chain data and rendered in the A2go telemetry palette, so they show the practice in our own work rather than in the abstract.

Ranking · cumulativeJuran · 80/20
2026-06-20T11:25:58.066917 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ Late supplier Capacity conflict Material shortage Quality hold Transport delay Demand spike System error Other 0 20 40 60 80 100 120 140 Failures (count) the vital few Pareto · drivers of OTIF failure 0% 20% 40% 60% 80% 100% 80%
Pareto. The bars descend by frequency while the line tracks cumulative share, and the reference line names the vital few. For A2go, this concentrates attention on the handful of drivers behind most OTIF failures.
Change over time · normalizedagency standard
2026-06-20T11:25:58.154967 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 6 12 18 24 Months 80 90 100 110 120 130 Index (month 0 = 100) Demand Inventory OTIF Indexed to a common base · month 0 = 100
Indexed series. Rebasing every series to a common period set to 100 makes change comparable across different units and scales. It lets demand, inventory, and OTIF be read on one honest axis.
Change over time · multiplicativelog vs linear
2026-06-20T11:25:58.451010 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 10 20 30 Months 0 100 200 300 Linear axis · looks like a sudden surge 0 10 20 30 Months 1 0 1 1 0 2 Log axis · steady % growth is a straight line Same series, two scales · MCP-layer call volume
Log scale. The same series on a linear axis can look like a sudden surge, while on a log axis steady percentage growth reads as a straight line. It allows adoption and call-volume growth to be shown without exaggerating the recent slope.
Distribution over time · uncertaintycentral-bank fan
2026-06-20T11:25:58.569356 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ −10 −5 0 5 10 Months (0 = now) 90.0 92.5 95.0 97.5 100.0 102.5 105.0 Forecast accuracy (%) actual forecast 50% / 90% Forecast with prediction intervals
Fan chart. A central forecast carried alongside prediction intervals that widen with the horizon. It lets every forecast ship with its interval, so confidence is shown rather than implied.
Deviation · controlled processSPC · ±3σ
2026-06-20T11:25:58.649232 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 5 10 15 20 25 30 35 Observation 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Error (MAPE pts) UCL LCL mean out of control Statistical process control · forecast error
Control chart. A center line with control limits, where a point beyond them is a signal rather than noise. It separates a real regime change from ordinary variation.
Change between two pointsTufte slopegraph
2026-06-20T11:25:58.705424 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ Q1 Q3 Site A 71 88 Site A Site B 78 90 Site B Site C 69 84 Site C Site D 83 91 Site D Site E 75 86 Site E Site F 80 72 Site F Slope · forecast accuracy by site, Q1 → Q3
Slope chart. Two periods with lines connecting categories, where slope and rank carry the story and a single decline is unmissable. It reads well before and after a deployment, site by site.
Comparison across a dimensionTufte / Cleveland
2026-06-20T11:25:59.010544 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ DC 1 DC 2 DC 3 DC 4 DC 5 trending down DC 6 DC 7 DC 8 Small multiples · OTIF by distribution center (shared scale)
Small multiples. Many small panels on a shared scale work better than one tangled chart, since the eye can scan for the exception. It shows OTIF across every distribution center at a glance, with the outlier marked.
Ranking · gap to targetCleveland dot plot
2026-06-20T11:25:59.166128 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 70 75 80 85 90 95 100 OTIF (%) Veltran Meridian Castellan Orbu Delphi Nyx Aterra Volt 72 79 81 84 88 90 93 96 Supplier OTIF vs target · gap to close Target (95%) Current
Dumbbell plot. Position on a common scale is read more accurately than bar length, and the dots with their connector show current, target, and the gap at once. It suits supplier scorecards, sorted, with the gap to close made plain.

These were generated with matplotlib for this document and themed to the A2go palette and IBM Plex. The same standards apply whichever engine renders them in production.

IV.3b · The signature set

The same rules, at full reach.

The gallery above proves the standard on the classical forms; this set shows its range — the console register on its abyss ground, the sequential ramp, the composition bridge, the distribution that carries the product's central argument, and the instrument row. Every figure still passes the nine standards, and every one is computed: the set renders from one committed script into this page and into the Office templates' specimen slides, so the chart in the deck and the chart on this page remain the same figure. All data is illustrative and labeled so in the figure.

