A brand voice matrix is a one-page table that tells every writer and reviewer how your brand sounds in each situation: a launch, a complaint, a promotion, a bad-news post. Build one in an hour with four columns: situation, tone dial, say this, never say this. Then store it where the writing happens, so it gets used.
TLDR: Rows are situations, columns are tone dial, a sample line, banned phrases and who approves. Fill 5 rows per brand, test the matrix on 10 old posts, then load it into your AI tool and your review step so it is applied, not just filed.
Start with situations, not adjectives
Most voice guides list adjectives: "friendly, bold, smart". Writers cannot act on those. A voice changes with the moment, so the matrix is built from moments.
Pick the five situations that produce most of your posts:
- Product or campaign launch
- Everyday value post (tips, behind the scenes)
- Promotion or offer
- Customer complaint or public reply
- Bad news (delay, outage, correction)
Add a sixth only if a real situation keeps breaking the first five, such as a regulated claim or a local holiday.
The matrix itself, filled in for one brand
Copy this structure. The example is a fictional outdoor-gear brand.
| Situation | Tone dial (1 formal, 5 casual) | Say it like this | Never |
|---|---|---|---|
| Launch | 4 | "The new 40L pack is out. We tested it on 6 trips." | Superlatives like "best ever" |
| Everyday tip | 5 | "Pack the heavy stuff close to your back." | Jargon, long intros |
| Promotion | 3 | "Free shipping this weekend, ends Sunday." | Fake urgency |
| Complaint reply | 2 | "Sorry about that. Send us your order number and we will fix it." | Defensive wording, emojis |
| Bad news | 1 | "Shipping is 3 days late. Here is why, and what we are doing." | Blame, jokes |
Two details make this work. The "say it like this" column holds a real sentence, not a description, because people copy sentences. The "never" column is short and concrete.
Add three fields that stop the matrix from drifting
A matrix with only tone and examples gets ignored in month two. Add:
- Audience in one line. Who is reading this brand's posts, and what do they want right now?
- Word list. Five words the brand uses, five it never uses.
- Approver per situation. Everyday posts need no sign-off. A complaint reply or bad news may need a manager. Put a name next to each.
For a team with several brands, copy the matrix per brand and change only the sample sentences and the word list. The structure stays the same, so reviewers learn it once.
Test it on 10 old posts before you roll it out
Take 10 posts you already published and ask a colleague to match each to a row, then judge: does the post follow the row?
- If they cannot pick the row, you are missing a situation.
- If most posts break the row, the tone dial is too far from how the brand really writes. Move it.
- If everything passes, your rows are too vague. Tighten the "never" column.
Run this test again every quarter. It takes 20 minutes.
Store the matrix where writing and review happen
A matrix in a PDF decays. Put it in the tools people use.
In Mydrop, each brand can hold its tone and audience notes so the AI caption tools draft in that voice by default, and you can save the situation rows as prompts through the AI content features. Reviewers then see the same rules in the approval step. If you want the template for prompts, our guide to reusable brand voice prompts shows the wording.
Know the limits
A matrix will not fix a team that disagrees about the brand itself. If the five rows spark a long argument, that is useful: you have found the real decision. Settle it once, write it down, and move on.
It also does not replace review. It makes review quicker, because a reviewer can say "row 4 says no emojis" instead of "feels off".
Fill in the first three rows today
Open a blank table, write five situations for your busiest brand, and fill three rows with real sentences from your best past posts. Share it with one reviewer for a gut check. If you manage several brands, start free with Mydrop and add the matrix to each brand so the AI follows it.
Google review
Trustpilot review
Product Hunt comment
