Quick answer
Data collection is the planned, systematic gathering of information about the variables you care about, in order to answer a question. Before collecting, decide what decision each field changes, how it will be recorded and on what basis.
What data collection means
Data collection is the planned, systematic gathering of information about the variables you care about, so that you can answer a question. The data glossary of NNLM (the Network of the National Library of Medicine) describes it as the planned, systematic gathering of information on variables of interest, with methods that run from manual to automated: surveys, interviews, instrument readings, recordings and records extracted from systems (NNLM).
Three words in that definition matter:
- Planned. You decide what to collect before you collect it. Collecting "everything, in case" is storage, not collection.
- Systematic. The same thing is recorded the same way each time, so the values can be compared.
- Variables of interest. Each item exists because some question needs it.
On a website this usually means the script and server calls that record visits, clicks and conversions; in a business it also covers forms, surveys, support conversations and imports from other systems.
Primary, secondary and first-party data
| Kind | What it is | Examples | Main risk |
|---|---|---|---|
| Primary | You gather it yourself for this question | Your own tracking events, a survey you send | Cost and time; design mistakes are yours |
| Secondary | Someone else gathered it earlier | Public statistics, a purchased panel, a partner's report | May not match your definitions |
| First-party | Collected by you from your own users | Site analytics, account data, chat transcripts | Consent and security are on you |
| Third-party | Collected by others across sites | Ad-network audiences | Little control over quality or lawful basis |
| Zero-party | Volunteered on purpose | A preference you choose in a form | Small volume; self-reported bias |
The same dataset can be primary and first-party at once. Before you collect, check whether secondary data already answers the question; if it does, you may not need to collect at all.
The five questions to ask before collecting
- What decision will this change? If you cannot name one, do not collect it.
- Which exact fields? Write each with its type, the allowed values and an example.
- How will it be recorded? Browser script, server call, form, import or survey; each has different gaps.
- On what basis and with what notice? Under the EU's GDPR, personal data must be collected for specified, explicit and legitimate purposes and limited to what is necessary for them (GDPR Article 5). Where consent is the basis, the controller must be able to demonstrate that the person consented (GDPR Article 7). This is a summary, not legal advice.
- How long will you keep it? Article 5 also limits keeping data in identifiable form to as long as necessary. Set the retention period when you plan the collection, not when the storage bill arrives.
A plan row can be as short as this, one per field:
| Field | Type | Allowed values | Decision it changes | Owner | Keep for |
|---|---|---|---|---|---|
| referral_source | text | utm_source value, or "direct" | Which channel to fund | Growth lead | 13 months |
If a row cannot be filled in, the field is not ready to be collected.
A worked example: cutting a signup form from 14 fields to 8
Sample data: a team plans a newsletter signup form and lists 14 data points. They write the decision each one would change.
| Data point | Decision it changes | Keep? |
|---|---|---|
| Where to send | Yes | |
| First name | Greeting in the message | Yes |
| Country | Language and send time | Yes |
| Referral source | Which channel to fund | Yes |
| Topics of interest | Which content to send | Yes |
| Company size | Which case studies to include | Yes |
| Consent timestamp | Proof of permission | Yes |
| Sign-up page | Which page converts | Yes |
| Last name, phone, birth date, gender, postcode, job title | None named | No |
Check: 8 kept (email, first name, country, referral source, topics, company size, consent timestamp, sign-up page) and 6 dropped, so 14 − 8 = 6 and 6 ÷ 14 = 43% fewer fields.
Each dropped field is one fewer item to secure, export on request or delete. The decision rule: if no one can name the decision a field changes, remove it from the form, not from the privacy notice.
A second effect to plan for is coverage. If 10,000 visits happen and 62% of visitors agree to measurement, the system records 10,000 × 62% = 6,200. If it counts 186 conversions among them, the observed rate is 186 ÷ 6,200 = 3.0%. That figure describes the consenting visitors, and it is only as good an estimate of everyone as the two groups are alike. Report the coverage next to the rate.
Common mistakes
- Collecting first, deciding later. Unused fields still need protection and can still be requested by the person they describe.
- Recording the same thing under different names or formats. Values that cannot be compared are not systematic. See tracking plan.
- Ignoring the gap between collected and complete. Ad blockers, consent choices and bots all change what the script sees.
- Mixing test and real data. Internal visits and bots inflate counts unless filtered.
- Treating raw data as the product. Raw data needs cleaning and definitions before it supports a claim.
- Leaving the definition unwritten. "Signup" can mean a form view, a submit or a confirmed email. Write which one, or two teams will report different numbers from the same data.
- Keeping data indefinitely by default. Set a retention period per dataset.
For the principle behind the shortlist above, see data minimization.
Planning collection in TapCub
Use the signup example and put the plan into TapCub.
- Write the plan first. All-platform analytics › Data › Tracking plan records the events you intend to collect, including ones not sent yet, with a trigger, properties (name, type, required, example), owner and status from Proposed to Live. Add
newsletter_signupand its properties before the developer ships. - Choose how data arrives. The website script, mobile or mini program SDKs, server calls and imports share one project. Autocapture is limited on purpose: form submission records only the form id, name and action, never what was typed.
- Set the privacy level. Configuration › Privacy center has collection tiers. Tier 1 is anonymous summary data with no persistent ID; the default is tier 2. For the EEA and UK the default rule requires consent, and without analytics consent collection drops to tier 1, where page views and custom events are still recorded. In China mode nothing is collected before consent. Retention and data subject requests (export and delete) are in the same place. This is configuration, not legal advice.
- Verify what arrived. Turn on live verification in the tracking plan to see each event's hits in the last 24 hours and the completeness of required properties, and check All-platform analytics › Data › Data quality for rejects and warnings.
- Keep bots out. Known bots and AI crawlers are counted apart, so the rates you read describe people.
TapCub does not decide what you need to collect or whether you have a lawful basis; it records what you plan, and shows how complete it is.
Sources
- NNLM (Network of the National Library of Medicine), Data glossary: Data Collection.
- GDPR, Article 5: Principles relating to processing of personal data.
- GDPR, Article 7: Conditions for consent.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Technical Engineer
Focuses on tracking code, SDKs, and API integrations, with practical guidance on implementation checks and troubleshooting.