TapCub is live — analytics, insights and live chat, free on one platform
Glossary

What is data conversion? Types, checks and a time zone example

Data conversion changes data from one format or structure to another. Seven types, the rows-in rule, a 12-order example where a correct time zone conversion moves the day, and six checks before you import.

Su He Reviewed by Ethan Brooks Also called data format conversion Last updated Nov 19, 2026 Glossary

Quick answer

Data conversion changes data from one format or structure to another, such as a file format, character encoding, unit, data type or time zone, without changing its meaning. It succeeds only when both sides agree on meaning: a conversion can write every row and still move numbers onto the wrong day.

Formula: rows in = rows converted + rows rejected

What data conversion changes, and what it does not

Data conversion changes data from one representation to another so that a different system, program or person can use it. The content is meant to stay the same; only its form changes. Turning a spreadsheet into a CSV file, a Latin-1 text file into UTF-8, pounds into kilograms, 03/04/2026 into 2026-04-03, or a text column into an integer column are all conversions. The Wikipedia entry on data conversion describes it as converting computer data from one format to another, and Microsoft's SQL Server Integration Services even ships a transformation with that exact name that changes a column's data type.

Three neighbouring terms are easy to confuse with it:

  • Data migration moves data from one place to another. Moving can need conversion, but it can also copy records as they are. Planning a whole move is covered in data migration plan.
  • Data integration combines data from several sources into one view and usually includes many conversions along the way; see data integration.
  • Data preparation and cleaning fix values that are wrong or missing. Conversion keeps wrong values wrong, just in a new format. For the fixing side, see data preparation and data cleaning techniques.

The practical lesson of the rest of this page: a conversion can be technically successful, with no error and every row written, and still be wrong, because the two sides never agreed what a field meant. The tool is rarely the cause. An undocumented convention usually is.

Common types of data conversion

Most conversions fall into seven kinds. Knowing which one you are doing tells you what to check.

TypeWhat changesExampleWhat usually goes wrong
File formatThe containerExcel to CSV, log file to JSONCommas inside values, line breaks inside cells, lost formatting
Character encodingHow characters become bytesWindows-1252 to UTF-8Accented letters turn into garbage characters
Data typeHow a value is storedText "007" to integer 7Leading zeros dropped, "N/A" cannot become a number
Unit and scaleThe measurement unitCents to dollars, seconds to minutesA factor of 100 or 60 applied twice or not at all
Date and timeFormat and time zoneLocal time to UTC, DD/MM to ISO 8601Dates shift by a day, day and month swapped
StructureThe shape of the recordsWide table to one row per eventDuplicate rows, lost rows, fields with no home
Value mappingCodes and labelsUS and USA to one country codeValues nobody mapped silently become blank

Two of these deserve extra attention in analytics work. Date and time conversions cause the most expensive mistakes, because a day boundary decides which day a number is counted on. Value mapping is where a data dictionary pays off: if both sides write down what each code means, the mapping table is a by-product rather than a guess.

The formula: nothing disappears without a record

A conversion has no formula in the sense of a ratio, but it has one rule that is easy to test:

rows in = rows converted + rows rejected

Every input row must end up in exactly one of the two groups, and rejected rows must be listed somewhere you can read. If 100,000 rows go in and 99,412 come out with no list of the 588 others, you do not have a conversion, you have a loss. Row counts are the cheapest check there is, and they catch truncated files, rows dropped by a failed type cast and duplicates created by a join.

For values, add a second check: pick 10 or 20 rows by hand, convert them yourself, and compare with the tool's output. Counts prove nothing was lost; the hand check proves nothing was changed.

Worked example: a time zone conversion that moves the day

Sample data: an online shop exports 12 orders from a system that writes local time without any time zone marker. The shop is on US Pacific time, which in January is UTC−08:00. The analytics tool expects UTC. Here are the 12 order times and the same moments in UTC (computed in Python with the IANA time zone database):

Local time (written in the export)Same moment in UTC
2026-01-14 09:102026-01-14 17:10
2026-01-14 13:452026-01-14 21:45
2026-01-14 17:202026-01-15 01:20
2026-01-14 19:052026-01-15 03:05
2026-01-14 20:402026-01-15 04:40
2026-01-14 22:152026-01-15 06:15
2026-01-14 23:302026-01-15 07:30
2026-01-15 08:002026-01-15 16:00
2026-01-15 12:302026-01-15 20:30
2026-01-15 16:102026-01-16 00:10
2026-01-15 18:552026-01-16 02:55
2026-01-15 21:202026-01-16 05:20

Count the orders per calendar day in each column:

DayOrders by local timeOrders by UTC
2026-01-1472
2026-01-1557
2026-01-1603
Total1212

The total is the same, so the row-count check passes. But 14 January falls from 7 orders to 2, a drop of 71.4%, and a new day appears. Nothing is "wrong" with either column; they simply answer different questions. The evening shoppers of the Pacific coast are, in UTC, already in tomorrow.

Daylight saving time adds a second trap. In 2026 US clocks moved forward on Sunday 8 March. A local 01:30 that day is UTC−08:00, while 03:30 the same morning is UTC−07:00, and times between 02:00 and 03:00 do not exist. A fixed "minus eight hours" rule is therefore right in January and wrong in July. The safe form is a timestamp that carries its own offset. RFC 3339 defines one: 2026-01-14T09:10:00-08:00, or the same instant as 2026-01-14T17:10:00Z.

