Module 2 settled what Fellwise should stand for; this lesson asks how that choice is expressed. The starting distinction is between brand identity and brand image. Identity sits with the sender: what the organisation intends the brand to be. Image sits with the receiver: the associations people actually hold in memory, which Module 1 treated as the source of customer-based equity (Keller, 1993). Identity comes first in the sense that managers can decide it; image is the test of whether the decision worked (Kapferer, 2012).

Kapferer's brand identity prism organises identity into six facets (Kapferer, 2012). Physique is the tangible and sensory: products, logo, colour and packaging. Personality is the character the brand would have as a person, a facet Jennifer Aaker (1997) measured along five dimensions: sincerity, excitement, competence, sophistication and ruggedness. Culture is the set of values and origins the brand draws on. Relationship is the mode of conduct between brand and customer, such as mentor, companion or expert. Reflection is the typical user the brand portrays, and self-image is how buyers see themselves when they use it. The prism is read along two axes. Physique and personality describe the sender, while reflection and self-image describe the receiver. Physique, relationship and reflection are the visible, social facets; personality, culture and self-image are internalised.

David Aaker (1996) offers a complementary model: a core identity that should survive new markets and products, an extended identity that adds texture, and four perspectives (brand as product, organisation, person and symbol) that stop planners reducing identity to product attributes. Both frameworks are planning heuristics rather than validated measurement scales. Their value lies in disciplined comparison: write the intended identity, research the perceived image, and treat the gaps as the brief for the assets, voice and governance that follow.

Identity describes what a brand means; distinctive brand assets are how buyers pick it out in the first place. Romaniuk (2018) defines them as elements other than the brand name, such as colours, logos, shapes, characters, fonts, taglines, sonic cues and advertising styles, that trigger the brand in memory. The argument sits within the Ehrenberg-Bass account of brand growth: brands grow largely by being easy to notice and recall across many buying situations, so being recognisable at a glance on a shelf or in a two-second scroll is itself a source of mental availability (Sharp, 2010). Lesson 2.2 set out that debate; the concern here is practical measurement.

Romaniuk (2018) assesses each candidate asset on two dimensions. Fame is the share of category buyers who link the asset to the brand. Uniqueness is the brand's share of all the brand mentions the asset receives, which shows whether buyers link it to that brand rather than to competitors. Plotting both gives a grid with four quadrants. High fame and high uniqueness is "use or lose": a strong asset that must be used consistently or it will decay. High uniqueness but low fame is "investment potential": the asset belongs to the brand, but too few people know it yet. High fame but low uniqueness is "avoid solo use": people connect it to the brand but also to rivals, so it needs the name alongside it. Low on both is "ignore or test".

Three cautions follow. First, measure rather than assume. The brand team sees its own assets every day and is poorly placed to judge them, so the grid needs survey data from category buyers. Second, assets are built through use, so changing them is costly; a redesign that discards a famous asset can discard years of accumulated memory. Third, the line between "high" and "low" is a judgement. Many published audits draw both lines at 50%: half of category buyers for fame, and half of the asset's brand mentions for uniqueness. Treat this as a convention rather than a rule, state whichever benchmark you use, and remember that a prompted survey measures recognition under ideal conditions, not noticing in a cluttered feed. Sonic assets and characters deserve the same testing as logos, because audio and video now carry much of a brand's exposure.

Verbal identity is everything a brand says and how it says it: its naming, taglines and key messages and, above all, its tone of voice. It deserves the same discipline as visual identity, because words now make up much of the brand experience: product pages, app notifications, customer-service chats, packaging and social replies. Read through Kapferer's (2012) prism, voice is where personality and relationship become audible, and a consistent voice can become a distinctive asset in its own right.

A widely used distinction separates voice from tone. Voice is the stable character of the brand, while tone adjusts to the reader's situation and emotional state. Mailchimp's public content style guide states the principle plainly: the voice stays the same, and the tone changes with context. A complaint reply and a product launch should both sound like Fellwise, but not in the same register.

Good voice principles are few, specific and testable. A common format pairs each principle with what it is not, so that writers can see the boundary: "knowledgeable, not technical", "warm, not chummy", "plain-spoken, not blunt". Each principle then needs worked examples: a before-and-after rewrite, a list of words the brand uses and avoids, and guidance on the grammar that carries the voice, such as sentence length, contractions, the use of "we" and "you", and whether humour is ever appropriate. Without examples, principles collapse into adjectives almost any brand could claim. "Friendly, honest and expert" distinguishes nobody; it is the verbal equivalent of an asset with no uniqueness.

Two tests help. The swap test asks whether a competitor could publish the sentence unchanged; if so, the voice is doing no work. The pressure test asks how the voice behaves when the news is bad: a delayed order, a faulty product, an apology. Voice under pressure will matter again in Module 6, where tone can protect or erode trust. Finally, the voice must stay clear in translation and for readers using assistive technology, so plain English is an accessibility requirement as well as a stylistic choice.

Identity decays quietly. Each off-brand leaflet, stretched logo or improvised social post seems harmless, but together they dilute the fame of the assets and the coherence of the voice built in Lesson 3.2 and Lesson 3.3. Brand guidelines, often called a brand book, codify the identity: a summary of the positioning and the prism, logo and colour specifications, typography, photography style, the approved distinctive assets and the tone-of-voice sheet. Guidelines alone rarely hold a brand together. Governance is the set of roles, processes and tools that gets them used: a named owner, a central asset library with approved templates, clear approval routes for high-risk material, training for agencies and new staff, and periodic audits.

The design problem is balancing consistency with flexibility. Consistency is what builds distinctive assets (Romaniuk, 2018), but over-rigid rules encourage workarounds and stale creative. A practical approach is to separate fixed elements, which never change, from flexible ones, which teams can adapt within stated limits. Module 4 will develop this further when it distinguishes consistency from integration.

Generative AI sharpens both sides of the trade-off. It lets small teams produce copy and imagery at a previously impossible scale, and it can be briefed with voice principles and example texts. But its output drifts towards generic phrasing, it can produce images that echo other brands' assets, and it can state product facts or environmental claims that are simply untrue. The UK Advertising Standards Authority has said that its codes apply however content is generated, and that disclosing AI use is very unlikely to cure a fundamentally misleading message (ASA, 2025). It also notes that UK law sets no blanket requirement to disclose AI use, so disclosure is a matter of judgement. In the EU, Article 50 of the AI Act has applied since 2 August 2026: providers must mark AI-generated content in a machine-readable form, and deployers, which would include a brand such as Fellwise, must disclose deepfakes (European Parliament and Council of the European Union, 2024). A 2026 amendment gave generative systems already on the market until 2 December 2026 to meet the marking duty, so check current guidance. Sensible governance treats AI as a production tool inside the system: approved tools, prompts and exemplars; human review against the voice sheet and the facts; a clear disclosure policy; and a record of what was generated, when and by whom.