Brand refresh: when and how to renew a brand without rebranding

A brand refresh renews how a company looks and sounds while the name, the position and the audience stay put. By the end you will have nine practices, each one taken from a company that documented its own identity update, and the single signal that tells you when a full rebrand is the honest answer.

Ten companies published a first-party account of their own visual update between June 2024 and September 2026. Several call the same kind of work a rebrand in their own post. This piece sorts them by Ciaran Jack's test instead: nine report no change to who the company sells to, so those nine qualify as refreshes here regardless of what the company itself called it.

Thinkific is the exception. Its own account describes narrowing who its relaunch targets, which by this piece's own definition makes it a rebrand. It stays in this set because the practices its case illustrates still hold on their own.

Several of the ten state in writing which visual element they kept and why; check each case, since the pattern doesn't generalize cleanly. The name stayed in all ten.

Inventory where the brand already contradicts itself, before touching the logo

Write down every surface where the brand disagrees with itself, before a designer opens anything. The audit is the brief.

Spike, an incident alerting product, published exactly that inventory on 2026-09-04. In the team's own words: "The website was blue, maybe too blue one could argue but the social media header was black. The changelog images ranged from white, gray, gradients, and blue backgrounds. Every marketing material looked different because Spike had never found a home."

The sequence matters as much as the list. Spike rebuilt the product first and turned to identity after: "Once the product felt right, we started looking at everything else."

Ask what the brand stands for today, and ask people outside the room

Collect the words your customers and your own team already use for you, then design toward those words. The refresh amplifies a read that exists; it does not invent one.

Cast AI, a Kubernetes automation platform, ran that as its first step. Yuri Frayman wrote on 2025-02-04: "we spoke with numerous customers to understand their perception of Cast. A clear trend emerged: they consider us experts in Kubernetes." The new logo, typeface, palette, imagery and site were built to amplify those attributes.

Statsig ran the same step facing inward. Designer Cat Lee described the process on 2025-03-05: the team audited its own brand alongside brands it found inspiring, then interviewed employees and leadership, and rolled the findings into one file that guided every visual decision that followed.

Name the one element you will not touch

Pick the element people use to recognize you and take it off the table in writing, before the exploration starts. Everything else becomes negotiable once that line exists.

Warp, a terminal for agentic coding, put the rule in a caption: "Warp's refreshed logo retains the same glyph while introducing an updated wordmark." The October 2024 post is signed by Kyle Ribant, Kim Pham and Alisha Mowder.

The reason they give is specific enough to argue with, which is what makes it useful: "did you know that Warp's glyph is a broken terminal window that cracked under extreme speed? Of course we have to keep that."

Warp's own post makes the case for the glyph in writing. It makes no equivalent case for the wordmark. That asymmetry is the whole decision, and it is worth writing down before anyone gets attached to a direction.

Redraw the mark for the smallest size it actually ships at

Test the mark at the size it lives at most: a favicon, an app icon, an avatar. A mark that only works on a slide fails everywhere it matters.

Thinkific, a learning commerce platform, published the reasoning on 2025-06-24. Evan LePage, Director, Content & Creative, describes the icon work as refining "for better scalability, ensuring clarity at all sizes by reducing whitespace, thickening its elements, and adjusting for better visual balance."

The wordmark got the same treatment: "we squared the terminals and gave it more weight... while optimizing legibility at smaller sizes."

Statsig reports the same constraint driving its own logo simplification, describing the new mark as one that "works better at smaller sizes and has more gravity."

Give the mark a meaning you can state in one sentence

If nobody on the team can explain the symbol in a sentence a customer would repeat, the symbol is decoration. Get the sentence and the drawing to agree, whichever one comes first.

Prefix.dev, a conda package platform, changed its symbol for a reason it could say plainly. Tim de Jager and Wolf Vollprecht wrote on 2025-10-09 that the old package block was being read as "a sugar cube or marshmallow," so they moved to a puzzle piece: "Trying to get all your software and dependencies to work together can feel like needing to complete a difficult puzzle."

Spike passes the same test in its own post: the double arrow is the spike going up when an incident fires and coming back down when it resolves, with the second arrow standing for escalation.

Neither mark's drawn shape can be verified from outside without rendering it, so treat both as what each company states about its own work, on the date it stated it.

Decide whether your color is equity or noise, and say which

Run one question on your primary color: does a customer already associate it with you, or is it just your category's default? The answer decides whether the palette is a starting point or a problem.

Stream, a real-time communication API company, treated its color as equity. Design Director Jared Laham wrote on 2024-06-20: "By using our legacy main color, Stream Blue, as the starting point, we were able to pair our secondary and tertiary palettes around it easily to bring a better sense of balance and tone."

