{"canonical_url":"https://engineer.company/glossary/recovery-point-objective/","content_markdown":"# Recovery point objective\n\n## What it is\n\nA recovery point objective, usually shortened to RPO, is the limit on how\nfar back in time a service can go when restoring data after a disruption.\nAn objective of fifteen minutes means that the organisation can tolerate\nlosing at most the changes from the final fifteen minutes before the event\nunder the agreed failure scenario. If the last verified recoverable state\nis fifty minutes old, the result misses that objective even if the service\nstarts immediately.\n\nSpecify the data and the scenario alongside the number. Losing a recent\nlocal cache may matter little when it can be rebuilt from a trustworthy\nsource. Losing accepted orders matters differently, particularly when a\ncustomer has already received a confirmation. A deleted row, a full\nregional outage and a compromise of operating credentials can leave\ndifferent backups available. A target attached to one data store and one\nwell stated scenario can be tested.\n\nThe recovery time objective measures how long the system can be unavailable\nwhile it is restored. Recovery point and recovery time therefore lead to\ndifferent questions. One asks which state survived. The other asks when\nuseful service returned. A daily backup that restores in minutes may meet\nthe time target and still lose a day's work. A copy that retains recent\ntransactions may meet the point target and still take too long to verify\nand serve.\n\n## Why it matters\n\nA backup schedule is an input to a design. The objective is a requirement for\nwhat can be recovered. A job advertised as running every five minutes might\nlag behind the application, fail without a useful alert or hand back a copy\nthat cannot be replayed. A team can test the difference by placing known\nchanges at recorded times and then tracing the last contiguous state it can\nrestore. Measuring the end result reveals problems which an ordinary\nsuccessful backup log cannot show.\n\nThe target should reflect what the organisation is willing and able to\nreconstruct. Ask what happens to a receipt, an accepted booking or a supplier\nevent produced during the lost window. Some records can be replayed from\na durable upstream log. Others exist only in the application that recorded\nthem. Estimate the cost and risk of reconstructing each record, and give\nthe work an owner. A small objective on a small set of important state can\nbe more useful than the same ambitious number across every cache and report.\n\nFrequent copying is only one approach. Depending on the data, a recovery\nstrategy may combine periodic snapshots with a transaction log, a replica,\nan immutable copy or independent upstream event retention. These options\nhave different properties under different failures. A replica can transfer\naccidental deletions as quickly as legitimate work. A separate copy helps\nonly if the restoration team can access, interpret and validate it when it\nis needed.\n\n## What it looks like in practice\n\nEngineer ApS describes [backups that have actually been restored](/portfolio/backups-that-have-actually-been-restored/)\nand the availability and handover terms in its [service levels](/service-levels/).\nThe portfolio account gives a reader evidence about exercising restoration.\nThe numerical example that follows is a hypothetical way to test a point\nobjective without assuming a target for the published case.\n\nAssume a shop adopts a fifteen-minute RPO for accepted orders in the event\nof a failed database. It snapshots the database and saves a sequence of\ntransaction records separately. At 14:00 a rehearsal removes access to the\nprimary system. Write known test orders before and after 13:45, as well as\na cancellation and a payment update. Start the recovery with the permissions\nthe duty team will actually have. Restore to a separate environment and\nchoose the latest consistent state whose data can be verified.\n\nIf that state includes all expected transactions through 13:50, the observed\ngap is ten minutes and the point target is met for the rehearsal. A visible\norder from 13:58 does not alone establish a two-minute gap if earlier records\nare missing. Compare row identities and totals and check the dependencies an\norder needs to be actionable. Preserve the restoration evidence and account\nfor what would happen to the orders in the remaining window. Record the time\nwhen useful service could return separately.\n\nA second rehearsal can change the circumstance. Remove the ordinary\noperator's access to the shared control plane or simulate an accidentally\ndeleted item by choosing a clean point before that edit. Inspect the copies\navailable under that condition. Results for one scenario are useful evidence\nfor that scenario, and a reason to test additional ones whose dependencies\ndiffer.\n\n## Where it goes wrong\n\nTeams often assume that replica lag is an end-to-end measure of loss.\nAn application may write order details to one store and attach evidence to\nanother. If their time boundaries differ, an apparently recent database\nmay reference files that did not survive. Include linked records and check\nconsistency across those boundaries in the actual recovery exercise.\n\nA recovery run may still rely on unavailable credentials, an undocumented\nencryption key or a human who remembers how to find the right copy. Give the\nauthorised duty role the actual runbook and exercise access with it. Write\ndown which timing and capabilities are assumptions. Time spent locating\ncopies is also useful evidence for the separate time objective.\n\nFinally, review the target when data becomes harder to reconstruct. The\nfirst order application may have started as a copy of a manual log, for\nwhich a day's loss was manageable. After the system becomes the only accepted\nrecord, the same target has a different meaning. Revisit the loss limit and\nrepeat the exercise with the affected process owner before treating the\nold measurement as evidence for the new service.\n\n## Further reading\n\n- [NIST contingency planning guide for federal information systems](https://csrc.nist.gov/pubs/sp/800/34/r1/upd1/final)\n- [AWS Well-Architected recovery objectives and disaster recovery](https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/plan-for-disaster-recovery-dr.html)\n- [NIST recovery-point objective terminology](https://csrc.nist.gov/glossary/term/recovery_point_objective)\n\n\u003chttps://engineer.company/glossary/recovery-point-objective/\u003e\n","date_modified":"2026-11-19T19:18:49Z","description":"Set a tolerable loss window, turn it into backup and replay requirements, and measure the most recent data you can actually restore.","formats":{"gemtext":"https://engineer.company/glossary/recovery-point-objective/index.gmi","gopher":"https://engineer.company/glossary/recovery-point-objective/gophermap","html":"https://engineer.company/glossary/recovery-point-objective/","json":"https://engineer.company/glossary/recovery-point-objective/index.json","markdown":"https://engineer.company/glossary/recovery-point-objective/index.md","pdf":"https://engineer.company/glossary/recovery-point-objective/index.pdf"},"kind":"page","language":"en","title":"What is a recovery point objective?","url":"https://engineer.company/glossary/recovery-point-objective/","version":1}