Brand Reflection Series · Episode 2 · companion to Color fingerprint & Sound fingerprint

Three templates, one institution.

What happens when an AI reads brand homepages instead of brand briefs? For 23 Czech brands, it draws a map where the boundaries don't match the sectors. This report shows that map, walks one brand (Air Bank) through what it means for their strategy, and gives you a tool to generate the same brief for any brand on it.

Data & cloud engineer at Alma Career. Built across a few weekends to test what brand competitive analysis looks like when it starts from public data instead of from pitch decks. Part of a five-piece Brand Reflection Series.

By Barbora Šandová. Source on GitHub.

Eight brands. Four sectors. One homepage template.

ČSOB, Česká spořitelna, Fio banka, Air Bank, Komerční banka, Generali, ČEZ, Škoda Auto. The boundary between them, on the homepage layer, is zero.

How it works

An AI language model reads each Czech brand's homepage and plots the brand on a 2D map (filled circles, coloured by neighborhood). Brands whose homepages read alike land near each other; brands that read differently land apart. The same model reads top rising Czech queries from Google Trends and drops them onto the same map (outlined diamonds, sized by how fast each query is rising), so you can see which empty zones a rising query lands in. Full technical breakdown in the appendix.

01

The map

Brands are filled circles, colored by neighborhood. Rising queries are outlined diamonds, sized by how fast each query is rising. Distance on the map approximates how differently the brand homepages read to the AI — brands close together share homepage language; brands far apart don't. Neighborhood names come from the dominant sector in each group, not from me.

established / heritage registerpromo / retail rhythminstitutional registerconsumer registerumělá inteligenceakce lidl tento týdenúroky hypotéka 2026chatgpt českyčerný pátek 2026inflace 2026elektromobily 2026vitamíny d3ceny energiífotovoltaika dotaceneomezená data tarifai generátor obrázkůgrilování receptyškoda octavia 2026kaufland letákymagnesium na únavunetflix česko ceníkspoření na důchodvoyo programrohlík sleva nový zákazníkkybernetická bezpečnostojeté auto kontroladr max akcebezlepkové potraviny5g pokrytíkofola nebo coca colaT-Mobile ČRO2 ČRVodafone ČRTelecomMattoniPilsner UrquellStaropramenBeverageČSOBČeská spořitelnaFio bankaAir BankKomerční bankaŠkoda AutoČEZGenerali ČPBanking + heavy industryLidlAlbertRohlíkMcDonald's ČRKFC ČRKofolaPilulkaSeznam.czAvastMass-market consumer
Click any neighborhood to inspect the brands and vocabulary that define it. The dashed red arrow shows Air Bank's recommended trajectory — from inside the Banking + heavy industry neighborhood toward the edge of mass-market consumer (the position held by Pilulka and Avast).

Telecom

3 brands — T-Mobile ČR, O2 ČR, Vodafone ČR

Beverage

3 brands — Mattoni, Pilsner Urquell, Staropramen

Banking + heavy industry

8 brands — ČSOB, Česká spořitelna, Fio banka, Air Bank…

Mass-market consumer

9 brands — Lidl, Albert, Rohlík, McDonald's ČR…

Rising Czech query

26 trends — sized by how fast each is rising

02

One brand, end to end: Air Bank

If the map is the diagnostic, this is what doing something with it looks like. The brand is Air Bank; the moves stay the same for any brand on the map. Four steps from "where am I" to "what slides go in the deck."

01

Where Air Bank sits on the map

Air Bank embeds inside the Banking + heavy industry neighborhood, next to ČSOB, Česká spořitelna, Fio banka, Komerční banka, Generali, ČEZ, and Škoda Auto. Three banks plus an insurer plus an energy company plus a carmaker all read as one brand to the homepage layer.

On the map, Air Bank tilts slightly toward the consumer side of the institutional neighborhood — closer to the boundary than its peers. That tilt is the only visible difference between Air Bank and Komerční, and it's small.

02

The vocabulary it shares with seven other brands

The words that distinguish the institutional neighborhood from the other three on the map, taken directly from how those homepages read, are: banka, škoda, čez, fio, česká.

If Air Bank's homepage shares those tokens with ČEZ's homepage, then to a buyer's first read, Air Bank is ČEZ. The challenger positioning — "the bank that's different from Komerční" — is decided in the campaign layer and quietly undone in the homepage layer.

03

Which playbook moves apply, in order

Move 02 — list the institutional vocabulary explicitly in the brief and rule it out. Not as creative aspiration; as a word-level constraint. The first draft that lands with banka, partnerem or silnejší on the homepage goes back.

Move 03 — stop pitching against named banks. The boundary that matters is the one between the institutional neighborhood and the mass-market consumer neighborhood. The target isn't the center of mass-market (Lidl, Albert) — it's the edge closest to Air Bank's current position, held by brands like Pilulka and Avast that read as consumer without reading as supermarket. That edge is the strategic ask.

