OBLYTECH Logo
OBLYTECH Logo
ServiceNow Upgrade Readiness Requirements Workshop Questions

ServiceNow Upgrade Readiness Requirements Workshop Questions

This page sets out the questions to settle before a ServiceNow upgrade is planned. The questions cover the current version and upgrade cadence, the custom code and integrations that must still work, downtime tolerance and testing.

Current Version and Upgrade Cadence

What version of ServiceNow are you currently running? Is it a recent version or several versions behind? If you are several releases behind, the release notes for each version in between need reading before you start.

Answer impact: If you are significantly behind, the upgrade effort is larger. Each version in between can carry changes that affect your instance, so it needs checking. The first upgrade after several skipped versions can carry more risk, because more changes arrive at once. Consider a phased upgrade strategy: upgrade to an interim version first, then to the target version.

How often do you plan to upgrade? One upgrade a year is a workable pace, because each step stays smaller. More frequent upgrades keep you current but require sustained resources. Less frequent upgrades (every 18+ months) risk accumulating breaking changes.

Customization Inventory and Version Dependencies

What custom code exists in your instance? Inventory all business rules, workflows, client scripts, and API integrations. For each, document: what it does, which version-specific features it depends on, whether it will work after the upgrade.

Answer impact: A heavily customized instance has upgrade risk. Each customization is a potential compatibility issue. An out-of-the-box instance has fewer customizations for the upgrade to conflict with; a highly customized instance requires extensive testing and remediation. If customization inventory is not known, conduct it before planning the upgrade.

Do you have ServiceNow partner support, or are you managing upgrades internally? Partners often have upgrade playbooks and experience; internal teams may not. Budget accordingly.

Integration and Third-Party Connector Compatibility

What systems integrate with ServiceNow? List every integration: CRM, ERP, ITSM tools, communication platforms. For each, confirm the vendor supports the target ServiceNow version. Do any integrations need updates?

Answer impact: If any vendor does not support the target version, you may be blocked from upgrading until they provide support. Contact vendors early; give them notice of your upgrade plans so they can provide updates if needed.

Downtime Tolerance and Cutover Timing

How much downtime can your business tolerate? If ServiceNow is down for two hours on a Saturday, is that acceptable? What about during business hours? Business hours downtime is much more disruptive than weekend downtime.

Answer impact: Tolerance for downtime dictates cutover strategy. If you cannot tolerate any business-hours downtime, you may need parallel running (old and new versions running simultaneously for a period, with data synchronized between them). This is complex and expensive but preserves continuity.

When is the best time to upgrade? Identify low-activity periods (weekends, holiday weeks, off-season for your business). Plan upgrade during these windows to minimize impact.

User Communication and Support Readiness

Do users know an upgrade is coming? Have they been informed of what will change? UI changes, new workflows, and new features should be communicated before the upgrade so users know what to expect.

Answer impact: Communication and training prevent post-upgrade confusion. Users who are surprised by UI changes or new workflows will struggle and create support tickets. Pre-upgrade communication reduces support load post-upgrade.

Is the support team (help desk, platform support) prepared for the upgrade? They need to understand what changed and be ready to help users. Warning sign: if support team members are not aware of the upcoming upgrade or its scope, communication is not adequate.

Testing Strategy and Resource Allocation

Do you have a test instance that matches production? Testing the upgrade in a realistic environment requires a production-like test instance. If you do not have one, building it adds effort and time to the upgrade project.

How much time can you allocate for testing? Comprehensive testing takes weeks; abbreviated testing takes days. The less time you spend testing, the higher the risk of surprises in production.

Post-Upgrade Stabilization Period

Plan for a stabilization period after upgrade where issues are monitored closely and support is ramped up. The first week after upgrade is critical; bugs and issues often surface. Have extra staff available for support. Plan to defer non-critical work during the stabilization period so people are available to address upgrade-related issues. After a week of stability, return to normal operations.

Document all issues found during stabilization. Track whether issues are upgrade-related bugs, customization incompatibilities, or existing instance problems coincidentally discovered during upgrade. This documentation helps prioritize fixes and provides feedback to ServiceNow vendor about issues with the upgrade process. Post-stabilization, document lessons learned and improvements for the next upgrade cycle.

