
Privacy tends to enter a software project late, as a checkbox on a security questionnaire or a sudden question from a large customer: "Where is our data held, who can access it, and how do you delete it?" By then the database is full, ex-employees' records are mixed with current ones, a third-party messaging tool has copies of customer phone numbers, and no one can answer with confidence.
Handled early, privacy is far less daunting than it sounds. Most of it comes down to sensible design: know what personal data you hold, collect what you need, let only the right people see it, keep it only as long as necessary, and be able to act on it when asked. This guide turns those ideas into concrete decisions for a custom business application.
A note on scope: this is general, practical information about software design, not legal advice. Laws differ by country, sector and circumstance, and they change. Have a qualified adviser confirm what applies to your organisation.
Privacy and security are related, but not the same
Security asks: can unauthorised people get to this data? Privacy asks: should we have this data at all, are we using it fairly, and who is entitled to use it? A system can be very secure and still be poor on privacy. If an app records the location of every technician around the clock for no clear reason and keeps it for ten years, encrypting that data does not make the collection reasonable. Our guide to security basics for business apps covers the protective controls. This one concentrates on the decisions about the data itself.
Which rules might apply?
You rarely need to become an expert in privacy law to design a good system, but you should know which rules your adviser may want to consider. Applicability often depends on where the affected people are, not only on where your company is registered.
| Framework | Broadly covers | A point worth knowing |
|---|---|---|
| GDPR (EU) and UK GDPR | Processing of personal data in the EU/UK, and certain organisations elsewhere that offer goods or services to people there or monitor their behaviour | Reaches beyond Europe; distinguishes controllers from processors |
| DPDP Act, 2023 (India) | Digital personal data processed in India, and certain processing abroad linked to offering goods or services to people in India | Uses "data fiduciary" and "data principal"; its implementing rules are being phased in, so check current dates |
| US state laws, such as California's CCPA/CPRA | Residents of the relevant state, for businesses meeting the law's thresholds | Scope varies by state; sector laws such as HIPAA may also apply |
The frameworks differ, but they overlap on several expectations: be transparent, limit what you collect and how you use it, protect it, keep it no longer than needed, respect individuals' rights and respond properly to breaches. Designing for these principles gives you a sound base, with local legal requirements layered on top.
Clarify your role as well. If you decide why and how data is used, as with your own employees and customers, you are usually the controller or fiduciary. If you run a SaaS platform and process data on your customers' instructions, you may be their processor, with different duties. Many software companies hold both roles at once.
Step 1: Build a data map
You cannot protect or delete what you have not found. Before designing screens, create a simple inventory of personal data. Here is an illustrative map for a field service application:
| Data | Whose | Purpose | Who needs it | Stored or sent to | Keep for |
|---|---|---|---|---|---|
| Name, phone, site address | Customer contact | Schedule and carry out jobs | Dispatch, assigned technician | Database, mobile devices, SMS provider | Contract period plus defined period |
| Job photos | May show people or homes | Proof of work | Technician, supervisor | File storage | Warranty period |
| Signature | Customer | Acceptance of work | Billing, support | File storage, PDF job sheet | As for invoice records |
| GPS location | Technician | Check-in and nearby job allocation | Dispatcher | Database, maps service | Short period, then summarised |
| ID and bank details | Employee | HR and payroll | HR, payroll only | HR module | As required by employment and tax law |
The map usually reveals surprises: phone numbers in free-text notes, identity cards attached to tickets, customer exports sitting in shared drives, personal data in log files. It also makes a practical working document. When a customer asks where their data goes, the answer is already written down. Give the map an owner and update it whenever a feature adds data or a vendor.
Step 2: Challenge each field
The cheapest data to protect is data you never collected. For each field in a form, ask what would break if it were removed. Many forms ask for date of birth when an "over 18" confirmation would do, or for a full address when a city is enough for the task. Do not scan an identity document when recording that it was checked would suffice. Free-text fields deserve special care, since people will type anything into a box labelled "Notes", including health details and card numbers. Offer structured choices where possible, and add a short reminder near the box.
