The eleven shapes of a market

The default segmentation

Open any consumer brand strategy deck from the last 20 years and you'll find a 5-segment customer model. Sometimes 4. Sometimes 6. Almost never more.

This isn't because markets have 4-6 segments. It's because survey research can resolve about 5 segments before sub-sample sizes get too small for statistical significance. The 5-segment output is a methodology artefact, not a market reality.

Search-based segmentation — clustering keywords × products via Leiden community detection — consistently finds more. 8-15 for a focused category. 50+ for a broad one. The Bose Germany headphone market resolved into eleven shapes.

The Bose finding

When we ran Leiden Surprise on the German headphone market — 3,796 ASINs, 38,127 customer voices, 25K keywords — eleven distinct segments emerged:

1. True wireless premium — Bose QC Ultra, Sony WF-1000XM5 ecosystem
2. Over-ear noise cancelling — Bose 700, Sony WH-1000XM, Sennheiser Momentum
3. Sport / wireless earbuds — Jabra Elite Active, Beats Studio Buds
4. On-ear lifestyle — Marshall, JBL Tune series
5. Audiophile wired — Sennheiser HD 660S, Beyerdynamic DT 1990
6. Studio / professional — Audio-Technica ATH-M50x, Beyerdynamic DT 770 Pro
7. Kids headphones — JLab JBuddies, Onanoff BuddyPhones
8. Gaming over-ear — HyperX Cloud II, SteelSeries Arctis 7+
9. Bone conduction — Shokz OpenRun, Aftershokz Aeropex
10. Sleep / ambient — Bose Sleepbuds II, Soundcore Sleep A20
11. Budget-tier wireless — Soundcore Q30, Anker Soundcore Life

A 5-segment customer model would collapse 7 of these into "other". A strategy built on 5 segments is structurally blind to where most of the category's growth is happening.

Why search reveals what surveys can't

Three reasons:

01 — Search is continuous and at scale. Every keyword search is a consumer voting for an intent. Aggregate billions of searches and the segment structure emerges from the data — no sample-size limit, no question-framing bias.

02 — Search is revealed preference, not stated preference. Surveys ask consumers what they want. Search reveals what they're already doing. The gap between those two is the difference between marketing intuition and category reality.

03 — Search clustering uses Leiden with Surprise. The algorithm has no resolution limit (unlike Modularity, which merges small clusters even when they shouldn't merge). Surprise finds the statistically most distinct partition relative to a null model — which surfaces every genuinely separate demand pocket.

Categories and their shapes

Pattern: narrower categories produce fewer shapes (pro printers: 3 — there really are only three meaningfully distinct buyer types). Broader categories produce more (parenting: 14, with everything from car seats to baby monitors to sleep aids to feeding).

Whatever the count, it's almost always more than 5, and a strategy that ignores the additional segments is leaving share on the table.

What 11 segments does to strategy

Three changes:

01 — Own + defend + ignore. With 11 segments you can be honest: own 3-4 where you have natural advantage, defend 3-4 adjacencies, ignore 3-4 where structural disadvantages mean modest investment will lose. The 5-segment world forces you to claim presence everywhere.

02 — Real competitive sets. In a 5-segment model, "Bose competes with Sony, Sennheiser, and JBL" is the answer. In an 11-segment model, the answer is segment-specific: Bose's gaming presence competes with HyperX and SteelSeries, not Sennheiser. Different fights, different strategies.

03 — Defensible niche identification. Some of the 11 segments are dominated by single brands with structural moats. Some are wide-open. Some are growing 30%/year, some are shrinking. A 5-segment view averages these and loses the texture that strategy is made of.

Sub-segment-level granularity

Within each of the 11 shapes, there's further structure. The over-ear noise cancelling segment in DE contains:

• Premium (Bose 700, Sony WH-1000XM5)
• Sub-premium (Sony WH-CH720N, Bose QC SE)
• Audiophile-tier (Sennheiser Momentum 4, Bowers & Wilkins PX7)
• Mid-market (Anker Soundcore Q45, Soundcore Life Q35)
• Budget (Tribit QuietPlus, OneOdio A70)

That's 5 sub-segments within 1 segment. The full structure is fractal — and the deeper you go, the more strategically actionable the sub-segments become.

When 5 segments is enough

To be fair: there are categories where 3-5 segments is genuinely the right number. Pro photo printers, for instance: there are 3 segments because there are only 3 buyer types — wedding/portrait pros, fine-art photographers, commercial reproduction shops.

But these are exceptions. For most consumer categories, the natural shape is 8-15. The 5-segment default is a research-firm habit, not a market reality.

The methodology piece

How does Theia find the shapes?

1. Build a bipartite keyword × product graph from search data
2. Weight edges by CTR-adjusted traffic
3. Apply per-keyword dampening so head terms don't dominate
4. Run Leiden Surprise
5. Reassign products to centroids until stable
6. Cross-language merge (same segment in DE + UK + FR + IT collapses)
7. Dissolve clusters < 5 products via LLM reassignment
8. Name each cluster via HHI-weighted distinctiveness

The result: 11 shapes for German headphones. 9 for UK mirrorless. 14 for UK parenting. Each refreshed weekly. Each defensible by the underlying data.

What this means for strategy

If you're running a category strategy on a 5-segment model in 2026, you're working with the methodology constraint of survey research from 2005. The data to do better is available. The algorithm to do better is open-source. The integrated stack to deliver it weekly exists.

The question is whether you're using it.

Coming next

Next post: Deep web research for B2B markets — why standard social listening misses 90% of where industrial purchase decisions are actually made.

---

Theia uses search-based segmentation to find market structure for consumer and B2B brands. See the Bose Germany case for the full 11-segment example, or read more about Leiden community detection.