What is data mapping? A plain-English guide
Every time data moves from one place to another, someone has to decide which field goes where. That decision is data mapping. It sounds small, and for one file it is. It becomes a real job when you do it every week, for many sources, with columns that never quite match.
The short definition
Data mapping is the set of rules that says how fields in a source correspond to fields in a destination.
A simple example: your supplier sends a CSV with a column called cust_nm, and your CRM
expects customer_name. The mapping is the rule that connects the two.
| Source column | Destination field | Rule |
|---|---|---|
cust_nm | customer_name | Copy as is |
dob | date_of_birth | Convert DD/MM/YYYY to YYYY-MM-DD |
st | state | Look up the full state name |
amt | amount_cents | Multiply by 100 |
Most real mappings are a mix of these four kinds of rule: copy, reformat, look up, and calculate.
Where data mapping shows up
- Imports and onboarding. A new customer hands you their data in their format, and it needs to fit yours.
- Integrations. Two tools need to stay in sync, and each has its own names for the same things.
- Migrations. You are replacing a system and everything in the old one needs a home in the new one.
- Reporting. Data from several sources has to line up before the numbers mean anything.
Why it gets painful
Names do not match. email, e-mail, EmailAddress and contact_email are the
same thing to a person and four different things to a computer.
Values do not match either. One system stores CA, another California, a third
06. Getting the columns lined up is only half the work; the values need translating
too. The table that holds those translations is often called a crosswalk.
Sources change without warning. A vendor renames a column or adds a new one, and a mapping that worked last month quietly produces bad output.
The same work gets repeated. If you mapped cust_nm to customer_name once, you
should never have to do it by hand again. In practice, most people do, because the
knowledge lives in a spreadsheet or in someone's head.
What good data mapping looks like
- Start from the destination. Know exactly what the target needs, including which fields are required and what format each expects.
- Look at real values, not only column names. A column called
datecan hold almost anything. Sample rows tell you the truth. - Keep translation tables separate from the mapping. State codes, product categories and status values change on their own schedule.
- Keep the raw input. If you store the data exactly as it arrived, you can fix a mapping and re-run it instead of asking for the file again.
- Reuse what you have already decided. Every accepted mapping is a decision that should make the next one faster.
Where KlaroKit fits
We are building KlaroKit around that last point. You pick a source and a destination, KlaroKit suggests how the fields line up, and each suggestion you accept is remembered, so the next mapping starts closer to done.
KlaroKit is not available yet. If this is a problem you deal with, join the early-access list and you will get a launch discount when we open up.