Also be cautious about re-using data for a new purpose. A phone number collected for a service appointment isn't automatically available for marketing. Whether you rely on consent or another lawful ground depends on the law and context, so ask your adviser. The software should then record the relevant choice, with its date and source, and let people withdraw it as easily as they gave it.
Step 3: Limit who sees what
Most privacy incidents in business systems are not movie-style hacks. They are everyday over-exposure: a support agent browsing records out of curiosity, or a salesperson exporting the entire customer list before leaving for a competitor. Design for these situations:
- Least privilege. Grant access by role and scope. A technician needs today's assigned job, not the whole customer database.
- Show only what's needed. Mask sensitive values by default, such as showing the last four digits of a bank account, and reveal the full value only to a permitted role, with the action logged.
- Control exports and bulk actions. Treat "Export to Excel" as a permission of its own, record it, and include it in your reviews. Exports are where carefully controlled data escapes into uncontrolled files.
- Audit sensitive reads, not just changes. Knowing who viewed a payroll record can matter as much as knowing who edited it.
- Remove access quickly. Connect leaving-employee steps to account deactivation, and review access periodically.
Mobile devices need separate attention. Cached customer lists remain on a phone that may be lost. Keep offline data to what is needed for upcoming work, protect it with device and app-level controls, and clear it when no longer required. For personal phones, decide what happens when an employee leaves.
Step 4: Don't forget the hidden copies
Teams often secure the main database while personal data accumulates in less obvious places:
- Test and staging systems populated with a copy of production. Use generated or properly masked data instead.
- Application logs and error trackers that record full requests, including names, tokens and form values. Mask them at the source.
- Analytics tools that receive customer details through page addresses or custom events.
- Email and chat tools, where exported reports and screenshots are shared. Prefer links with access control.
- AI services used to summarise or classify records. Check the data use terms, retention and location, and send only what's needed.
- Backups, which are covered by your retention plan as well.
Step 5: Decide when data dies
Systems are good at remembering and terrible at forgetting. Unless someone designs for deletion, data stays forever, increasing both your risk and your workload. Set a retention schedule with your adviser, then have the application enforce it through scheduled jobs. Tax, employment and sector rules may require you to keep certain records for specified periods, so this is not simply a case of deleting everything early.
Deletion is more difficult than it sounds. Removing a customer may break the invoice that refers to them, and an invoice may have to be retained. A practical design chooses among three actions per data type:
| Action | What happens | Good for |
|---|---|---|
| Delete | The data is removed completely | Old leads, expired location trails, unneeded uploads |
| Anonymise | Identifying details are removed so that the remaining record can no longer reasonably identify the person | Historic job statistics, aggregated reports |
| Restrict | Data is retained for a required reason but hidden from everyday use | Invoices or contracts within a statutory retention period |
Note that merely replacing a name can fall short of true anonymisation if a unique address or job description still points to a person. And plan for backups: keep them on a defined, documented expiry cycle, and ensure that restoring a backup does not silently reinstate records you deleted. One way is to re-apply a stored list of deletions after any restore.
Step 6: Be able to answer people's requests
Depending on the applicable law, individuals may have rights to see their data, correct it, have it erased, withdraw consent or nominate someone to act for them. Requests are straightforward with the right tools, and painful without them. Build these basics:
- A search that finds a person across modules, including linked files and communications.
- An export that produces a readable copy of their data.
- Correction with an audit trail, and deletion or restriction following your retention rules.
- An internal request log with due dates, identity-check notes and the outcome.
- Care around third parties, because a person's record may also contain information about other people.
