One naming convention for campaign, ad set and ad.

a campaign name is not a label, it is a row key. every ad platform export and every dashboard hands you the name as one string, and the only way to answer a question like "did lookalikes beat interest targeting across both channels" is to split that string on one character and treat each piece as a column. if the pieces sit in a different order in every row, or there are four of them in one name and six in the next, nothing splits and you are back to reading names one at a time.

the test is mechanical. drop the export into a spreadsheet, run =SPLIT(A2, "_") or the split-text-to-columns menu item, and see whether column three is the offer in every single row. if it is, you can pivot the account by channel, by objective, by offer and by quarter in about a minute. if it is not, the convention failed, and no amount of care in the platform interface fixes it after the fact.

that is also why the separator can never appear inside a field. one lookalike_1pct in an underscore-joined name adds a phantom column and shifts every field after it by one, which is worse than a messy name, because the split still runs and quietly produces wrong groupings.

the value of a convention comes from position, not from content. field one is always the channel, field two is always the objective, field three is always the offer. once that is true, anyone can read a name without a key, and a pivot table built last quarter still works on names created this week. the three levels here are nested on purpose: the campaign name carries channel, objective, offer and period, the ad set name repeats those and adds audience and geo, and the ad name adds the creative angle and the variant. most exports give you the ad row on its own, so a name that stands alone is worth the extra characters.

blanks are the part people get wrong. leaving a field empty does not just shorten the name, it shifts every field after it into the wrong column, so one row splits into six pieces and the next into four and the pivot silently mixes them. this builder skips empty fields rather than emitting a double separator, but the better habit is to fill every slot with a real short value: all for a geo that runs everywhere, evergreen for something with no end date, none for a level that genuinely has no variant.

use two different characters: one to join fields, another to join words inside a field. meta_signups_pro-trial_2026-q4 reads cleanly and splits cleanly because the underscore only ever means "next field" and the hyphen only ever means "same field, next word". that single rule prevents most of the mess a naming convention is supposed to prevent.

teams lose real weeks arguing about underscores against pipes. it does not matter. what matters is that one document exists, that it lists the fields in order, and that every person launching in the account has seen it. a mediocre convention followed by everybody beats an elegant one followed by two people, because the reporting only works if the account is consistent, not if it is clever.

campaign names usually end up inside utm_campaign on every link the campaign points at, so the same rules apply twice over: lowercase, no spaces, and no characters that need encoding in a URL. this is the practical argument for the underscore. a pipe is readable in the platform interface and becomes %7C the moment the name lands in a link, which puts two versions of the same campaign in your analytics. build the tagged links with the UTM builder, and run anything you inherited through the UTM checker before trusting the numbers.

agree the convention before launch, not after. renaming a campaign later does not annotate history, it rewrites it: most platforms report on the current name, so an export you pulled last month and an export you pull today will disagree about what the same campaign was called. cheap to decide on day one, permanently annoying on day ninety.