The rule that follows: before converting a time, write down the zone it was recorded in, and decide which zone the reports will use to cut days.

Six checks before you import converted data

  1. Name the source and the target. Which system wrote the file, which one reads it, and who owns each. Conversion errors are mostly misunderstandings between two owners.
  2. Fix the encoding and the delimiter. Save as UTF-8, quote values that contain commas, and open the file in a plain text editor, not only in a spreadsheet, which can hide problems.
  3. State the time zone of every date column. Prefer ISO 8601 with an offset or UTC. If a column has no zone, write down what you believe it is and test it with a few rows where you know the real time.
  4. Convert units and types explicitly. Keep identifiers such as phone numbers and order numbers as text so leading zeros survive; convert money to one currency and one unit, such as cents or dollars, never both.
  5. Map codes with a written table. List every distinct value in the source column, give each a target value, and decide what happens to values that are not in the table. "Blank" should be a decision, not an accident.
  6. Reconcile counts and totals. Rows in equals rows converted plus rows rejected. Compare a total, such as revenue or orders, per day before and after, and look at the first and last day of the range most closely, because boundary days are where time zone mistakes appear.

Common mistakes

  • Trusting a spreadsheet to convert for you. Opening a CSV in a spreadsheet and saving it can reinterpret dates, strip leading zeros and turn long numbers into scientific notation. Convert with a script or the destination tool, and keep the original file untouched.
  • Converting twice. A file that was already converted to UTC gets converted again by a second tool. Keep the original and one documented chain of steps.
  • Mixing DD/MM and MM/DD. 03/04/2026 is 3 April in one country and 4 March in another. Any date where both numbers are 12 or less is ambiguous, so check a row where the day is above 12.
  • Calling a migration "done" because there were no errors. An import that never complains still needs the checks above; a missing mapping is not an error.
  • Treating conversion as cleaning. Duplicates, typos and impossible values pass through a conversion unchanged. Clean before or after, but do it on purpose.

How to see conversion problems in TapCub

Suppose you are importing the exported orders from the example into TapCub as history. The Tracking plan page (All-platform analytics › Data › Tracking plan) is where you write down what each event and property means before you send anything; the import itself is on All-platform analytics › Data › Data import. TapCub does not convert data for you beyond the mapping step; it reports what it could and could not read.

  1. Choose the file type and keep each file under 100,000 rows. Data import accepts CSV events, CSV user attributes and Nginx or Apache logs, one file at most 100,000 rows. Split larger exports before uploading.
  2. Map the columns in the preview. The preview shows the first 20 rows and guesses a target field from the column name. Map time, event name and the user identifier, and set properties for any column you want to keep; unmapped columns are ignored. Times can be 10-digit or 13-digit timestamps, ISO format or common date formats. If you map no time column, TapCub uses the moment of import; rows whose time cannot be parsed count as failed. If you export with an explicit offset or in UTC, there is nothing to guess.
  3. Know what happens without a user identifier. If no user column is mapped, each row becomes its own anonymous visitor and rows cannot be joined into one person. Events from the same user within 30 minutes count as one session.
  4. Check the day boundary against the project time zone. Daily, weekly and monthly reports are cut in the project's time zone, set on Configuration › Settings › General. In the example above, compare the per-day order counts in the report with your 7 / 5 / 0 and 2 / 7 / 3 columns to see which convention you actually imported.
  5. Read the result and filter by platform. The result shows total, imported and failed rows. Imported events carry the platform import, so you can filter them in or out of a report. If you later send history through the server interface, an event time more than 72 hours away from the server's clock is replaced with the receive time, so backfill old events through Data import instead.

TapCub does not automatically correct time zones or unit mistakes in your file. It makes the failures countable, and the Tracking plan gives the conversion rules a written home.

Sources

Was this explanation clear?

We use your answers to fix and expand this page.

Su He

Data Analyst

Explores traffic sources, conversion funnels, and retention, using clear metric definitions and concrete examples to explain what the numbers mean.

FAQ

Questions people ask

Short answers to the follow-up questions that come up most.

Still have a question?

Chat with the team behind TapCub — we usually reply within a few hours on business days.

What is data conversion in simple terms?

Data conversion is changing data from one form to another so another system can use it, for example a CSV file into UTF-8, cents into dollars, or local time into UTC. The meaning should stay the same. Only the representation changes.

What is the difference between data conversion and data migration?

Migration moves data from one place to another and may copy it unchanged. Conversion changes the form of the data. A migration often includes conversions, but you can convert without moving anything, and move without converting.

What are the main types of data conversion?

Seven common kinds: file format, character encoding, data type, unit and scale, date and time, structure, and value mapping. Date and time conversions cause the costliest errors because they decide which day a number is counted on.

How do you check that a data conversion worked?

Rows in must equal rows converted plus rows rejected, with the rejected rows listed. Then compare a daily total before and after, focusing on the first and last day of the range, and convert 10 or 20 rows by hand to check the values themselves.

See this number on your own site

Add one script tag for web and apps, and the first reports fill in within minutes.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics