Raw data: what it is, raw vs processed data, and when you need it

What is raw data?

Raw data (also called primary or source data) is data as it was originally recorded, before it has been cleaned, aggregated, filtered or transformed. In analytics, raw data means individual events: one row per pageview, click, sign-up or purchase, with its timestamp, identifiers and properties.

Everything you see in a report is processed data built from those rows: counts, rates, averages and charts. Raw data is what you go back to when a report cannot answer your question, or when you need to check how a number was produced.

Raw data vs processed data

Example: one question, two levels

A report shows 1,240 sign-ups last month and a 3.1% sign-up rate. Someone asks: "How many people signed up within ten minutes of their first visit?"

The report cannot answer that, because it does not store the time between each person's first visit and sign-up. The raw events can. Sample raw rows:

Visitor v_51 signed up 5.5 minutes after arriving; v_88 took two days and returned through a different source. Answering the question means grouping raw events by visitor and comparing timestamps, something only row-level data supports.

Why raw data matters

• Custom analysis. New definitions, joins with CRM or billing data, models and machine learning.
• Auditing and reconciliation. Checking a dashboard number against orders in your database.
• Ownership and portability. Keeping a copy in your own warehouse, independent of any tool.
• Debugging. Finding duplicate events, wrong properties or bot traffic that totals hide.

Working with raw data: common pitfalls

• It is messy. Expect duplicates, test events, bots, time-zone differences and missing properties. See data cleaning techniques.
• It is large. A busy site produces millions of rows a month; analyze with SQL or a notebook, not a spreadsheet.
• Definitions are yours now. If you recount "sessions" or "active users" yourself, your numbers may differ from the tool's because the rules differ. Write your definitions down; see data dictionary.
• It contains personal data more often than reports do. Identifiers, IP-derived fields and free-text properties need access control and retention limits. See keep personal data out of analytics.
• It expires. Raw rows are usually kept for a fixed retention period, after which only aggregates (or nothing) remain.

Raw data in TapCub

TapCub reports are aggregates built from raw events. When you need the rows, All-platform analytics › Data › Data export creates an export job for events, sessions, people, autocapture records or consent receipts. Files are gzip-compressed JSONL, a job can span up to 365 days, and finished files are kept for 48 hours. Export is available on paid plans; rows older than your retention period are no longer available. For ad hoc questions without exporting, the SQL query feature runs against the same event data; see query data with SQL and export raw data.

Sources

• Wikipedia. Raw data.
• Georgia State University Library. Raw data: finding and using it.
• TapCub documentation. Data export.