How to build a brand kit in 2026

Open almost any guide to building a brand kit and the first instruction is to pick your colours. It is the most satisfying step, which is exactly why it comes first, and it is also why so many kits stop being useful about a week after they are finished.

What follows is the build sequence we would actually follow, in order, with the reason the order matters at each stage. If you want the definition before the procedure, what a brand kit actually is covers the five parts and why the thing is a decision system rather than a folder. This post assumes you are already convinced and want to build one this afternoon.

The argument underneath the order is short. Most brand-kit advice produces a style, which is a set of visual constants, when what you need is a brand, which is a described position that a style follows from. Assemble colour-first and you get a style. It holds up beautifully right until somebody has to write a caption, at which point there is nothing in the kit to consult, and the caption gets written from scratch by whoever is at the keyboard. Every step below exists to stop that happening.

Step one: positioning and audience

This is the step everybody skips, and skipping it is the single reason kits look generic.

Write four things down, in plain sentences, before you open a colour tool.

Who this is for, specifically enough that they would recognise themselves. Not "small business owners". Something closer to "solo consultants who bill for their time, have no marketing help, and dread writing about themselves". Specificity is not a stylistic preference here. It is the input every later step reads.

What they are struggling with, in their words rather than yours. If you have had the conversation even a few times, you already have the sentences. Use theirs, not the tidied-up version.

What you sell, and what you deliberately do not. The second half does more work than the first. A brand that has decided what it refuses is a brand with edges.

What you want to be true after somebody has read three of your posts. Not a mission statement. A belief and a feeling, written as a sentence you would say out loud.

Then apply the swap test. Put a competitor's name at the top of what you just wrote. If it still reads as true, you have written a category description rather than a position, and everything you derive from it will be exactly as interchangeable as it is.

Step two: voice

Voice comes second, before anything visual, and this is where most kits and almost every tool part company with reality. You write far more than you design: captions, replies, headlines, the small sentence under a button. Voice is the part of a brand people meet most often and the part least likely to be written down anywhere.

The full framework is in how to define your brand voice, so here is the procedural short version. Define the character doing the talking. Define the tone range, how you sound at the calm end and at the urgent end, so the flex has bounds. Write a vocabulary list of the words you reach for, and a banned list of the words you refuse. The banned list is the fastest-acting part of the entire kit and the one people find easiest to write.

Then prove it. Take three sentences you have actually published, rewrite each one to match, and keep both versions in the kit. A voice section with before-and-after pairs settles arguments. A voice section made of adjectives does not, because "friendly and professional" is not a rule anybody can follow, including you.

Step three: palette

Now you can choose colours, and now they will not be arbitrary, because there is a position to derive them from.

Choose by role rather than by swatch. One primary that carries the brand. A small set of neutrals, which will quietly do most of the actual work. One accent, used rarely enough that it still means something when it turns up. Next to each colour, write where it is allowed to appear. That note is the part that survives contact with a real layout; the hex code on its own never has.

Two practical constraints. Check your text and background pairs for contrast at the moment of choosing, not after you have fallen in love with them. And pick fewer colours than you think you need, because every extra one is a decision you will have to re-make under deadline.

Step four: typography

One or two typefaces. A workhorse for body text, and optionally a display face for headings. Choose the body face first, which is the reverse of what most people do, because body text appears on every surface you will ever make and a display face appears on a fraction of them.

Then set an actual scale. Define the sizes, name them, and write the line height and weight for each. "Big, medium, small" is a guess repeated forever; a named scale is a decision made once. Test the smallest step on the smallest surface you publish to, because type that reads beautifully in a design tool at full width routinely falls apart at phone width.

Step five: logo direction

Direction, not a finished mark. Most people do not need a final logo this afternoon, and waiting for one is the most common way a kit never gets finished.

What you do need is the direction and its constraints: wordmark or monogram, the feeling it carries, the clear space around it, the minimum size, the backgrounds it may sit on, and the things it may never do. A mark without usage rules gets stretched, recoloured and crowded until it stops reading as yours.

Repic's branding engine proposes a logo direction alongside the palette and the type pairing, all derived from the same profile, which is the point: the three agree with each other because they came from one description rather than three separate afternoons.

Step six: the usage rules

A kit fails at the moment somebody has to make something the kit did not anticipate. Usage rules cover that case, and they are the whole difference between a kit that survives real work and one that gets admired once and then ignored.

Write them as defaults with named exceptions. "Headings use the display face at scale step two, except in the feed, where the workhorse reads better at small sizes." Write down the answer to every question you have had to decide twice. And when you hit a choice the kit does not cover, do not simply decide it, add it.

The most practical rule of all is that the kit has to live where the work happens. A kit in a slide deck three folders deep gets skipped under deadline, which is exactly when consistency matters most. Inside Repic your brand kit is not a document you remember to open, it is context the creation engines already carry when they generate carousels, static posts and stories.

Step seven: stop it decaying

Kits do not break. They decay, and the decay is invisible because every individual decision that caused it was reasonable at the time. Three habits prevent most of it.

Date everything, so you can see which parts have not been touched since your offer changed. Review on a business trigger rather than a calendar one, because the events that matter are the offer moving and the audience moving, and neither of them sends you a reminder. And score the parts rather than the whole, because a single brand-health percentage tells you nothing you can act on, while "audience: underspecified, add two concrete struggles" tells you which sentence to write next. That is what Repic's Quality Engine does per component, alongside flagging work that was generated before you changed the field it depended on.

What is actually different in 2026

One thing, and it changes what a kit has to contain.

Your kit is now read by machines at least as often as by people, and anything you leave unspecified does not get politely skipped. It gets filled in with the most statistically ordinary option available, because that is what a model trained on everything produces in the absence of instruction. Every gap is a small pull toward the average, applied once per gap per generation, always in the same direction.

That has two consequences for how you build. The kit has to be structured rather than prose, because a field can be inspected and scored and a document cannot. And it has to contain voice, because voice is the largest unspecified surface in most brand setups and therefore the one that gets averaged hardest.

If you would rather start from a coherent draft than a blank page, Repic holds your brand as a structured profile, generates the kit from it, and then generates carousels, static posts and stories from that same profile. The describe-once walkthrough shows the sequence and where the confirmation steps sit. Repic is in early access, so places go out in weekly batches in signup order rather than instantly.

The order is the part that transfers whatever you build it in. Position, voice, palette, type, logo direction, rules, upkeep. Do those in that sequence and you end up with a brand that your style follows from, rather than a style you spend the next year explaining.