Naming Conventions: How to Write One That Actually Holds
A naming convention fixes the components, order, delimiter and allowed values of a name. Here's what makes one hold, and why most decay within a year.
A naming convention is an agreed rule for how names are constructed: which components appear, in what order, separated how, using which allowed values. A good one is short enough to apply from memory and closed enough that guessing is impossible.
Most conventions are written by people who will follow them and imposed on people who will not. That asymmetry, rather than any weakness in the rule, is what determines whether the convention survives its first year.
What is a naming convention?
An agreed rule for how names are built: components, order, separators, allowed values.
The idea is old and near-universal. Research data management uses it for files, so that a folder sorts meaningfully and a colleague can identify a dataset without opening it (Harvard Medical School Data Management, “File Naming Conventions”, accessed 2026-09-11). Software has them for identifiers. Marketing has them for campaigns, placements and creative.
The mechanics transfer between those contexts. The enforcement problem does not.
A file naming convention is applied by the researcher who will later need to find the file. A code naming convention is applied by the engineer who will later have to read the code. In both cases the person following the rule is the person who benefits from it, and the convention is largely self-enforcing for that reason alone.
A campaign naming convention is applied by someone launching under deadline and benefits a different person entirely, several weeks later. Nothing about the rule is harder. Everything about getting it followed is.
What a convention has to specify
Four things: the components, their order, the delimiter, and the permitted values for each.
| Element | Decides | Failure if left vague |
|---|---|---|
| Components | Which facts the name carries | Names that answer different questions record different things |
| Order | Which position each component occupies | Sorting and pattern-matching both break |
| Delimiter | What separates components | Mixed underscores and hyphens make names unparseable |
| Permitted values | What each component may contain | The one that matters most, and the one most often omitted |
| Case | Upper, lower, or mixed | Email and email become two values in every report |
| Length limit | Where the name stops | Truncation in downstream systems, silently |
Most published conventions cover the first three carefully and the fourth not at all. That is the difference between a format and a convention. A format tells you the shape of a name; a convention also tells you what may go in it. IT Glue’s treatment covers the structural elements well (IT Glue, “Naming Conventions: Examples, Formats and Best Practices”, accessed 2026-09-11).
A worked example
A campaign name broken into its parts, with the allowed value list for each.
NA_acme_springlaunch_2026Q2_paidsocial_v2| Position | Component | Permitted values | Owner |
|---|---|---|---|
| 1 | Region | NA, EMEA, APAC, LATAM — closed | Regional lead |
| 2 | Brand | Approved brand list — closed | Brand team |
| 3 | Initiative | Free text, lowercase, no delimiter characters | Campaign ops |
| 4 | Period | YYYYQn — pattern | Campaign ops |
| 5 | Channel | 8 values, closed | Media ops |
| 6 | Version | v + integer — pattern | Campaign ops |
Four of the six are closed lists or patterns. One is genuinely free text, and it is placed where free text does least harm: in the middle, not at the start where it would break sorting, and not at the end where it would collide with the version.
The version component earns its place by preventing the most common workaround. Without it, the second version of a campaign gets named by appending something (_final, _v2, _NEW, _USE THIS ONE) and those appendages are where conventions visibly start to fail.
The taxonomy underneath: Read: what a data dictionary is — where the components and their permitted values are defined.
Best practices
Short, positional, delimited, and closed, with every component drawn from a fixed list.
- Carry only what you will filter on. Every component is a field someone types. If no report groups by it, it is costing entry time and buying nothing.
- Fix the order and never change it. Position is what makes a name machine-parseable. Reordering invalidates every name already created.
- Use one delimiter, and reserve it. Underscore between components, hyphen never inside one. A delimiter appearing inside a value destroys parsing.
- Close every component you can. Regions, brands, channels and periods are all knowable sets. Free text is a last resort, not a default.
- Fix case, and prefer lowercase. Case-sensitivity bites downstream in analytics tools that group by exact string.
- Keep it short enough to type from memory. Six components is workable. Ten is a document somebody consults, which means a document somebody stops consulting.
