1. Why RPO and RTO must be defined per application
RPO and RTO are two measures. RPO (Recovery Point Objective) is the maximum amount of work you can lose, measured as the interval between the last valid copy and the moment of failure. RTO (Recovery Time Objective) is the maximum time within which the system must be back in service. Setting a single value for the whole business is a common mistake, because a production management system, an archive and an internal tool do not have the same needs. Each application should therefore be assessed on its own. Ask each owner to explain the reasoning behind their choice.
Every application has a different cost of downtime and of loss. That is why targets should be decided case by case, together with the people who use the system every day. The result is a cheaper plan, because you do not pay for protection beyond what is needed, and a more robust one, because the truly critical systems receive the protection they deserve. The result is also easier to explain to management.
2. Prerequisites: inventory, owners and the cost of downtime
Before setting a target you need three things. The first is an inventory of applications, with the data they handle and the systems they depend on. The second is a business owner for each application, that is, a person who knows the process and can decide how much work it is acceptable to lose. The third is a qualitative estimate of what an interruption means for the business: orders blocked, customers without a response, legal obligations not met.
With these elements you can group the applications into a few categories. The table below is a generic example: it must be adapted to your business and is not a technical reference. The indications are qualitative on purpose, because precise values depend on context and must be verified with the supplier and the owners. Remember that the class is decided by the impact on the business, not by the technical complexity of the system.
| Class | Generic example | Target tendency |
|---|---|---|
| Critical to production | System that governs orders or shifts | Almost no data loss and very limited downtime |
| Operational | Tool used every day for documents or cases | Limited loss and downtime that does not stop the business |
| Supporting | Internal tool that is not indispensable | More tolerant targets and less frequent copies |
| Archive | Historical documents consulted rarely | Long-term retention and non-urgent restore |
3. The five-step method
The method is simple and is repeated for each application. Start from questions to the business owner, not from the parameters of the backup software. Ask how much work can be lost without serious damage and how long the process can stay down before the business suffers. Use concrete examples, such as a typical working day or the closing of an accounting month, because abstract answers tend to be too optimistic.
Then compare the answers with what the infrastructure can do today. This is where the IT supplier comes in: if the target requires very frequent copies or replication, check that it is technically possible and what it costs. If there is a gap between wish and possibility, you have three options: invest, accept a compromise or lower the ambition. Choose explicitly and write it down, because the choice is a business decision.
- List the applications holding valuable data and assign each a business owner.
- Ask the owner how much work loss and how much downtime are acceptable, using concrete examples.
- Compare the answers with what the supplier can guarantee today, with evidence or documents in hand.
- Where there is a gap, choose whether to invest, accept a compromise or reduce the target, and note it.
- Have management approve the targets and review them when the system or process changes.
4. Common mistakes when setting targets
The first mistake is declaring very ambitious targets without knowing whether they are achievable. A near-zero RPO always sounds desirable, but it requires architectures and costs that must be assessed together. The second mistake is the opposite: targets set by the IT department without input from those who use the system, who then discover they can lose much more, or much less, than expected.
The third mistake is confusing copying with restoring. An RTO target only makes sense if it includes the time needed to recover data, reconfigure the system and verify that it works. The fourth mistake is never updating targets, even when the business changes processes or adds dependencies. A target written some time ago for a system that has since changed protects nobody.
- The same targets for all applications, with no distinction.
- Targets set without the business owner.
- An RTO that does not include verification and reconfiguration time.
- Targets never reviewed after changes to the system.
- Targets declared by the supplier without evidence of a restore.
5. How to check that the targets are realistic
A target only becomes credible once it has been verified. After setting the targets, plan a restore test for each application class. Measure the real time from the restore order to the system being ready for use, with a user checking that the data is correct. Compare the result with the written RTO. If the test is slower, the problem lies in the target or the infrastructure, and you must decide which of the two to change. Record the test with date, timings and result. Repeat the test after every significant change.
For the RPO, verification is more subtle. Check that the last valid copy really contains recent data, for example by searching for a record entered shortly before the test. Ask the supplier how the interval between copies is measured and how a failed copy is reported. If they cannot say, the written RPO is a hope, not a fact. Ask for the answer in writing, together with the date of the last test.
6. How to get started: one system at a time
The first step this week is to choose one critical system and define its targets with its business owner. Prepare three questions: how much work can be lost, how long it can stay down, and what concrete consequences a one-working-day outage would have. Write down the answers as they are, even if imprecise, and mark the points where there is no agreement.
Then ask your IT supplier whether the current infrastructure can meet those targets and how they would verify it. After the first system, repeat the method for the others, starting with the most critical. There is no need to complete everything at once: a partial document that is approved protects far more than a complete plan that stays in a folder. The first system becomes the model for the others.
Frequently Asked Questions
Are RPO and RTO the same thing?
No. RPO measures how much data you can lose, that is, how far back in time you can return with the last valid copy. RTO measures how long you can be without the system, that is, how long it takes to bring it back into service. A system can have a low RPO and a high RTO, or the reverse. This is why they should always be defined together and against separate criteria.
Who should decide the RPO and RTO targets?
The targets are a business decision, so the application owner must take part, together with management for choices that carry costs. The IT department and the supplier provide the technical side: they say what is possible and at what cost. A document approved by both sides is stronger than one written by only one.
Can I have an RTO of a few hours for all applications?
You can ask for it, but it is often not sustainable for every system. A uniform target leads either to overspending on systems that do not need it, or to under-protecting those that really do. Differentiating by application class is more effective. Set tight targets only where downtime stops the business, and leave the rest with more tolerant targets.
How often should a target be reviewed?
Whenever the system, the process it supports or its dependencies on other systems change. It should also be reviewed after an incident or a restore test that produced different results from those expected. Without changes, a periodic review, for example annually, confirms that the targets are still valid and that tests are still being carried out.
Can the supplier guarantee these targets?
The supplier can declare targets and commit to delivering them with an appropriate architecture, but the guarantee must be read carefully in the contract: what it covers, what exceptions apply and how it is demonstrated. Ask to see the results of the latest restore tests. A declaration without evidence is not sufficient for a serious decision.