Oblytech’s Continuous Improvement services aim to help you identify areas for enhancement and review changes against your business goals.
Continuous improvement is a phrase commonly attached to support contracts, and it can give sponsors very little to hold a supplier to. A sponsor who approves an improvement budget without defining its outputs often discovers, a year later, that a steady stream of small changes has been made and that nobody can say whether the platform is better used. The useful way to frame the work is as a cycle with visible products at each stage: a short list of observed problems, a decision about which ones justify change, a set of changes released in a known window, and a measure that shows whether each change did what it was meant to do. If any of those products is missing, the improvement is activity rather than progress.
The cycle also needs a stopping rule. Some improvements are worth making only once. Others need to be re-examined each quarter, because the business process they support keeps moving. A retirement step, where an improvement is removed because the need has passed, belongs in the cycle as much as the addition of new capability. Platforms that keep growing with no retirement step become expensive to change, and the cost is felt most by the people who must maintain them.
Improvement decisions are only as good as the evidence behind them, and no single source is complete. Platform records show what users did. Surveys and feedback show what users felt. Support data shows where they got stuck. Each source has a characteristic blind spot, and the cycle works best when the blind spots are known in advance.
| Source | What it reveals well | Where it misleads |
|---|---|---|
| Platform usage and transaction records | Which paths are used, how long steps take and where records stall | It cannot show work that happens outside the platform, such as side conversations |
| Reopened and reassigned records | Where ownership or resolution is unclear | A reassignment can be correct when the first owner was wrong about the category |
| Short user feedback at the point of use | Friction that users notice immediately | Users rarely report the problems they have already learned to avoid |
| Service desk conversations | Recurring questions that the platform should answer without help | Agents may answer the same question differently, which blurs the signal |
| Interviews with process owners | The reason a step exists and what would break without it | Owners often describe the process as designed rather than as performed |
A practical habit is to test each suspected improvement against two sources before it enters the backlog. If the platform data and the service desk conversations point in the same direction, the case is often strong. If they disagree, the disagreement is itself a finding, and it is often a sign that a process is being performed differently in different teams.
A continuous improvement backlog fills quickly, and much of what goes into it will look worthwhile. Prioritization needs a small set of criteria that the sponsor and the platform owner accept in advance. Useful criteria include how many people the problem touches, how often it occurs, whether it creates risk to records or compliance, how much effort the change requires, and whether the change makes a later change easier or harder. A change that removes a common source of rework can be more valuable than a visible feature, and the criteria should make that visible.
The discipline lies in recording why an item was rejected as well as why one was accepted. A declined request that comes back six months later is easier to handle when the earlier reasoning is on file, and it prevents the same argument being repeated with different people in the room.
An improvement backlog fails in one of two ways. It either becomes a dumping ground that no one reviews, or it becomes a queue that only the platform owner can move. A governance rhythm avoids both. A short standing review, with the sponsor or a delegate, the platform owner and one representative from the largest user group, meets on a fixed cadence. Decisions are recorded, owners are named and anything not decided within a set period is escalated rather than left to age. The review should be short enough that busy people will attend, and it should have authority to say no as well as yes.
Release timing is part of governance. Changes should be grouped into predictable windows so that testers, trainers and the business can plan around them. Frequent small releases feel responsive, but they tire users and make it difficult to tell which change caused a new problem.
Users stop noticing improvements when the communication about them is poor. A release that quietly improves a form may be valuable, but if no one tells the people who use the form, the benefit is lost and the next change is treated with suspicion. A short note for each release, written in terms of the task the user performs rather than the feature that was built, works better than a long list of technical changes. Where a change alters a familiar step, a brief walkthrough or a pinned example helps users find the new route.
The same principle applies to feedback. Asking for input and never showing what happened to it teaches users that feedback is pointless. A simple practice is to publish, after each review, the requests that were accepted, the ones declined and the reason for each. That record also reduces the number of repeated requests that arrive from people who had not seen the earlier decision.
Continuous improvement works best on the existing shape of the platform. When a request would add a new process, a new module or a substantial change to how a team is organized, it is a project, with its own business case, funding and delivery approach. Treating such requests as improvements hides their cost and leaves the sponsor unable to judge whether the budget is being used well. The boundary should be agreed early, with a test that the review can apply: if the change alters who does the work, what data the work depends on, or which decisions the platform supports, it probably needs project treatment.
The cadence should match the rate at which requests arrive. A review that meets more often than requests come in will spend its time on repetition. A cadence that is too slow lets urgent items age. Some teams settle on a rhythm that lines up with the release windows, so that each decision has a visible route into delivery.
The sponsor or a delegate with budget authority, the platform owner and a representative of the largest user group are the minimum. Including someone who runs the service desk is useful, because they see the recurring questions first.