Move 01 — after the rebrand, re-score. If the homepage didn't move neighborhoods, neither did the brand. Soft success at the campaign layer with no movement at the homepage layer is the failure mode this episode is built to catch.

04

The brief this produces

Below is the literal one-page brief the diagnostic outputs. Not a description of what the brief would say — the brief itself, ready to hand to a copywriter on Monday. Substitute any brand on the map and the layout stays the same; only the token lists and the trajectory change.

03

Generate the same brief for any brand

The Air Bank brief above is one application of the diagnostic. This generator runs the same logic for any brand in the set. The "drop" and "adopt" word lists come from the words that distinguish each neighborhood from the others; the "where the map points next" targets come from the brand's nearest neighboring neighborhood and the brands in it ranked by distance.

The generator is mechanical. Geometrically nearest does not mean strategically right. Pilsner Urquell's nearest neighborhood is mass-market consumer, but recommending a premium beer drift toward Lidl is obviously wrong. Use the output as a structured starting point for a strategist's judgment, not as a directive. The Air Bank brief above shows what the human edit looks like.

04

Appendix

The reference material behind the diagnostic. Open whichever you need; the report above stands without them.

Four patterns the map reveals

Each pattern is a finding that the map made visible. Patterns explain why the four neighborhoods matter; the Air Bank walk-through shows how to act on them.

01

One institutional brand wearing eight logos.

All five Czech banks (ČSOB, Česká spořitelna, Fio, Air Bank, Komerční), plus Generali, plus ČEZ, plus Škoda Auto land in one homepage neighborhood. Eight brands, four sectors (banking, insurance, energy, automotive), one register. The boundary lines we draw between these brands at the campaign level do not survive the homepage layer.

Why it matters"Established Czech institution" is a homepage template, not a sector. A challenger that wants to feel distinct from Česká spořitelna has to break out of the institutional template, not out of the banking sector. A bank that wants to feel different from ČEZ has the same problem.

02

The mass-market neighborhood is a homepage-shape problem.

Nine brands across six sectors — Lidl, Albert, Rohlík (retail), McDonald's, KFC (QSR), Kofola (beverage), Pilulka (pharma), Seznam.cz (media), Avast (tech) — share one homepage template: short positioning line on top, product grid below, "akce / novinky / aktuální nabídka" everywhere. Avast (security software) lands here because their homepage looks like a supermarket aisle.

Why it mattersThe buyer who lands on any of these reads the same shape first. The campaign creative might be distinctive; the homepage is not. A brand that runs distinctive campaigns onto a generic homepage trains buyers to forget the distinctive part on contact.

03

Telecoms own a tight, isolated neighborhood of three.

T-Mobile, O2 and Vodafone embed close to each other and far from every other brand. The "always-on connectivity / esim / mobilní tarif" language is differentiated enough to be its own territory, but homogeneous enough that the three competitors are essentially interchangeable inside it.

Why it mattersNo fourth brand can join the telecom neighborhood without becoming the fourth-best telecom. A new entrant could occupy any direction outside it (a telecom that reads as a lifestyle brand, a productivity brand, a cultural brand). The neighborhood boundary itself is the strategic opportunity.

04

The neighborhood boundary IS the strategic decision boundary.

The most distinctive Czech brands sit close to a neighborhood boundary, not at its center. Air Bank holds the institutional neighborhood but barely — its language tilts toward consumer. Pilsner Urquell holds the food & beverage neighborhood but tilts toward heritage / luxury. Strong brands are edge brands; safe brands sit at the center of their neighborhood.

Why it matters"Which competitor do we want to look like" is the wrong question. The right question is "which homepage neighborhood do we want to leave." Most brand strategies adjust within the neighborhood (Air Bank vs Komerční); the bigger move is between neighborhoods (Air Bank vs Avast vs Albert).

Full playbook (five moves)

Five moves a strategist or creative director can run on next week's pitch. Moves 01–03 already show up in the Air Bank walk-through; 04 and 05 cover trends and re-scoring.

01

Score the client's homepage before writing the brief.

Run the client URL through the same pipeline as Episode 2. If they're in the mass-market neighborhood, every campaign starts with the same handicap: the homepage trains buyers to expect a promo template. If they're in the institutional neighborhood, campaign creative has to actively pull buyers out of "Czech legacy institution" framing.

02

Identify your homepage's defining vocabulary, then refuse to use it.

Each homepage neighborhood has a vocabulary that distinguishes it from the others ("akce / novinky / magazín / leták" for mass-market; "bankou / silnější / partnerem" for institutional). A brief that wants to escape its neighborhood lists those words explicitly and rules them out. Escape is a word-level constraint, not just a creative ambition.

