Structure your Meta ad names so Kickbite can group performance by every concept variable you're testing — angle, offer, product, format, variant — using simple contains() queries.

Free-text ad names break creative reporting. Names like "shop the spring deal" or "TF best video v2 final FINAL" tell Kickbite nothing about what's being tested. Every ad becomes its own bucket. You can't ask "which angle is winning across all our products?" or "how does video compare to image, holding angle constant?" without manually tagging hundreds of ads after the fact.

When you're running 50–200+ creatives, this is the difference between answering "which angle wins?" in seconds vs. hours. A structured ad name carries the test hypothesis inside the name itself: every concept variable you care about — angle, offer, product, format, variant — appears in the string, ready to be grouped.

The convention below makes ad naming a system instead of a habit. Once it's locked, your team's media buyers can't accidentally drop a dimension, invent a synonym, or break the report.

Kickbite groups performance by running contains() queries on the ad name string. A query like contains('ErvaringenKlanten') returns every ad whose name includes that substring, regardless of where it sits in the name.

Two consequences:

That's the whole mechanic. The rest of this article is about making the values inside the name collision-proof and consistent.

A Meta ad name is a single underscore-joined string of dictionary values:

Worked example:

Reading the example: brand TF, format Video, funnel Sales, offer Deal, product ToiletPietro, angle ErvaringenKlanten, variant Text1. The recommended reading order for humans is <brand>_<format>_<funnel>_<offer>_<product>_<angle>_<variant> — but again, Kickbite doesn't care about order.

Kickbite's segmentation can only group by values that appear in your dictionary. Lock the dictionary before you start naming ads.

These four variables drive Kickbite's primary segmentation. Every ad name must include exactly one value from each.

For the offer field, use NoOffer rather than None — None would collide with the substring None appearing in other contexts.

For genuine multi-product creatives, concatenate product values alphabetically: BadgoedBeddengoed, ApparelKidsBedding. There's a trade-off here, covered later.

Angle is where 80% of repeatable creative-testing wins come from. The other dimensions slice your data; angle tells you which message resonates. Give angles room to grow — you'll keep adding new ones as you test. Common extensions to the four starters: Probleem, Demonstratie, VoorNa, Schaarste, Educatie, Influencer.

Every ad in a test (which is almost every ad) gets a variant identifier in the format <Type><Number>.

Allowed types: Text, Hook, Thumb, Cta, Music, Voiceover, Concept.

Even a single ad in a test of one gets Text1 or Concept1. Never blank — when a test concludes you need to know whether you tested text variations, hook variations, or thumbnail variations.

You can add your own optional dimensions (audience, placement, geo, lang) as long as the values follow the formatting rules and don't collide with anything else in the dictionary.

Sit down with whoever owns paid-social analytics on your team and answer one question: "Six months from now, what dimensions do I want to group creative performance by?"

The answers become your dictionary. For each dimension, write down the complete list of allowed values. Once the dictionary is locked, no freestyle additions — every new value goes through a quick collision check (covered below) and gets added formally.

A starter template you can copy:

Lean on the starter values where they fit, replace where they don't. The product list is yours; the angle list will grow with every test you run.

Because Kickbite uses contains(), the same collision risks that plague UTM tracking apply here. Three rules:

1. Distinct values across dimensions. A value should appear in exactly one dimension's dictionary. If Bundle is an offer value, do not also use Bundle (or any string containing Bundle) as an angle. A contains('Bundle') query would otherwise return both offer-bundle and angle-bundle ads, mixed together.

Worked counter-example: a media buyer asks the team to add Deal as a new angle. Refuse. Deal is already in the offer dictionary — using it as an angle would make contains('Deal') queries return both offer-deal ads and angle-deal ads, mixed together. Suggest something more distinctive: LimitedTimeDeal, WeeklyDeal, or just use a different angle name entirely.

2. Numeric values get a unit suffix. If you encode discount or price tiers in the ad name, write 10pct, not 10. Without the suffix, contains('10') falsely matches 100, 1099, or 2010. Same logic as the UTM convention.

3. Multi-word PascalCase for ambiguous values. Single short words like Sale are collision risks. Multi-word PascalCase values like ErvaringenKlanten or BundelVoordeel land cleanly in contains() queries because they're unique enough that nothing else accidentally matches.