A hard case: tracking employees
Field service, delivery and sales apps often want location data, and it illustrates how privacy decisions actually work. Compare two designs for a team of technicians:
| Decision | Careless design | Proportionate design |
|---|---|---|
| When is location collected? | All the time, including after work | During working time, or at check-in and check-out only |
| How precisely and how often? | Every few seconds | Only as often as dispatching requires |
| Who can see it? | All office staff | Dispatcher and relevant supervisor |
| How long is it kept? | Indefinitely | Detailed trail for a short period, with job check-ins retained as records |
| Are people told? | Buried in a policy, if at all | Clear notice and a visible indicator in the app |
The second design still solves the business problem: dispatchers can allocate the nearest technician and customers get proof of attendance. It also leaves a much smaller store of sensitive data and is easier to defend to staff, whose trust directly affects whether they use the app. Check employment and privacy rules in each country before finalising such a feature, as they can vary significantly.
Vendors and data crossing borders
Each outside service in your data map is a part of your privacy posture: cloud hosting, SMS and email providers, maps, analytics, support tools and AI services. For each one, record what personal data it receives, why, where it is processed, and which contract terms cover it. Send a vendor only what it needs. An SMS provider requires a number and a message, not the customer's entire record. Message wording matters too, since text like "your oncology appointment" reveals more than "your appointment".
Cross-border transfers need explicit checking. Some laws restrict or attach conditions to moving personal data abroad, and some customers set contractual requirements about hosting location. Remember that remote access by support staff in another country may count as a transfer. Your architecture should let you state clearly where production data, backups and support access are located.
When something goes wrong
Breach readiness is mostly preparation. Some laws impose tight timelines: under GDPR, for example, certain breaches must be reported to the regulator within 72 hours of the controller becoming aware of them. That leaves no time to work out your process during the incident. Prepare the following in advance:
- Named people responsible for technical response, management decisions and legal advice.
- Logs that show what was accessed, by whom and when, retained long enough to investigate.
- A way to quickly list the affected records and individuals, using your data map.
- The ability to lock accounts, rotate keys and close an exposed feature.
- Contact details for your advisers, key vendors and insurer, stored outside the affected system.
- A short rehearsal, using a realistic scenario such as a lost phone or a misdirected export.
A checklist for your software project
| Project stage | Privacy question | Done? |
|---|---|---|
| Requirements | Is there a data map, with a purpose for each personal data field? | |
| Design | Are roles, data masking, export permissions and retention rules specified? | |
| Vendors | Is each third-party service listed, with its data, location and terms? | |
| Build | Are logs clean, and is test data generated or masked? | |
| Testing | Have cross-user and cross-customer access been tested and blocked? | |
| Launch | Are privacy notices reviewed, and old data migrated only where still needed? | |
| Operation | Are access reviews, retention jobs, requests and incident rehearsals scheduled? |
Include these points in your requirements document from the start. Adding a retention rule to a new table takes little effort; finding and cleaning five years of personal data scattered across a live system and its integrations takes a project of its own. A business that can say exactly what data it holds, why it holds it and when it will be gone is in a stronger position with regulators, customers and its own staff — whatever the next law or security questionnaire requires.
Frequently asked questions
Does data privacy law apply to B2B software?
Often, yes. Privacy laws usually protect individuals rather than companies, and B2B systems hold names, phone numbers and email addresses of contacts as well as employee data. How a particular law applies varies, so ask a qualified adviser about your own situation.
Is encrypting the database enough to protect privacy?
No. Encryption is an important security control, but privacy also covers why data is collected, who is allowed to use it, how long it is kept and whether people's rights can be honoured. A properly encrypted database can still hold far too much data for too long.
Can we use a copy of production data for testing?
It is best avoided. Test environments often have weaker access controls and more users. Use generated data or a properly masked copy, and if real data is ever necessary, restrict and record its use and remove it as soon as possible.
Do we have to delete personal data from backups immediately?
Requirements and regulators' expectations vary. A common approach is to delete data from live systems, let backups expire on a short documented schedule, and make sure deleted records are not reinstated if a backup is restored. Confirm the approach with your legal adviser.
Does software built by a development company make us compliant automatically?
No software makes a business compliant on its own. Software can support privacy through features like access controls, retention rules and export tools, but compliance also depends on your policies, notices, contracts, staff behaviour and legal basis for using data.