Make the output a log of outcomes, not a list of prompts

A question list is a prompt, and a prompt on its own rarely changes anything. What a requirements workshop has to produce is a record in which every question ends in one of three states: decided, deferred with a single accountable person and a revisit date, or rejected as out of scope for this upgrade. Anything still open when the room empties means the session has not finished, however productive the discussion felt. Set this expectation before the first slide, because many workshops drift into discussion and leave the recording of outcomes to a note-taker who is already tired.

The practical test is simple. Ask at the end whether a person who missed the session could read the log and know what the team has agreed, what it is still arguing about, and who is responsible for each open point. If the answer is no, the output is minutes, not a plan.

Matching attendees to the decisions they own

The people who answer the questions matter more than the number of people present. A useful room includes someone who can make a cutover call, someone who knows the custom code well enough to say which items are fragile, someone from operations who has handled outage calls, and a business representative for the process that cannot tolerate interruption. Sales or account contacts from the vendor side can answer product questions, but they should not steer the scope, and internal stakeholders who only want updates are better served by a short written summary afterwards.

Question area Who must answer it Output the log should hold
Starting version and target release Platform owner The confirmed current release, the target and the reason the target was chosen
Customization inventory Technical lead with the developers who built the code Each item marked keep, fix or remove, with the reason written beside it
Integration support status Integration owner Per-integration confirmation that the vendor supports the target, in writing
Downtime tolerance Business owner with operations An agreed window, and a stated rule for what happens if the window overruns
Staff and user notice Service desk lead Message dates, an audience list and a named owner for each message
Test capacity Project manager A test calendar with named testers and the environment each test runs in

Spotting answers that close a question without answering it

Some answers sound complete and carry no information. “We test everything” and “the vendor will tell us if something is wrong” are two common examples. Ask for the test list, or ask for the vendor’s confirmation in writing. A related trap is treating silence as evidence: “nobody has reported problems with that workflow” means there are no tickets, which is not the same as there being no problems. Check the incident history for the area before accepting the claim.

Where the workshop depends on memory, ask the owner to point to a record. A configuration export, a support ticket or a test result is a stronger answer than a recollection, and the log should note which kind of evidence each answer rests on.

A short procedure for running the session

The order below keeps the session moving and makes the output reusable for the test plan and the change record.

  1. Send the question set and the inventory request well ahead of the session, so owners arrive with evidence rather than opinions.
  2. Open with the stop conditions, so the group knows in advance which findings would delay the upgrade.
  3. Take the question areas in dependency order, from the target release through to communication.
  4. Record each answer in the decision log with its source and its status: decided, deferred or rejected.
  5. Read the deferred items back aloud and give every one an owner and a revisit date.
  6. Circulate the log and ask for corrections within a fixed window, after which the log becomes the baseline for the test plan.

The stop conditions deserve particular attention. Typical examples include no test instance that resembles production, an unresolved vendor position on a core integration, or a business peak that falls inside the planned window. The notes on release acceptance criteria are a useful reference for setting them. Writing these down before the session means the team is not negotiating them under schedule pressure, when everyone wants the date to hold.

Questions people ask after the workshop

How long should the decision log be kept?

Keep it until the upgrade has been stable for long enough that the post-upgrade review is complete, then archive it with the change record. Its value is highest in the next cycle, when the team can see which assumptions were wrong and which questions were never asked.

What if the business refuses any downtime at all?

Record that position as a constraint and bring the options back to the same group. Zero tolerance usually leads to a discussion of parallel running or a phased rollout, both of which cost more. The business should choose between those costs knowingly rather than discover them during cutover.

Should integrations be covered in the same session as the platform questions?

Cover the support status in the main session, then hold a separate working session with each integration owner. Integration testing involves different people and different evidence, and a single session tends to run out of time before the detail gets reached.

Who signs off the deferred items?

The person who owns the risk should sign off, not the person who raised the point. If the owner of a deferred item is unclear, the deferral is not finished, and it should go to the sponsor for a decision before the log is accepted.

The output of this session feeds the testing plan and the licensing review. Read the testing and acceptance checklist next, and take the deferred items from the log into the licensing and cost planning so that open points carry their owners forward.

How to judge ServiceNow upgrade and release readiness before you commit to a date

An instance is ready for an upgrade when four things can be shown, not just stated: the team knows what is customised, every integration has a confirmed position, a production-like test environment exists, and someone can reverse or absorb a bad outcome. If any one of these is missing, the honest status is “not ready yet”, whatever the calendar says.

