Skip to content
Home Posts About us Privacy Policy
What a DPIA Is and When a Web Data Project Needs One

What a DPIA Is and When a Web Data Project Needs One

Share ๐• in f W

A data protection impact assessment, usually shortened to DPIA, is a structured review that the GDPR requires before an organization starts processing personal data in a way that is likely to result in a high risk to people. A common assumption in web data projects is that a DPIA belongs to hospitals, banks and surveillance systems, and that collecting information people chose to publish could never need one. That assumption breaks in a specific place: the GDPR looks at the risk created by what you do with the data, not at whether the data was visible to begin with. Large-scale collection, profiling and combining sources are all features that can push a scraping project into DPIA territory, even when every page involved was public.

What a DPIA is, and what it is not

A DPIA is a documented process, carried out before the processing begins, that describes the project, tests whether it is necessary and proportionate, identifies the risks to the people whose data is involved, and records the measures that reduce those risks. It is written by the controller, the organization that decides why and how the data is processed, and when the organization has a data protection officer, that officer's advice has to be sought.

A DPIA is not a permission slip and it is not filed anywhere by default. It stays with the organization as evidence of its reasoning. It also does not replace a lawful basis: a project that relies on legitimate interests still needs its own assessment of that basis, and the DPIA sits next to it. The two documents overlap, but they answer different questions. The legitimate interests assessment asks whether the processing can rely on that basis at all; the DPIA asks what could go wrong for people and what has been done about it.

The minimum content is fairly stable across guidance: a systematic description of the processing and its purposes, an assessment of necessity and proportionality, an assessment of the risks to people's rights and freedoms, and the measures planned to address them, including safeguards and security.

When a DPIA is required

The GDPR names three situations in which a DPIA is required: systematic and extensive evaluation of people based on automated processing, including profiling, that leads to decisions with legal or similarly significant effects; large-scale processing of special categories of data, such as health, political opinions or religious beliefs, or of data about criminal convictions; and systematic monitoring of a publicly accessible area on a large scale.

Those three are examples, not the whole rule. European guidance adds a set of criteria that signal high risk, and the general reading is that a project meeting two or more of them is likely to need a DPIA. The criteria most relevant to web data are:

  • Evaluation or scoring: ranking, rating or predicting something about individuals.
  • Large scale: judged by the number of people, the volume of data, the geographic reach and how long the processing lasts.
  • Matching or combining datasets: joining records from different sources in ways people would not expect.
  • Sensitive or highly personal data: including information that reveals sensitive traits by inference.
  • Vulnerable people: such as children or people in a weaker position than the controller.
  • Innovative use of technology: new methods whose effects on people are not yet well understood.

National supervisory authorities also publish their own lists of processing that always needs a DPIA, and those lists differ from country to country. The list of the authority that applies to your organization is the document to check first.

How the question varies for web data projects

The same scraper can produce very different answers depending on what it collects and what happens next. Three hypothetical projects show the range.

A team collecting product prices and stock levels from retailer pages touches almost no personal data. Seller names on a marketplace might appear, but nothing about the project evaluates people. A DPIA is very unlikely to be needed, though a short note recording why it was considered and rejected is good practice.

A research group collecting public posts on a political topic from several thousand accounts processes data that can reveal political opinions, a special category, and it may do so at scale. Two or more criteria are met, and a DPIA is the expected step.

A recruitment startup that gathers public professional profiles from several sites, merges them into one record per person and scores candidates for employers meets several criteria at once: combining datasets, evaluation, large scale, and people who were never told. This is the clearest case for a DPIA, and a careful one may conclude that parts of the plan should change.

What happens when the risk stays high

The point of the assessment is to change the project where needed. Typical measures in web data work include collecting fewer fields, dropping free-text fields that carry incidental personal details, pseudonymizing identifiers early in the pipeline, shortening retention, restricting who can query the raw data, and offering a clear way for people to object.

If, after those measures, the controller still finds a high risk that it cannot reduce, the GDPR requires it to consult the supervisory authority before starting the processing. In practice this outcome is uncommon, because most projects can be scaled back to a level of risk the organization can accept. Reaching it is a signal to reconsider the design rather than a box to tick.

A DPIA should also be reviewed when the project changes. Adding a new source, a new field or a new use of the output can move the assessment, and the version on file should describe the project as it actually runs. The article on ethical web scraping and data privacy standards covers the broader privacy questions that sit around this review.

How a DPIA relates to neighbouring documents

Several documents in a data project look alike, and confusing them leads to gaps.

  • Legitimate interests assessment: tests whether the lawful basis applies. It feeds the DPIA but does not cover risks such as security or misuse.
  • Records of processing: an inventory of processing activities that many organizations must keep. It describes; it does not assess.
  • Terms of service review: records what each website allows. It concerns the relationship with the site owner, not the rights of the people in the data.
  • Data specification: defines fields and quality targets. It is a useful input, because the DPIA needs to know exactly which fields are collected.

Outside the GDPR, similar ideas appear under other names. Several US state privacy laws require data protection assessments for certain activities, such as processing sensitive data, targeted advertising or some kinds of profiling. Scope, thresholds and timing differ between states and change over time, so the current text of each law, or a lawyer who works with it, is the reliable source.

Common mistakes with DPIAs in scraping projects

  • Skipping it because the data is public. Public availability affects people's expectations, which matters in the risk analysis, but it does not remove the requirement when the high-risk criteria are met.
  • Writing it after collection has started. The assessment is meant to shape the design. One written to justify a running project rarely changes anything.
  • Describing the purpose too broadly. "Market insight" cannot be tested for necessity. A purpose such as "weekly counts of public job postings by region and occupation" can.
  • Forgetting inference. A dataset without a religion field can still reveal religion through the groups people follow or the events they attend.
  • Assuming the provider carries it. When a company commissions collection from a service such as ScrapingIsNotACrime, the company deciding the purpose is normally the controller and remains responsible for the DPIA. The provider can supply facts about how collection works, which makes the description more accurate.
  • Treating it as a one-off. A DPIA that describes last year's sources and fields is not evidence of anything. Schedule a review whenever the scope changes, and confirm with a data protection lawyer when the result is borderline.
Share ๐• in f โœˆ