On 1 October 2026, Supermetrics moved its campaign-management claim onto its product page, and it arrived with a sentence worth reading twice: “Every change is logged in Supermetrics against who made it.”
Read that as a business, not a feature. There is now a Supermetrics audit log in everything but name: write access switched on per ad account, four approval modes, and a change history attributed to whichever person or credential triggered the change. On 20 August we published the opposite — that the write capability sat in a blog post while the canonical product pages carried no write vocabulary at all. That record is expired, and the update is what this post is about.
Supermetrics has made the first genuine product-level move into the space Ott occupies, and it did it in Ott’s own vocabulary. The honest reply is not to argue that the gate exists. It is to ask what the gate leaves behind.
What Supermetrics actually shipped
Give the company its due: this is a real capability, documented on the surface a technical buyer evaluates rather than buried in a release note. The page at /platform/campaign-management was last updated on 1 October and carries three claims that matter.
| Claim | Verbatim |
|---|---|
| Per-account writes | “Write access is also switched on per ad account, so only the accounts you enable can be changed at all.” |
| Approval modes | “Decide when a person has to sign off: never, before a campaign goes live, before budget changes and launches, or before any change to a live campaign.” |
| Change history | “Every change is logged in Supermetrics against who made it, whether it came from your AI chat or a Supermetrics Studio dashboard.” |
The per-account write scoping is the right default and we will say so: enabling a connector should not silently hand an agent every account in the workspace. New campaigns being created paused is the correct direction too. So is logging changes somewhere rather than nowhere. Every one of those is an improvement on what the category shipped in the spring.
Now look at where the log lives. Not that it exists — that it exists inside Supermetrics. A vendor keeping a record of its own writes is a real thing, and it is not the same thing as a record that survives the vendor.
Four approval modes, and the one that approves nothing
The blog post that accompanies the launch (30 September) spells out the four modes by name. Three of them are gates. Read the fourth:
| Mode | What it does |
|---|---|
| No human approval | Changes apply immediately. |
| Going live | Requires approval before enabling or unpausing a campaign. |
| Budget + going live | Also requires approval for budget changes on live campaigns. |
| All live changes | Requires approval for every change to a live campaign, including enabling it. |
This is a shape we have documented across the vendor market all year: every governance property is a setting, and every setting has a position that produces no evidence. A retention window can be set to zero days. An approval gate can be switched to “never.” When the first option is selected, there is no authorisation event for anything to record — so the gap is structural, not a completeness problem. The setting is not the flaw. The point is that the setting, on its own, tells you what the vendor will let you do, not what the vendor can prove you did.
The two questions the copy does not answer
Both questions are answerable from the two sentences Supermetrics published. Neither is answered by them.
- Does the log leave the platform? “Logged in Supermetrics” is a location. An export you can only read by signing into the system that produced it is a report; a record is something you can hand to someone who does not have your login.
- Does it name the approver, or the actor? “Who made it” is the credential that fired the write. “Who approved it” is a person who accepted responsibility for it. In the fourth approval mode those are the same person; in the other three they are not.
That second distinction is not pedantry; it is the entire question. When a supervisor asks why a claim ran on a particular client’s account in a particular market, the useful answer is a name attached to a decision, not the service account that executed it. A log that records the trigger is a debugging tool. A log that records the decision is evidence.
Grading their own homework
At Programmatic IO in New York on 1 October, S4 Capital’s Sir Martin Sorrell made the independent-record argument better than we have. As AdExchanger reported it: “‘Ownership of data brings problems,’ Sorrell said, because it opens agencies up to accusations that they’re grading their own homework.”
Substitute “agency” for “platform” and the sentence is the whole problem. A tool that records its own writes is marking its own paper. That does not make the record dishonest — it makes it non-independent, which is a different and more precise objection. The regulator, the client and the ad platform are three parties who might dispute an operation. A record held by the fourth party that performed it cannot settle the dispute.
