Set up UTM conventions that last

Five values appended to a destination link, carried into whatever counts sessions and orders on your side. They are how a visit arriving at your site can be attributed to the advertising that produced it, in your own system, under your own definitions.

What they are not is a second copy of platform reporting. Tempio holds the platform-side counts — one row per entity per day, spend, impressions, reach, clicks, conversions, conversion value, with the account's currency stamped on the row and every ratio computed at read time. The parameters travel somewhere else entirely. The convention is the join between the two, and the reason a platform's claimed conversions and your own order data can be compared at all.

They are also the only part of this whole subject that is entirely within your control. The platforms decide their own windows, definitions and restatement behaviour. The tagging is yours.

Source: where the click came from. The platform or the property, one token, and the same token forever. This is the value most often spelled three ways.

Medium: what kind of traffic it is. Paid search, paid social, display, email, affiliate. A small closed list, decided once and written down, because an open list here turns every report into a vocabulary exercise.

Campaign: the thing you are running. The name you would give it in a plan. Keeping this aligned with the campaign name inside the ad account is what makes a joined report readable rather than a translation exercise.

Content: the variable inside the campaign. The creative, the audience, the placement — whichever one you will actually want to compare. Picking the wrong dimension here is the most common way a convention becomes useless while still being followed.

Term: the keyword, where there is one. On search, the matched keyword. Elsewhere, usually empty, and an empty value is better than a filled one that means something different per campaign.

If you cannot say, for two links, which differs in campaign and which in content, the convention has not been decided — it has been improvised.

Lowercase, everywhere. Most analytics tools treat these values as case-sensitive strings, so one source spelled two ways is two rows in every report that groups on it. This is not pedantry; it is the single most common cause of a fragmented report.

Hyphens, never spaces. Spaces become encoded characters, which are not what anybody typed and not what anybody will search for later.

A closed vocabulary per parameter. Source and medium especially. Write the permitted values down where the person building a link will see them, because the moment the list is in somebody's head it has three synonyms in it.

Building the links by hand is how all three rules get broken, which is the argument for generating them: the UTM builder applies the same shape every time, and the consistency is the whole feature.

A new person. Somebody who was not there decides where a link belongs. If the rule is not written down, they will invent a reasonable-sounding value, and reasonable-sounding is what fragments a report.

A new channel. Adding a platform should not require rethinking the scheme. It requires one new source value from the closed list, which is why the list being closed and short matters.

A year. The values are in your historical data permanently. A scheme that encodes this quarter's team structure is a scheme that stops parsing when the team changes.

A handover. The convention belongs in the same place as the rest of the account's reasoning. A brief that explains why the split is what it is and does not say how links are tagged is a brief that leaves the next person guessing.

Worth stating plainly, because tagging is often sold as a solution to the wrong problem.

It does not reconcile the platforms. Two platforms can each honestly claim the same sale under their own windows and definitions, and no tagging scheme changes what they claim. What it gives you is a third account — your own — to hold both against.

It does not make the numbers agree. Your own system will count differently from every platform, permanently, and the useful target is a stable ratio rather than equality. Why platform numbers disagree is the sorting of which differences matter.

And it does not change anything in the ad accounts. The arrangement there, stated once: your connections read your accounts, a change is written as a draft on the record with the state it would change from and the checks it passed, applying it is gated on approval — by you, or by a rule you wrote — and no platform write reaches a customer account until that platform has granted write access. Applying a cleared change to a platform is in private beta.

Do not rewrite history. Rewriting live links breaks nothing forward and fixes nothing backward, and a half-migrated scheme is worse than either version of it.

Treat the changeover as a break in the series: pick a date, apply the new convention to everything created from that date, and write the date down wherever the numbers are quoted. A series with a documented break is readable. A series that silently changed meaning is not, and nobody will remember which quarter it happened in.

Then leave it alone. The value of a convention is entirely in its stability, and a scheme improved twice a year is a scheme that has never been useful for a full year.

Build a weekly performance review is the rhythm this makes possible, and the naming convention template is where the campaign-side half of the same decision lives.