Brand architecture decisions: house, endorsed, or let go

Brand architecture questions arrive at the studio disguised as design questions. "Can you give the new product its own logo?" is really "should the new product be its own brand?" — and the honest answer is a strategy conversation that happens to end in a logo, not a logo conversation that happens to touch strategy. Get the architecture right and the identities are easy; get it wrong and you'll spend years paying the interest: confused cross-sell, duplicated marketing spend, and a naming tree that looks like a family argument drawn in PowerPoint.

There are three canonical architectures — branded house, house of brands, and the endorsed middle — and countless consultant diagrams grading them on matrices. Here's how we actually run the decision inside brand strategy engagements, with the trade-offs we argue about in the room.

The three shapes, honestly

Branded house. One master brand on everything: the company, the products, the features. Product names are descriptive and lowercase-energy — think of how a certain fruit company names things. Strengths: every marketing dollar compounds into one reputation; launches borrow trust instantly; one design system, one voice, one budget. Weaknesses: one scandal burns the whole house; the master brand's positioning constrains every product; and naming gets hard, because descriptive names age as products evolve beyond their description.

House of brands. A quiet parent and loud children — each product with its own name, identity, site and audience. Strengths: each brand can occupy a position the master can't (cheaper, weirder, younger); a crisis is quarantined; acquisitions slot in without surgery. Weaknesses: you build each reputation from zero, so marketing spend doesn't compound; internal teams duplicate everything; and customers may never learn the siblings exist, which makes cross-sell an act of introduction.

Endorsed brands. Children with the parent's surname on the letterbox: "ProductName, by Company." Strengths: borrowed trust with room for the child to have its own character; flexible enough to sunset a product without wounding the parent. Weaknesses: it's the most expensive to maintain well — two identities to keep coherent per product — and the endorsement line erodes under use until nobody's sure how attached the brands really are.

The dirty secret: most companies with three or more products are accidentally endorsed. Marketing uses the parent name, the product team uses the product name, and customers pick whichever they met first. Naming the architecture — even "we're now officially a mess, here's the rules" — is progress.

The decision framework we run in the room

Matrices with twelve scoring rows produce false precision. Five questions, in order, decide most real cases:

1. Who's the buyer, and is it the same person? One buyer evaluating a suite pushes toward a house — every product the same buyer touches is a reason to consolidate. Genuinely different buyers (procurement for one product, a teenager for another) license separate brands.
2. Does the new thing need shelter or distance? A radical, risky or reputationally spiky product wants distance. A solid adjacency that would benefit from the parent's credibility wants shelter. Be honest about which instinct is operating: usually founders want distance when they're proud and shelter when they're scared, which is backwards.
3. What's the acquisition expectation? If you're building to be acquired, a branded house is worth more — a house of brands is a harder purchase to integrate. If you're building to acquire, you need architecture rules that can absorb other people's names (more below).
4. Can you fund it? A separate brand is a second marketing P&L, forever — not a one-off logo fee. If the honest budget for brand-building next year is "the founders post on LinkedIn", you have one brand. Choose accordingly.
5. What would un-doing it cost? Reversible decisions (a sub-brand line under the master) can be made bravely. Hard-to-reverse ones (a fully separate brand with its own domain equity and packaging run) deserve the slower, more evidence-based pass. As with rebrand rollouts, switching costs — not taste — set the timeline.

Score the answers and the architecture mostly chooses itself. When it doesn't, the disagreement is real and strategic, and should go to the exec team with the questions, not the colour swatches.

Sub-brands: where houses quietly subdivide

The pure branded house rarely survives Series C. Products accrete names — internal codenames leak, big features demand marquee status — and suddenly there's "Acme Insights" and "Acme Pulse" and someone in sales has invented "Acme for Banking". Sub-brands are how houses accommodate gravity; the craft is in the rules:

• One naming pattern, enforced. Pick the grammar — "Master + descriptive word" is the durable default — and write down what qualifies for a name at all. Our test: a nameable sub-brand must have a distinct audience or a distinct revenue line. A feature with neither gets a sentence, not a name.
• Tier the identities deliberately. Master brand: full system. Named sub-brands: wordmark variants using the master's type and palette — no second colour worlds. This is where colour systems engineered for change pay off: architecture rules expressed in tokens survive staff turnover.
• Designate the lockup once. The "Product by Company" endorsement lockup, typeset and spaced once, shipped everywhere as an asset. The day three teams hand-set it three ways, you've lost the endorsement's entire value — consistency.
• Sunset clause. Every sub-brand gets a written condition for absorption back into the master (e.g. "if X% of buyers can't distinguish it from the master in the annual survey, it merges"). Sub-brands without sunset clauses are how navigation bars become museums.

