Tags vs Fields in a Contact Database: Which Is Which
Tags vs fields in a contact database: one value per contact belongs in a field, zero or many belongs in a tag. Here is the rule and how to apply it cleanly.
Tags vs fields comes down to one rule: a field holds a fact every contact has exactly one value of, and a tag holds a fact that only some contacts have and that a contact can have several of. Confusing the two is the root of most messy bases.
The rule fits in a sentence, but applying it takes some care. This post gives the rule, the tests that make it easy, and the tag habits that keep a vocabulary trustworthy.
Tags vs fields in a contact database: the one rule
Most contact systems offer two ways to record a fact: a structured field with defined values, and a free-form tag. The rule is short. One value per contact goes in a field. Zero or many goes in a tag.
| Question to ask | If yes | If no |
|---|---|---|
| Does every contact have exactly one value of this? | Field | Ask the next question |
| Do some contacts have it and others not? | Tag | Probably a field |
| Can one contact have several values at once? | Tag | Probably a field |
The guide's own examples help. Country, lifecycle stage, original source and type of organisation are fields, because each contact has one. Attended an event, interested in a topic and part of a pilot are tags, because some contacts have them, others do not, and a contact can carry several.
The classic error: a field built out of tags
If you are creating tags called germany, france and spain, you have built a country field out of tags. It will drift out of sync with the real one within a month. Someone moves, someone is corrected in one place and not the other, and now two records disagree about the same person.
The same thing happens with stage and source. As tags they multiply, drift and contradict the real field. The symptoms are easy to spot.
- Two tags that cannot both be true for one person, such as two different countries.
- A tag whose name is a value of a field you already have.
- A segment that gives a different count depending on whether you use the field or the tag.
The repair is to convert: put the value in the field, remove the tags, and check what depended on them first. The check matters, and it is covered in segmentation mistakes founders make under renaming.
Fields give you something a tag cannot
A field has defined values. That means a stage can only be one of the stages you chose, and a country can only be a country. A tag is free-form, which makes it quick to create and easy to misspell.
Because a field has exactly one value per contact, it also gives you something to count against. Every contact is in the base once, in one stage, from one source. Add the counts for each value and they equal the base. Tags make no such promise, and that is by design.
Tags record facts, never feelings
The most corrupting tag encodes somebody's opinion at a moment: hot, promising, follow up. It is true on the day it is applied and false soon after. Nothing removes it, and two people apply it to different standards.
| Avoid (a feeling) | Prefer (a fact that stays true) |
|---|---|
| Hot | Requested a quote |
| Promising | Attended the onboarding session |
| Follow up | Came from the partner referral |
If a judgement matters, express it as a rule over facts, such as a segment or a score, so it updates itself when the facts change. This is also why fit and behaviour are kept apart, as explained in how to segment by fit and behaviour.
A naming shape for the tags you do keep
Without a convention, a vocabulary grows into near-duplicates: webinar, webinar attendee, attended-webinar. Every segment built on one of them silently misses the others. Choose one shape for everything.
A simple pattern is a category, a separator, then the value. The guide's categories are source, interest, event and programme. Written with a colon, a tag for an event would begin with the word event, then the specific event after it.
- It sorts related tags together.
- It makes each tag's purpose obvious.
- It makes a stray tag that fits no category visible as a problem.
Keep it lower case, add no dates unless the date is the point, and use no abbreviations that only one person understands.
Keep the vocabulary small, described and owned
A scheme survives when three things are true.
- Small. Every tag has to earn its place.
- Described. Each tag has a one line description of when to apply it, written where everyone who tags will see it.
- Owned. Each category has someone who decides whether a new tag is warranted.
It takes seconds per tag and saves the afternoon, later, of working out what forty undocumented tags were for. The test for a new tag is simple: will anything ever be sent, filtered, scored or counted differently because of it. If not, it is noise.
What to do with the tags you already have
If your base already holds a pile of tags, a single pass usually sorts them into four piles.
| Pile | Example of the problem | Action |
|---|---|---|
| Fields in disguise | One tag per country | Move the value to the field, then retire the tags |
| Feelings | Hot, promising | Replace with a fact or a rule |
| Near-duplicates | Three spellings of the same event | Merge into one in the agreed shape |
| Strays | Almost no contacts, no description | Describe or remove |
Before you merge, rename or delete anything, find out what depends on it. Scoring rules, sequence triggers and saved segments keep running, report no error and match nobody when a tag they rely on has changed. A tidy minded person renames a tag and three automations go dark.
Where tags earn their keep
Used well, a tag does more than label. The most valuable tags start things: a contact is tagged because they requested the guide, and the follow-up sequence begins. Or the next message in a sequence depends on whether the contact now carries a tag. That is where tagging stops being administration and becomes a system that responds faster than any person could.
This only works if the tag means one thing. A tag with five spellings starts the follow-up for a fifth of the people it should.
Reviewing the scheme monthly
The vocabulary will drift unless someone looks. The monthly pass takes thirty minutes and has four tag checks.
- Tags with almost no contacts, usually typos or experiments to merge or remove.
- Tags nobody can explain, to be described or retired.
- What scores, triggers or segments depend on a tag before it is renamed or deleted.
- Tags in use that feed nothing, in case a rename broke a rule.
The first time you do this, expect the findings to be worse than you hoped. That is normal, and it is the reason to do it monthly.
How fields and tags work together in one segment
The two are not rivals. A good segment usually leans on a field for its backbone and a tag for its detail. Take the guide's own example: leads tagged as interested in the advanced offer who visited the pricing page in the last two weeks. The interest is a tag, since some contacts have it and others do not. The lead status and the recent visit are facts about stage and behaviour.
Because the field has one clean value, the rule behaves predictably. Because the tag is a fact, the rule stays true. If either half is replaced by an opinion or by a duplicate, the count looks confident and means very little.
Questions to ask before creating any new tag
Pause for a moment each time someone proposes a tag. The pause costs seconds and spares the later clean up.
- Is this a property every contact has exactly one value of? If so, it is a field.
- Is it a fact that will stay true, or an opinion that will expire?
- Does it fit one of the existing categories, and does the name follow the agreed shape?
- Is there a one line description of when to apply it?
- Will anything be sent, filtered, scored or counted differently because of it?
- Who owns this category and approves the addition?
If the answers to the last two are no and nobody can name an owner, do not create the tag. A short vocabulary that everyone understands is worth more than a long one that captures every passing thought.
Frequently asked questions
Can a tag ever hold a stage or a country?
It can, but it should not. Those are one value per contact, so they belong in a field. As tags they multiply and drift until they contradict the real field.
Who decides whether to add a new tag?
One named person per category. Without an owner, every exception becomes a new tag, and the vocabulary grows without control.
How do I know a tag is worth keeping?
Ask whether anything will ever be sent, filtered, scored or counted differently because of it. If the answer is no, it is noise, and it will cost you in the next clean up.
Where does lifecycle stage go?
In a field, with exactly one value per contact. It is the backbone that every other segment leans on, so it deserves its own treatment in how to define lifecycle stages.
The full method, with the checklists, is in The Segmentation Playbook, a free 35-page playbook.