When someone proposes a new dictionary value, run a quick mental collision check: does the new value appear as a substring of any other dimension's values? If yes, suggest a more distinctive alternative.

Pick the path that matches whatever LLM tooling your team already uses. The output is identical; only the install differs.

A skill auto-activates whenever someone asks Claude to name a Meta ad. It enforces the dictionary, asks for missing fields, and outputs a clean ad name plus a TSV row for the log sheet.

For team members who use chat UIs without skill support.

The prompt enforces the same dictionary and produces the same output. Most teams ship both paths — skill for media buyers using Cowork, prompt for occasional users on Claude.ai.

Every ad name the builder produces comes with a TSV row. Pasting it into the next empty row of a Google Sheet builds your historical creative log automatically.

The headers are:

The hypothesis column is one short sentence the buyer fills in before launch ("does customer-review framing beat discount framing for the toilet line?"). Leave status and notes blank unless you have something to say.

Once your ad names follow the convention, every concept variable is one query away.

[SCREENSHOT: Kickbite segmentation view showing groupings by angle]

Examples:

Combine AND/OR to drill into any combination. The reports answer the questions your team is actually asking — "which angle is winning?", "is video beating image holding angle constant?", "which hook in the bundle test is converting?" — without anyone manually tagging anything.

A note on multi-product ads. contains('KidsBedding') will return both single-product KidsBedding ads and multi-product creatives that include KidsBedding (e.g., ApparelKidsBedding). This is intentional: when you ask "how is KidsBedding performing across all our creative", you want bundles in the answer. If you ever need to isolate single-product ads, add a secondary filter that excludes the bundle string (e.g., NOT contains('ApparelKidsBedding')). For most reporting, the inclusive behavior is what you want.

Two short examples of the convention in practice.

Test social proof vs. discount framing for the same product. Hold every other variable constant; vary only angle:

After launch, contains('ErvaringenKlanten') and contains('Korting') return each variant's performance. The test result is unambiguous because nothing else in the names varies.

Four hook variations of the same concept. Hold every variable constant except the variant identifier:

contains('Hook1') through contains('Hook4') directly answer which hook won.

When the team needs a new value — a new angle they're testing, a new product launch, a new format like Gif — follow this sequence:

Never rename an existing dictionary value. Renaming retroactively breaks Kickbite's historical aggregations: yesterday's Probleem ads will stop matching tomorrow's ProblemFocus query. If you need a different name, leave the old value in the dictionary, mark it deprecated in writing, and use the new value going forward.

What if my ad runs across Meta, TikTok, and Google?
This convention is Meta-only. TikTok and Google have their own naming systems and aren't covered here. Kickbite's Meta-side segmentation works once your Meta names are structured — your TikTok and Google reporting follows whatever conventions those platforms (and your team) use separately.

How long is too long?
Aim to stay under ~120 characters total. That's where Meta's Ads Manager display starts truncating names visually. Kickbite parses past that limit, but humans reading the names in Ads Manager won't. If you're getting close, drop optional context (brand code, funnel) before dropping required concept variables.

What if a media buyer ignores the convention and types a free-text name?
Two safeguards. First, the data validation on your log sheet rejects free-text entries — the sheet itself becomes a gate. Second, Kickbite filters out un-conforming names from segmented reports, so the offending ad simply won't appear in any angle/offer/format grouping. The buyer will notice their work is missing from reports and self-correct.

Can I add my own concept variable beyond the required four?
Yes. As long as the values follow the formatting rules (PascalCase, alphanumeric, underscore-separated) and don't substring-collide with values in any other dimension, add whatever dimensions help your reporting — audience, placement, geo, lang, anything else.

What if I want to rename an angle?
Don't. Renaming retroactively breaks every historical Kickbite aggregation that relied on the old value. Add a new value, mark the old value deprecated in writing, and use the new value going forward. Reports across the renaming boundary will simply show two values until the deprecated one ages out.

Does the order of fields matter?
Not to Kickbite. Field order is a team convention for human readability — the recommended reading order is <brand>_<format>_<funnel>_<offer>_<product>_<angle>_<variant>, but you can pick any order as long as the team uses it consistently.