The acquisition scenarios that flip the answer

Architecture debates assume organic growth. Acquisitions break the assumptions in three recognisable ways:

You acquire a brand stronger than yours in its niche. The instinct is to absorb it; the maths often says endorse or park it. If the acquired brand owns its search demand and community goodwill, renaming it burns the asset you paid for. The right shape is frequently "Acquired, by Parent" for two years, then re-test.

You acquire five small brands nobody's heard of. Fold them into the house fast, transition sentences and all — the rename playbook applies unchanged. Equity their names don't have, you don't owe.

You get acquired. Your beautiful branded house becomes someone else's sub-brand decision. The single best preparation is boring: keep your architecture rules written and your asset inventory current, so the integration team inherits a manual instead of a mystery. Companies with legible architecture get to keep more of their identity; the illegible ones get standardised.

A real-world composite from our files: a fintech (call it Northwind-scale, after our ledger client) acquired a beloved community tool with a passionate user base. The deck said "rebrand to the master in 90 days". The churn model said the community was 40% of the tool's value and 90 days wasn't long enough to transfer trust. The compromise — endorsement for two years, master applied to the new paid tiers first, community tiers renamed last and gently — kept the migration inside the churn budget. Architecture is churn mathematics wearing a creative director's jacket.

Making it stick: the one-pager

Whatever you decide, the decision dies without documentation people can find. We end every architecture engagement with a single internal page — not a deck, a page:

1. The chosen shape, in one sentence ("We are a branded house; products get descriptive names; exceptions require exec sign-off").
2. A diagram of the current portfolio mapped onto it, including the embarrassing orphans.
3. The naming grammar and sub-brand qualification test.
4. The endorsement lockup asset, linked.
5. The sunset clause and the date of the annual architecture review.

Then the discipline: architecture goes on the roadmap like tech debt. Every quarter, someone asks whether reality has drifted from the page. It always has; the question is whether you correct the work or the page.

Key takeaways

• Choose architecture with five questions: same buyer, shelter vs distance, acquisition expectations, honest budget, and the cost of reversal.
• Most multi-product companies are accidentally endorsed brands; naming the architecture is half the fix.
• Sub-brands need enforced grammar, tiered identity rules, one canonical lockup and a written sunset clause.
• Acquisitions flip the answer: absorb the weak, endorse the strong, and keep your own house legible for the day you're acquired.
• Fund what you choose. A second brand is a second marketing P&L for as long as it lives.
• Document it on one page and review quarterly like debt — because that's exactly what it is.

FAQ

We already have a mess of six names. Where do we start? With the census and the questions, not with new logos. Map the portfolio, run the five questions against each product, and sort everything into consolidate / endorse / sunset. Most portfolios we audit lose two names in the first pass, and nobody misses them by the next quarter.

Is the endorsed model ever the permanent answer? Yes — when the parent's credibility materially changes purchase decisions but the products genuinely need separate personalities (different tone, different buyers). B2B suites with one flagship consumer product are the classic case. Just budget for maintaining two systems properly.

How do we stop teams inventing new names? You don't stop creativity; you route it. Publish the qualification test and a one-step approval path. If naming something takes a form and a fortnight, teams self-select: the features that deserve names survive the fortnight, the rest ship with good descriptive labels — which was the right outcome anyway.

Does this apply to a two-product startup? Applied lightly, yes. Two products is the cheapest moment to set the grammar — renaming at two products is an afternoon; at twelve it's a program. Write the one-pager now, keep it to twenty minutes, and thank yourself at Series B.