Thinkific reached the opposite conclusion about the same color family. Its post is blunt about why: the brand had always been blue, "like so many other software companies," and the team wanted primary colors "that really separate us from any other company in our industry." The new pair is Rooted, a dark plum, and Beacon, an off-white. Both are live in thinkific.com's own markup as of 2026-09-16.

Write the voice next to the visuals, in the same project

Define how the brand sounds while the palette and the type are still being chosen, so the two halves ship together. A refreshed look on unchanged copy reads as a paint job.

Merge, a product integrations platform, did both at once. Visual Designer Lanna Wang published the result on 2025-03-10, and the post calls the work a brand refresh even though its headline says rebranded.

The voice is three named characteristics with a definition attached to each: "Confident: Clear and conveying expertise. Clever: Witty and evoking energy. Committed: Empowering but not at the expense of being over-sentimental." The post also says where the characteristics flex, naming a social post and an article as different applications.

The typography moved in the same release: FH Oscar Pro for headlines, Inter for body and buttons, IBM Plex Serif for customer quotes. All three are declared in merge.dev's live CSS as of 2026-09-16.

Publish the design system in the same release as the identity

Ship the tokens, the type roles and the component library on launch day, in a place the team can open. Without that, the refresh has an expiry date measured in quarters.

TanStack, the open-source application stack, brought on Andy Beutler as its first lead designer, the first person on the team whose full-time job was to own these problems, and shipped the refresh on 2026-07-29 as "a new logo system, global type, design tokens, /ds, a component kit, Phosphor icons..."

Tanner Linsley's reason for the system is maintenance, and he says so: "Colors have primitive ramps and semantic tokens now... components have one home... when the site starts drifting, we have something shared to bring it back to."

That system is public. tanstack.com/ds returns a live catalog as of 2026-09-16, with sections for logos, colors, typography, iconography, Figma tokens, semantic tokens and roughly twenty components.

Sequence the rollout, and publish what is not changing

Tell people the order the change will arrive in, and which parts of the company are staying exactly as they are, in the same post. Both halves cut how much guessing your customers have to do.

Benchmark Email gives the second half its own section, titled "What's New (and What's Not)." Allie Wolff, its VP of Product Marketing, wrote on 2025-10-06: "Yes, our new look is different, but the team behind Benchmark isn't. You'll still find the same passionate people, the same mission..." Cast AI closes its post the same way, stating that everything beyond the look and feel remains the same.

Stream handles the first half. Its post reports that the website was refreshed already, then sets the expectation for the rest: "Over the next few weeks, you will see all other visuals around Stream aligning with this new direction in our marketing materials and products."

TanStack is even more direct about the unfinished parts: "This isn't finished. Some pages still need work, some things still feel halfway between the old world and the new one, and that's okay."

How to choose which one to run first

Start by confirming a refresh is the right project at all. Ciaran Jack's test is two sentences long: "A refresh is the right answer when the position is sound but the expression has decayed," and "A rebrand is the right answer when the underlying position has shifted, or needs to."

He names a third answer people skip: "A new logo will not fix a broken sales motion." His precondition for either project: a leadership team that can't agree on what the brand stands for today gets nothing out of it.

Once the project is confirmed, the inventory of contradictions you wrote down first (where the brand disagrees with itself) picks your order for you: the surface a customer hits first, and the one furthest from the rest, go first. Naming the untouchable element and testing it at its smallest size come next in every case, since everything decided after them has to work inside that constraint.

What goes wrong

Refreshing the expression when the position is what decayed. This is the failure the test above exists to catch: confirm the position is sound before starting, or the refresh becomes cosmetic cover for a strategy problem.

Ciaran Jack describes the tell: "A refresh that has to be revisited as a rebrand within twelve to eighteen months usually means the strategy was unsettled and the refresh papered over it." His recommendation in that case is to resolve the position in a smaller engagement first.

Refreshing a brand that is still building recognition. If the brand has been live for under three years, Ciaran Jack's read is that it is probably not broken: "Most early-stage brands are not yet broken; they are still building recognition. Refreshing too early restarts that clock."

Treating launch day as the finish line. This is what happens after every practice above ships correctly. Jason Suttie's subject is rebrands specifically, and his diagnosis is that they "rarely collapse at the strategy stage," while the handoff from launch into daily practice is where they break.

His list of the five most common reasons opens with no operational ownership and includes unrealistic rollout timelines and no measurement of brand adherence. His summary of the mechanism: "If the only thing holding the rebrand together is goodwill and a PDF, drift becomes far more likely..."