- Set a length limit and know what truncates. Downstream fields have limits, and truncation is silent.
General naming-convention guidance in software design converges on similar principles — readability over abbreviation, consistency over cleverness (Microsoft Learn, “General Naming Conventions”, accessed 2026-09-11). The marketing case adds one constraint those guidelines never have to handle: the person applying the rule often does not work for you.
Why conventions decay
The person naming the thing rarely pays for naming it wrong.
Every convention we inherit is correct in the document and wrong in about a third of the rows.
That incentive gap is the whole mechanism. A campaign manager launching under deadline gets no feedback from naming something imprecisely — the campaign runs either way. The cost lands weeks later, on an analyst who was not in the room and cannot reconstruct what was meant.
Everything else follows from it:
New people arrive continuously. The convention was explained at a kickoff attended by people who have since moved teams. Their replacements receive a document.
Agencies apply it from outside. A significant share of campaign names are typed by people in tools you do not administer, working from a PDF you emailed. Cross-agency governance and taxonomy compliance failures come up across 74 enterprise accounts in our customer conversations, the largest recurring source of naming drift.
Edge cases arrive without an owner. A campaign spans two regions. There is no rule, the person invents one, and the invention becomes precedent because nobody was available to decide. Every convention accumulates these, and the accumulation is invisible until someone counts distinct values and finds ninety where the document says eight.
The spreadsheet ages. Convention drift in tracking parameters is raised across 81 accounts, and it almost always traces to a reference document that is one version behind the practice.
Carhartt’s analytics team described the maintenance burden precisely.
“…it was always an uphill battle for people to have the latest Excel version, follow the rules and replicate that campaign after campaign.” — Andrew Laycock, Analytics Manager – Direct to Consumer, Carhartt
The standards layer: Read: what are data standards? — permitted values that travel with the workflow, not the document.
Documenting vs enforcing
A document asks for compliance; a form that rejects a bad value produces it.
| Documented convention | Enforced convention | |
|---|---|---|
| Lives in | A PDF, wiki or spreadsheet | The system where the name is created |
| Applied by | A person, from memory or by looking it up | The form, automatically |
| Reaches agencies? | Only if they read it | Yes, if they use the form |
| Handles a new joiner | Requires onboarding | Requires nothing |
| Failure mode | Silent drift, discovered in reporting | An error message, at entry |
| Cost of a violation | A permanent mapping table | Thirty seconds |
The documented column is not wrong — a convention still has to be written down, because the form has to be configured from something. The mistake is treating the document as the control rather than as the specification for one.
A simple test for which you have: if the convention changed tomorrow, how would a campaign manager in another market find out? If the answer involves anyone remembering to tell them, the convention is documented. If the answer is that the form would simply start offering different options, it is enforced.
The same logic governs tracking parameters, which are names by another route, and sits inside the broader marketing taxonomy that decides which components exist at all.
See standards enforcementThe convention applied where names are created.Explore Claravine Data StandardsFrequently asked questions
What is a name convention?
An agreed rule for how names are constructed.
What are some examples of naming conventions?
File naming, code naming, and campaign naming. The last is the hardest to enforce, because the people applying it are not the people who read the reports.
How do you create a naming convention?
Decide the components, fix their order and delimiter, then close each component to a list of allowed values.
Do SQL and database naming conventions apply here?
They are the same idea in a developer context. In campaign and marketing data the names are typed by many people rather than written once by an engineer.
Why does our convention keep breaking?
Because it is documented rather than enforced, and the people naming things are not the people reading the reports.
Sources
Outbound citations, named and dated:
- Harvard Medical School Data Management, “File Naming Conventions” (accessed 2026-09-11) — the definition and purpose of a naming convention in research data management.
- IT Glue, “Naming Conventions: Examples, Formats and Best Practices” (accessed 2026-09-11) — the structural element set.
- Microsoft Learn, “General Naming Conventions” (accessed 2026-09-11) — convention design principles from software design guidelines.
