Skip to main content

How to Evaluate an In-House On-Call Tool Before Replacing It

GuideWritten by oncall.fyi editorialPublication approved by Burak YApproval recorded 11 September 2026
Sources and verification
Source dates
Oldest source check: 10 September 2026.
Technical verification
Separate technical verification has not been recorded.

Publication approval and technical verification are recorded separately. Automated link checks establish reachability, not accuracy. A checked example verifies only its stated test cases, not the whole article or your production setup.

In short

List the specific failures and operating costs of the current tool before choosing whether to improve or replace it. Test missed events, incorrect recipients, retries, failed channels and unavailable maintainers. Price the remaining engineering work as well as the subscription alternative. A replacement is justified when it closes demonstrated gaps at an acceptable total cost and can pass the same acceptance tests.

Key takeaways

  • Dissatisfaction needs concrete failure evidence before a replacement decision.
  • The system needs a maintenance owner beyond its original author.
  • Evaluate the old and proposed tools against the same failure cases.

Turn “we hate the tool” into findings

Collect a few real examples with timestamps: an event vanished, an override paged the wrong person, or an engineer spent half a day repairing a connector. Record whether the failure affected detection, delivery, ownership or administration. An awkward interface is a legitimate cost, but it is different from a lost page.

Do not invent missing capabilities just because a tool is homegrown. Equally, do not accept “the script has worked for years” as evidence that its backup path works.

Build a bounded acceptance suite

ScenarioEvidence requiredLikely owner
Repeated source eventIntended deduplication, no lost incidentIntegration maintainer
Primary unavailableBackup actually receives escalationOn-call owner
Invalid integration credentialVisible failure and accountable responsePlatform owner
Process restartIn-flight work has documented outcomeTool maintainer
Original author absentSecond operator can restore serviceEngineering manager

Use a lab service and test recipients. Include a delayed recovery event and a configuration rollback. Keep the same suite for every replacement you evaluate.

Price the work you still own

Estimate connector maintenance, dependency updates, credential rotation, backup and restore, delivery troubleshooting, support and training. Record who covers the tool itself when it breaks outside working hours. The Google SRE on-call chapter (opens in a new tab) provides general operating context; the worksheet here is a proposed decision method, not a published benchmark.

For an illustrative decision, a team might find the core route works but only its author knows how to restore it. Writing and testing a restore procedure could be a smaller fix than migrating every integration. Another team may discover that reliable delivery requires more work than it can staff. Those findings support different decisions.

Choose repair, contain or replace

Repair a narrow fault when ownership and remaining maintenance are sustainable. Contain the tool's scope when it can safely serve a smaller noncritical use case. Replace it when required behavior cannot be supported with the available people and budget.

Write the decision with demonstrated gaps, alternatives, cost assumptions and a revisit date. Do not compare an audited old system with an untested vendor demo.

Make the next step reversible

For repair, change one failing path and rerun the suite. For replacement, follow the pilot migration plan and preserve the existing route until the new one passes. In both cases, assign someone besides the original developer to run the paging test from the documentation.

Did this help?

Your answer helps us improve this guide. We save only the page and your choice for 30 days.

No name, email, or incident details are requested.

Frequently asked

Is building paging software always a mistake?
No. The decision depends on demonstrated requirements, available maintenance capacity and tested behavior, including operation when the original author is absent.
What is the first useful measurement?
Start with a concrete missed or delayed response and the engineering time spent operating the tool. Classify the failure before comparing solutions.

Sources

Vendor facts change. Each source below shows the date this page last checked it.

  1. Being on-call Google SRE. Checked 10 September 2026.

Related

One practical idea, occasionally

The On-Call Brief: short field notes, templates, and operational lessons.

Follow the field guide

Email subscriptions are not open yet. Read new guides in your feed reader — no email address needed.

Subscribe with RSS