03

Position against neighborhood boundaries, not against named competitors.

A pitch slide showing "we differ from Komerční banka because X" is intra-neighborhood wallpaper change. A pitch slide showing "we will move from the institutional neighborhood toward the consumer one — here's the homepage shape we'll adopt" is a real positioning move. Re-score after the rebrand. If the homepage didn't move neighborhoods, neither did the brand.

04

Use Brief Generator for category-specific numerical targets.

The map tells you which neighborhood the client sits in. The Spot Scorer's Brief Generator tells you the numerical targets a distinctive spot in that sector needs to hit. Topography identifies the trap; the brief generator gives the spec for breaking out of it.

05

Read trends through the neighborhood lens, not the brand lens.

A rising query that lands in an empty zone of the map is a real opening for the brand whose neighborhood boundary touches that zone. A rising query that lands inside a dense neighborhood is an opportunity only for the brand willing to break its homepage vocabulary. Saturation score in the opportunity table is the diagnostic.

What this is and isn't

A few honest things to know before you act on the report or commission the diagnostic for a client.

  • The 24-brand demo is a method test, not a Czech market census. When you commission this for your client, we agree the brand set on the call (typically the client plus 3–6 competitors, plus a few reference brands). The public demo shows what the method does; a client engagement runs it on your actual competitive set.
  • The map groups brands by how their homepage text reads, not by what sector they're "supposed" to be in. Two brands near each other use similar homepage language and structure; two brands far apart don't. That's why a bank, an insurance company, an energy company and a carmaker can land together — the homepage layer doesn't care about your sector classification.
  • The top words in each neighborhood are descriptive, not prescriptive. They show what register a neighborhood shares (some are inevitably brand names, since a cluster of banks will have "banka" in its signature). Read them as "the language this neighborhood already speaks," not as "ship these specific words."
  • Homepages are one signal, not the whole brand. A brand can play very differently on TV, OOH or social and still get classified by its homepage snippet here. The map describes "how the homepage reads." If you want to score how a specific spot lands, that's the Spot Scorer, not this report.
  • The brief generator is mechanical and sometimes wrong. It points Pilsner Urquell toward Lidl, which is obviously not the move for a premium beer. Use the auto-generated brief as a structured first draft; the strategist's edit is where the judgment lives. The Air Bank brief above shows what that edit looks like in practice.
How it was built (technical specifics)

For the reader who wants the AI / ML pipeline named.

Each Czech brand homepage (title, meta description, first headings, first non-promo paragraphs, capped at 600 characters per brand) is read by a multilingual AI language model — intfloat/multilingual-e5-base — which turns the text into a 768-dimensional vector capturing meaning across languages. The same model reads the top rising Czech queries (pytrends + hand-curated fallback). The vectors are projected to 2D for the map via UMAP and grouped into neighborhoods via KMeans (run on the original 768-dim AI vectors, not on the 2D projection). Neighborhood labels are rolled from sector composition. The opportunity score for each trend is velocity × (1 − mean similarity to the three closest brands in the AI vector space) — a high score means the trend lives in an empty zone of the map.

This is a text-mode AI pipeline. The sister projects in the series use vision AI (Color Fingerprint) and audio AI (Sound Fingerprint); the Spot Scorer combines all three into one multimodal score for any video URL.

Rising trends and where they fall on the map

For each rising Czech query, the three brands whose homepage reads most similarly. Rows flagged low‑signal mean the top three brand matches are essentially tied — the "winner" is arbitrary, which is itself a finding (a sector gap in the brand set).

Trend Velocity Nearest brands Saturation
umělá inteligence 280 Vodafone ČR(0.83) · Avast(0.82) · Lidl(0.80) 0.81
inflace 2026 220 Vodafone ČR(0.79) · ČSOB(0.78) · Generali ČP(0.78) 0.78
chatgpt česky 240 Česká spořitelna(0.81) · Generali ČP(0.81) · Avast(0.81) 0.81
úroky hypotéka 2026 low signal 250 ČSOB(0.83) · Generali ČP(0.82) · Vodafone ČR(0.82) 0.82
černý pátek 2026 230 Vodafone ČR(0.82) · Seznam.cz(0.81) · Albert(0.81) 0.81
akce lidl tento týden 260 Lidl(0.86) · Seznam.cz(0.83) · Albert(0.82) 0.84
fotovoltaika dotace 190 Vodafone ČR(0.80) · Pilulka(0.80) · ČEZ(0.79) 0.80
elektromobily 2026 200 Škoda Auto(0.82) · Vodafone ČR(0.81) · O2 ČR(0.80) 0.81
voyo program 160 ČSOB(0.77) · Vodafone ČR(0.76) · Avast(0.76) 0.76
vitamíny d3 200 Pilulka(0.82) · Albert(0.82) · Lidl(0.81) 0.81