The simplest way to use the workshop answers is to grade each area as ready, ready with conditions, or not ready. Grade by evidence. A written vendor confirmation or a completed inventory counts; a verbal assurance does not.

Grade What it looks like What to do
Ready Evidence exists and an owner has seen it Move to test planning
Ready with conditions Evidence is partial, but the gap has an owner and a date before testing starts Proceed, and record the condition against the date
Not ready No evidence, or the owner cannot be named Hold the date until the grade changes

The rule that keeps this honest is that the overall grade is the lowest grade of any area, not the average. A strong test plan does not make up for an integration nobody has checked, because the failure will arrive in production in the one place you did not look.

Readiness also depends on what you ask the upgrade to carry. Moving a lightly changed instance to the next release is a different job from retiring old customisations at the same time. Combining them makes failures hard to trace, so many teams keep the upgrade itself separate from clean-up work and schedule the clean-up afterwards.

Finally, decide before the work starts what “done” means. A common choice is that agreed business processes run end to end on the new release, open defects are triaged by severity, and the support team has confirmed it can handle the first days. Treat this as a draft and adjust it to your own instance and release notes.

  • Check the target release’s own notes on your instance for changes that touch your customisations, rather than relying on general advice.
  • Rehearse the cutover at least once in the test environment and time it.

For the sequencing around releases, see how update sets and release management are handled, and the wider upgrades and release readiness guides. If performance is a concern, the platform performance and upgrade readiness page covers that angle.

Settling the Go or No-Go Call and the Reversal Route Before Cutover Starts

Decide two things before cutover begins: who makes the go or no-go call, and what the team will do if the call goes wrong after the switch. The workshop log should record each answer as a named person and a written rule.

The reversal route deserves more attention than it usually gets. ServiceNow does not provide a universal rollback, and rollback covers updates such as patch-to-patch moves, within a default window of 10 days for patch upgrades and 15 days for plugin activations. Restoring a production instance from backup is typically a last resort, because it returns the data to an earlier point, so records created or changed after the switch are at risk. A team that has not worked out how long that restore takes, or who holds the access to run it, does not yet have a rollback plan. Confirm the options on your own instance and release with the platform owner, and write the answer into the log. Changes that arrived through update sets are one thing to check when a reversal is being weighed. The reversal route itself is the rollback context, which the platform keeps for patch upgrades and plugin activations.

Tie the go or no-go rule to the stop conditions already agreed. Pick one checkpoint, usually after the smoke tests on the new release and before users are invited back, and make the call there rather than later. Keep the checks short: the core business processes complete end to end, the integrations confirmed in writing still respond, and the support desk can see the tickets it expects. A failed check returns the decision to the named owner, who chooses between fixing forward and reversing, with the cost and time of each stated. The change governance approach to release day is a useful model for who signs what.

Three roles keep the window from turning into a debate:

  • The decision owner, who can say go or no-go and is not also running the technical work.
  • The technical lead, who gives the decision owner the results of each check and how long each option takes.
  • The business representative, who says whether a failed process can wait or must be restored first.

Keep the reversal decision apart from the defect list. Teams often reverse over a single fault that could have been absorbed with a documented workaround and a revisit date. The opposite mistake is absorbing a fault that stops a core process for users. The log should name the faults that count as reversal triggers, so the choice is made against a rule rather than under pressure.

Frequently asked questions

Should we test the upgrade multiple times or once?

Test at least twice. First test identifies major issues and incompatibilities; you fix them. Second test validates the fixes and confirms the upgrade procedure is repeatable. If you find critical issues in the second test, you may need a third test cycle.

What if users find issues after the upgrade that testing did not catch?

This happens; user workflows are often more complex than test scenarios. Develop a post-upgrade support plan: extra staffing for a week or two, channels for users to report issues, and a priority process for addressing critical issues. Have a rollback plan, but rolling back should be a last resort after diagnosing the issue.

How do we decide whether to upgrade to a new version or stick with the current version?

Factors: security (current version is supported and secure), features (target version has features you want), vendor support (can you get help with issues), and resources (can you afford the upgrade effort). If current version is supported and secure, and there is no pressing need for new features, sticking with it is defensible. However, falling too far behind makes eventual upgrades much harder.

Talk to Oblytech

Planning, running or fixing a ServiceNow programme? Tell us where you are and we will suggest a sensible next step.

Contact us