Context
Anyone can ask a company what data it holds on them. Companies have a deadline to answer. This is a Data Subject Request (DSR). Usually, a data request form is available on a company’s website under its privacy page. Miss the deadline and the penalties are serious. I designed a system that automates most of the process.
A customer, employee or partner submits a request through the company’s privacy page.
The team hunts for the person’s data across every tool the company uses.
Each record is checked, then corrected, deleted, or kept for legal reasons.
They receive a summary of what was found and what was done.
Painpoints and goals3 Painpoints, 3 Goals, 1 System
Scanning them by hand is the biggest bottleneck.
1–1.5 months → ~1 week
Searches every source and matches the requester from their form.
Hand-offs between teams slow everything down.
Assignment, tracking and audit trail in one place.
Identity, financial, medical and more, ready to check, correct, delete or keep.
Each request type has different legal requirements.
GDPR CCPA + more
Applies the right rules per request, then generates the report.
A request that could take up to the full 30–45 day legal window now closes in under 5 days.
WorkflowTwo manual steps. The rest runs itself.
Initial design conceptsSchematics of layouts we tried before the final flow
Three zones: a header with the setup wizard, a configuration panel with a library of fields, and the form itself, built in sections.
The data discovery step, laid out around data sources so each one can be handed to its IT owner. Right for the hand-off, but the report is organised by instance, not source.
The same step, laid out around every instance found. Right for the report, but handing work to IT hides in a second tab.
Instances, data sources and validated items stacked with equal weight. It serves both jobs, so nothing leads, and it gets overwhelming at real volumes.
One block per stage, in order: verification, approval, discovery, then the report. Opening a block shows its rules: here, discovery erases data automatically where a source allows it, and only validates where erasure has to be done by hand.
Assumptions vs realityWhat the first designs got wrong
SolutionThe shipped flow, in four parts
Intake form, deployable on any website
One branded form per customer, embedded on their site.
- Build it by adding fields
- Preview it before it goes live
Requests list with bulk actions
Sort, filter and act on many requests at once.
- Approve or reject requests
- Share completed reports
- Close requests as complete
A single request, end-to-end
Every request moves through three steps:
- Verify identity, then approve or reject
- Review the data found, then delete, modify or keep it
- Share the report and close the request
Automation workflows with conditional logic
Set up once, so most requests run on their own.
- Choose what happens at each stage
- Branch on conditions like location or regulation
ImpactWhat it led to
ReflectionWhat I’d do differently
- Simplify conditional logic and automated workflows more. This flow does most of the heavy-lifting in the workflow. For customers interested in using it, the current designs are still a bit complex to understand, especially the way the if-then-else logic works.
- Design to fit into the customer eco-system. Initial designs didn’t have any integration like JIRA or ServiceNow where most customers track their tasks. Designing as if LightBeam was the only destination cost us some re-architecture and effort.
- Talk to varied set of customers. Only a handful of interviews captured only 2-3 working styles and customers’ workflows. Differing regulations and orgs with different sizes operate differently and we had to introduce a lot of flexibility in the design later on, which was not very easy to do.
- Champion for reduction in set-up time. As a startup a lot of the initial setup was done by the customer success team for almost all customers. This led to the overall onboarding aspect to be overlooked in my designs and I wish I had pushed back and championed for this aspect more.