Count the words the page does not use
Here is the measurable part. Supermetrics is precise about the mechanism and silent about the artefact. Running a character-exact scan over the served text of the product page returns the following.
| Present | Absent | |
|---|---|---|
| On the page | “approval” ×3 · “who made it” ×3 · “Change history” ×2 · “export” ×1 | audit 0 · record 0 · jurisdiction 0 · retention 0 · client engagement 0 · governance 0 |
The absence of the word “audit” is not a marketing choice. It is a specification. A company that has built an audit trail says the word; a company that has built a change history says “history.” Same for jurisdiction, which is the field that turns one change into three compliance questions when a brand runs across three markets. And same for retention, which is the field that decides whether the record still exists when the question arrives eighteen months later.
This is the same gap we have tracked across the vendor market all year, arriving at the vendor whose product page just moved closest to us.
What actually gets asked, months later
None of the following is a performance question. All five land on a team that has to produce an answer from a system, not a memory.
- Who approved it? A person who accepted the decision, not the credential that executed it.
- For which client engagement? One agency runs many clients; the record has to say whose account it was.
- Under which jurisdiction? Promotion rules are national. A change made for a UK-facing brand is a different question than the same change for a German one.
- What was the value before and after? “A budget changed” is not the same as the number it changed from.
- Can you export it without logging in? If the only copy lives inside the platform that made the change, the platform is also the arbiter of it.
The gate is now commodity
Supermetrics is not alone, and that is the more useful observation. Over the last quarter, an approval gate before the write became the default line across the MCP vendor layer: markifact, a hosted approval-gated ad MCP vendor that shipped two new servers at the end of September; Toffu, whose DV360 MCP leads with “Every write waits for your approval”; Windsor, where writes “require the end user’s explicit approval”; and Synter, promising a “Human Spend Gate on Every Mutation.”
Four vendors, one phrase, and not one of them names the approver, or stores the record anywhere the platform that accepted the write cannot reach. When a gate is on every README, the gate stops being the differentiator. The question moves one level down, to what the approval leaves behind — the actor the platform reports, the value before and after, and the timestamp. Those are the fields Ott keeps, and it is why the answer to a stronger Supermetrics is not a louder claim about having a gate.
Where Ott fits
Ott’s Activity Logging records the four fields a regulator-facing question resolves to: the actor, the value before and after, the timestamp, and the linked object. It is stored independently of the ad platform, so the answer survives a platform change, an account ban or a reporting outage, and it stays readable outside the tool that made the change. An MCP connection means an AI assistant working on your client’s accounts reads through the same structure rather than beside it.
Three limits, stated plainly, because we hold vendors to this and should hold ourselves to it. Ott is read-only — it records what happens in the accounts and in your operations; it does not write to ad platforms. On coverage, Meta is connected today with Google Ads on the roadmap, so an agency running four platforms is getting the record for the surface it uses most. And we publish no retention number, which is the same gap we criticise above — it is a limit, not a hidden detail.
The reason scope matters more than volume here: Ott prices by brands and seats, with no per ad account, per data source or per Telegram join fee. Adding accounts to a client does not change the cost of keeping the record, which is the wrong incentive for a compliance surface to have. See how that is metered, or read the side-by-side comparison.
The correction, and the question
We told you in August that Supermetrics’ write capability was announced in a blog post while the canonical product page stayed silent. That was true then and it is false now, and we would rather say so than quietly delete the take. Read the earlier post and you will see the surface it described has changed under it.
What has not changed is the shape of the question. Supermetrics now states, in its own published words, that every change is logged against who made it. That is a real sentence and a real capability. Ask the two things it does not say: whether that log leaves their platform, and whether it names the person who approved the change or only the account that fired it. A log you keep about yourself is a dashboard. A regulator asks for the change history on one regulated campaign, months later, by a named approver, under a named jurisdiction, from a system that is not the one that made the change.
That is a record. It is worth building before anyone asks for it.