Legal sources and digital issue spotting · Lesson 2

Identify data and parties in a service flow

7 min read

Lesson objectives

  • Map a data flow and distinguish who determines purposes and essential means from who processes on instructions, while recording missing facts.
Reading tools

Before assigning privacy obligations, define the unit of processing. A registration service can involve several operations: entering an email address, storing an appointment, sending a reminder, analyzing attendance, and deleting a record. Each has a purpose, data set, actors, and time boundary. Saying only that a supplier ‘processes data’ does not identify who makes the governing decision. Draw arrows among the person, interface, database, message supplier, and support desk. Label each arrow with the fields transferred and the purpose. Under the GDPR, personal data concerns an identified or identifiable individual, including indirect identification. An account number or device identifier may distinguish someone in context even without a name. Ask whether a field relates to an identifiable person and how it is used, rather than whether the team considers it sensitive. Final legal characterization still depends on the applicable text and actual facts.

In GDPR terminology, a controller determines the purposes and essential means of processing; a processor handles data on its behalf. The title on a contract does not by itself settle the real function. Ask who decided to send reminders, chose the categories of people and fields, set essential retention, and can issue binding instructions or stop the operation. A technical supplier may choose an implementation detail such as a storage system without controlling the core purpose. But if the supplier uses addresses for its own advertising, that use cannot simply be described as following the customer's reminder instructions. Where two parties jointly determine the central decisions for one operation, examine possible joint control in light of their arrangement and behavior. Data exchange or a shared tool alone does not establish it. Map the role by operation, not by attaching one permanent label to the entire company.

Use a five-column flow table: operation, data fields, decision about purpose and essential means, executing party, and evidence of the role. Evidence may be a campaign setting, contract clause, instruction log, or interface diagram. If the evidence conflicts, do not select only what favors the contract label. Record the conflict and seek the decision or document showing who controls the use. A wrong role map has practical costs: an individual's request may reach an entity unable to answer, or a processing agreement may omit the supplier's independent use. The map alone does not establish that the GDPR applies; perform the territorial test from the previous lesson first. The concepts can be used for comparative learning outside EU scope, but they must not be announced as binding local law where they do not apply.

After mapping a role, ask which evidence would overturn it. If a message supplier chooses campaign segments and sells analysis to its own customers, it is not merely following platform instructions for that use. If two partners jointly decide who can register and which data must be collected for a shared program, joint control may arise for registration specifically. Do not carry that conclusion into payment if a different party decides payment purposes. Break a flow down by recipient, purpose, and retention. A party can act as controller for one function and processor for another. This distinction should precede the notice and supplier agreement. In practice, send the map to counsel or a privacy officer with a column for missing evidence, not a declaration of compliance unsupported by the record. A dated map can be revisited when product behavior or a contract changes.

Lesson summary

Exercise 2 solution: the inputs are forty rows, a stable code, town, date, and a retained mapping. Row one covers the code: it can join to the mapping, so prevent analyst access and consider rotating or aggregating the code. Row two covers town and date: the combination may be rare in a small set, so test the need for that precision and reduce it. Row three covers the mapping: keep it separately under narrow permissions, while recognizing that this does not automatically anonymize the source file. The provisional result is that name deletion is insufficient and personal-data precautions are needed in the GDPR example. To call the revised file anonymous, assess the realistic means of identification available to relevant parties after changes; a count of forty is not a legal threshold. A complete answer has three distinct rows and a conditional sharing decision, not an assertion that removing names settles the matter.