Forensic Brand Architecture: Fix Agency Onboarding

Why onboarding keeps failing

A new agency asks for your brand guidelines. You send a PDF with a logo, colours, three adjectives ("bold, human, trusted") and a mission statement. Six weeks later the copy is technically on-brand and commercially wrong: it stresses benefits customers do not care about, makes claims sales cannot back up, and misses the objections that actually stall deals.

The guidelines are not wrong; they are incomplete. They describe how the business would like to sound. What a new provider needs is how the business actually wins: the words customers use, the proof that closes deals, the claims that are safe, and who approves what. That knowledge usually lives in the heads of a founder and a few salespeople.

What "forensic" means here

"Forensic brand architecture" is the name we use for rebuilding a brand from evidence rather than opinion. It is not an industry standard term. The idea is simple: treat every brand statement as a claim that needs evidence, and build the onboarding pack only from things you can point to.

The pack you are building

Components of an evidence-based onboarding pack

Component	Built from	Question it answers for a new provider
Customer language bank	Reviews, testimonials, sales call notes, support emails	How do buyers describe the problem and the result?
Claim register	Case records, reports, contracts, certifications	What can we say, with what evidence, and what must we never say?
Objection map	Lost deal notes, sales team interview	What stops people buying, and how do we answer it honestly?
Voice rules	Your best-performing emails, pages and ads, compared with your weakest	What does good copy from us look like, with examples?
Offer boundaries	Founder interview, proposals, terms	What do we do, for whom, and what do we not do?
Approval rules	Leadership decision	Who signs off what, and how quickly?

Seven steps

1. Collect the evidence

Gather, in one folder: public reviews, testimonials, a sample of sales call notes or recordings, reasons for won and lost deals, frequent support questions, your five best and five worst performing pieces of marketing, and any published results. Remove or redact personal data before any of it goes to an outside provider. If a provider will process personal data for you, UK GDPR requires a written controller and processor contract.

2. Extract customer language

Read the evidence and copy exact phrases into three lists: the problem in their words, the outcome in their words, and the hesitations. Do not paraphrase. The value is in the specific words buyers use, which are almost never the adjectives in the brand guidelines.

3. Build the claim register

List every claim your marketing makes or might make. For each, record the evidence, its date, where it is published and a status: approved, needs evidence or never use. The ASA has said advertisers need evidence to back up the claims they make, and holding that evidence before a claim goes out is the safe habit; a claim register is the practical way to do that. For the exact requirement, read the substantiation rules in the current CAP Code on the ASA website. This single document prevents more problems than any other part of the pack.

4. Map objections

Interview the person who sells most. Ask what makes deals stall, what competitors say about you, and what customers wish they had known earlier. Pair each objection with the honest answer and the evidence behind it.

5. Derive voice rules from your own best work

Compare your best-performing material with your weakest and write five to eight rules that explain the difference. Each rule needs a good and a bad example from your own material. "Be human" is not a rule; "name the customer's situation in the first sentence, as in this email" is.

6. Draw the offer boundaries

State what you do, who it is for, and what you do not do or do not promise. Providers over-promise most often where boundaries are unwritten.

7. Set approval rules

Who approves copy, claims, spend and anything legal or technical, and within what turnaround. Slow approvals undo fast production.

Test the pack with a blind brief

Give the pack and a one-paragraph brief to someone who does not know your business well (a new colleague, a freelancer on a paid trial, or an AI assistant) and ask for a short page of copy. Then check:

• • Does it use customer language from the bank?
• • Does every claim appear in the register as approved?
• • Does it follow the voice rules without being told twice?
• • Does it stay inside the offer boundaries?

Every miss points to a gap in the pack, not in the writer. Fix the pack and test again.

Using the pack in onboarding

With the pack in place, onboarding becomes a review of the pack rather than weeks of discovery. A new agency or provider should read it, ask questions, produce a test piece, and receive feedback against the pack. The same pack feeds AI tools as stored context, which is why it is the foundation of any business content system.

This is also how AfM starts with a client: making it easy to supply context is the first priority, because good output depends on it. The client provides evidence; AfM's AI operating system organises it into a working source of truth, and the founder reviews it before the work relies on it. See how it works.

FAQ

How long does it take to build the pack?

It depends mainly on how quickly you can gather evidence and get time with the people who sell. The collection step is usually the slowest; the writing goes quickly once the evidence is in one place.

Should we share sales call recordings with an agency?

Only with care. Recordings contain personal data about customers and staff. Share redacted notes or summaries where possible, and if recordings must be shared, make sure a proper processor contract is in place and your privacy information covers it.

What if our best-performing marketing contradicts our brand guidelines?

Take it seriously. It usually means the guidelines describe an aspiration the market has not bought. Decide deliberately whether to follow the evidence or to invest in moving perception, but do not leave providers to guess which you mean.

Can an AI tool build the pack for us?

AI is useful for sorting evidence, pulling out repeated phrases and drafting first versions of each component. The decisions (which claims are approved, which objections to answer and how) need the people accountable for the business.