Change over time · small multiplesConsole register
2026-07-23T07:35:39.716990 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ ORDER ATP coverage 96.2% NOMINAL SUPPLY supplier OTD 87.0% WATCH INVENTORY days of supply 41 d NOMINAL LOGISTICS tender acceptance 61% WATCH DEMAND rolling MAPE 31% BREACH EXTERNAL port dwell 3.1 d NOMINAL SIX SIGNAL FAMILIES · TRAILING 24 H ILLUSTRATIVE DATA · A2GO CHARTING STANDARD · MAPE BAND ≤ 22% PER THE PUBLISHED CASELET
Signal lanes on the console ground. Six families, one lane each — position carries every value, the worded chip carries the state, and Phosphor is spent only on the lit numerals. The demand lane draws its control band, so the breach is a fact of the picture rather than a color's claim. The register for booth screens, dark deck slides, and simulated product staging.
Magnitude over time · sequentialOne hue, stepped
2026-07-23T07:35:39.775735 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 84 99 M T W T F S S W1 W4 W7 W10 W12 84 99 % DAILY OTIF · 12 WEEKS ILLUSTRATIVE DATA · A2GO CHARTING STANDARD
Calendar heatmap. One teal stepped light-to-dark by value — monotonic in lightness by construction — and the episode surfaces on its own; the pale fortnight needs no arrow. Only the extremes are labeled. Daily operational measures with a weekly rhythm: OTIF, fill rate, dock utilization.
Composition · bridgeIBCS lineage
2026-07-23T07:35:39.800317 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ $0.48M Base +$0.22M Carrier detention +$0.31M Expedited freight +$0.19M SLA credits −$0.35M Reallocation recovery $0.85M Net PENALTY EXPOSURE ACROSS ONE QUARTER · $M ILLUSTRATIVE DATA · A2GO CHARTING STANDARD
Waterfall. The comparators stay neutral, and the single accent is spent where the judgment sits — the recovery that reallocation earned back. Bars keep the zero baseline; the running level is a dashed hairline. Explaining a period's movement: exposure, variance, and cost bridges.
Distribution · stripPosition over color
2026-07-23T07:35:39.827864 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 8 24 48 72 h DECISION SLA · 8 H ESCALATION · 24 H < $10k $10k – $100k > $100k one $240k decision waited 61 h AGE AT DECISION, BY VALUE AT STAKE ILLUSTRATIVE DATA · A2GO CHARTING STANDARD · N = 120 DECISIONS
Age at decision, by value at stake. Every dot is a decision, and its position against the labeled thresholds is its status — the hues only underline what position already says, so the figure survives grayscale untouched. One annotation names the finding. The argument drawn as a distribution: the decisions worth the most wait the longest.
Magnitude vs targetFew · bullet
2026-07-23T07:59:48.849800 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ OTIF 94.1% WATCH ROLLING MAPE 31% BREACH band ≤ 22 · lower is better TENDER ACCEPTANCE 61% WATCH INVENTORY TURNS 8.2× NOMINAL FOUR KPIS VS TARGET · TICK = TARGET · BANDS = POOR / FAIR / GOOD ILLUSTRATIVE DATA · A2GO CHARTING STANDARD · MAPE + TENDER FIGURES ECHO PUBLISHED CASELETS
The instrument row. Four KPIs against target ticks and qualitative bands — length reads the measure, position reads the target, the worded chip reads the state, and the inverted metric says so in words. Scorecards and review headers, where a dashboard would overclaim.

IV.4 · Tooling

Authoring once, and rendering everywhere.

The aim is a single charting standard that serves the analyst's notebook and the product interface without anything being redrawn. A grammar-of-graphics engine comes closest, because the same specification renders in both places and themes to the brand by default. The recommendations below run from the shared standard outward, and none of the core stack carries a license cost.

Vega-Litefree · the shared grammar
The recommended standard for both surfaces. Analysts author in Altair, the Python binding, inside notebooks, and the same Vega-Lite specification renders in the product through vega-embed or react-vega. It brings grammar-of-graphics rigor and built-in statistical transforms, and it themes to the brand tokens, so one specification serves the notebook and the application alike.
visx · D3free · signature charts
For the brand-custom charts that have to be pixel-precise, such as the signal-trace and the console-register figures. It offers full control and themes completely, and visx embeds natively in React. Best reserved for the figures that must be exactly on-brand.
EChartsfree · dashboards
For the fast, broad operational charts of the console register. Its coverage is wide, including Pareto, heatmap, Sankey, and geo, and it performs well on large data. A sound default where speed matters more than a custom look.
matplotlib · Altairfree · publication exports
For static, print-grade figures such as case studies, PDFs, and the analyst's working charts. Altair keeps the grammar consistent with the product, and matplotlib produced the gallery above.
Datawrapper
For colleagues who are not engineers to produce agency-grade charts without development time, which suits marketing and case-study visuals. It is sold by subscription, and current pricing is worth confirming before committing.
Highcharts
The option worth buying if a single supported, accessible, embeddable engine is preferred over maintaining custom code. Internal-use licenses run from roughly $400 to $500 per developer, and embedding in a distributed product is negotiated separately and tends to run into several thousand and beyond. It brings a strong accessibility module, a Python integration, and vendor support, with AG Charts and amCharts as comparable commercial alternatives.
On Figma. Figma is a design tool, and the right one for shaping the interface and the brand system itself, including the mockups, the component library, and the look of each register. What it is not is a charting engine. A chart drawn by hand in Figma is an illustration of a chart rather than a computed one, so it cannot hold to the underlying data or to the standards set out in this part. The sound division of labor is to let Figma define how the charts should look, and to let the charting engine render the real ones to that design, so that the picture in the mockup and the chart in the product remain the same thing.
The build pattern. Whichever engine is chosen, the palette, the IBM Plex type, and the nine rules above are encoded into one shared theme, an "a2go-charts" preset, so that every chart arrives on-brand and rule-compliant by default. That theme and the component library can be scaffolded in Claude Code on top of the selected engine, which keeps the analyst's output and the product's output identical by construction.

Closing

Decisions of record.

Three decisions carry the identity, and the reasoning for each is in the sections above.

On color, Signal Teal #00707A — the Brand 3a palette, adopted June 2026 as the live identity; the values in §0 are canonical. On type, IBM Plex across the three registers, for the cohesion it brings. On charts, a standard built on Vega-Lite, so one specification serves both the analyst and the product, with a small brand theme encoding the rules so consistency holds without constant effort.

Addendum — decisions of record, 18 July 2026. On the loop: the five-step frame — Signals → Decision → Approval → Execution → Outcome — is the notation of record; Sense · Decide · Act is its compact shorthand, and this document's palette keys its step mapping to the shorthand. On surfaces: the product surface is the decision notebook, everywhere; “Console” names a visual register in this document and never a product surface, and “command center” vocabulary is retired. On marks: A2go is a trade name of Analytics2Go, Inc. — the wordmark carries no registered-mark symbol; ® never appears with it.

Anyone producing material in A2go's name — a document, a deck, a page, a chart — should start from §0 and hold to it. The aim throughout is the one named at the start, to let the market recognize the judgment we sell before a word is exchanged.