Yes, provided the budget is sized for the kind of work it covers. Small, well-scoped changes can be absorbed in a standing allocation. Anything that needs new design, new data or new people should be reviewed separately.
Success is a measurable change in something the business cares about, such as the time a common request takes, the share of requests handled without escalation or the number of repeat contacts about the same task. Counting completed changes alone tells the sponsor how busy the team was, not whether the platform improved.
If the current support arrangement produces a steady flow of changes without clear results, the sensible move is to agree the evidence sources, the prioritization criteria and the review cadence. Discuss how a continuous improvement cycle would be governed to agree those terms before the next budget period starts. The support and maintenance overview is the wider support page, and the roadmap prioritization page is the related pointer for sequencing larger decisions. For the reporting side, the guide to dashboards explains what a ServiceNow dashboard is.
An improvement has worked when the measure you wrote down before the change moves in the intended direction, over a window long enough to separate the change from normal variation, and when no related measure got worse at the same time. If any of those conditions is missing, the result is not yet known, and the item should stay open on the backlog as unverified rather than be closed as a success.
Write the measure into the change request before release. A goal such as making the form faster cannot be checked later. A usable note names the figure, the population it covers and the direction you expect, for example the time from submission to first assignment for one request type falling. If the platform has no recent baseline of its own, a platform health-check readiness review gives you one before the change goes in.
Choose the comparison window with care. Ticket volumes move with month-end cycles, seasonal demand and staffing, so comparing one busy week with one quiet week proves little. Compare comparable periods on either side of the release, and list any other change that landed in the same window, such as a new assignment rule or a staffing change.
Check for side effects. A change that speeds up one step can push work into the next step and leave the total time unchanged. One symptom to watch for is a rise in reassignments or reopened records, so read those alongside the target measure. If the instance itself slowed down after the release, start with the performance troubleshooting steps, because a slow platform can hide the effect of a process change.
Record the result either way: whether the measure moved, by how much, over what window, and what else changed. A change that produced no effect is still useful, because it shows which problem was not the cause, and the record stops the same idea returning with no evidence against it.
Four checks cover the gaps that usually appear in a review:
ServiceNow optimization aims to help a platform that already runs do its job better: fewer steps per request, fewer misrouted records, forms people complete without help and reports people read. Depending on scope, it may focus on existing processes rather than adding new ones. A review can start with places where people wait, re-key data or ask the same question twice.
Possible areas to investigate include performance, configuration, adoption and process design. If the process appears sound but the platform is slow or unstable, consider performance work. If people struggle to use the platform, consider a configuration or adoption change, such as a clearer form or a knowledge article. If the process no longer matches how the business works, consider redesign and review whether it needs a separate project.
| Symptom | Possible cause | Possible response |
|---|---|---|
| Pages slow down at busy times | Heavy queries, background jobs or reports that run against live tables | Performance tuning, starting from the slowest transactions |
| Requests reach the wrong team | Assignment rules written for an earlier structure | Rule review tested against recent real records |
| Users raise tickets to ask how to do a task | A form with unclear fields or no written guidance | Form change and a short knowledge article |
| The workflow is followed but the result is still wrong | The process was designed for other volumes or policies | Process redesign, handled as a project |
When a review starts, sort every symptom into a table like this before anyone proposes a fix. Then take the cheapest response that addresses the cause, and hold the redesign items for a separate decision. A common mistake is starting with the most visible item. A new dashboard is easy to show a sponsor, but if records are going to the wrong team, the dashboard can show the misrouting without changing where records go.
Where the symptoms point to speed, performance tuning and upgrade work covers that separately, and a ServiceNow health check is a sensible way to gather the evidence before any change goes in. For the service side, see ServiceNow optimization and enhancement services.
Once the remaining items need a decision about what the work is for, rather than how it runs, treat them as new work.
One practical way to judge a continuous improvement cycle is to follow one complaint through it. Take a common case: the service desk reports that requests in one queue keep moving between teams before anyone works them. The first step is to turn that complaint into something observable. Pull the reassignment history for the affected records and note which group handed each one on, and at which step. If the same cause keeps showing up, it may belong in a problem management record rather than in a one-off change.
The second step is the decision. Suppose the history shows that the first owner was usually wrong about the category. Investigate whether the routing rule contributed before deciding on a rule change. Assess the scope, reversibility and testing needs of a routing change before choosing a release window. A proposal to retrain a whole team can be considered in the governance review.
Write down the decision and the reason for it, including the option that was turned down. If someone asks in three months why the queue was not redesigned, the answer should already be on file. The wider cycle this fits into is set out in ServiceNow continual improvement management.
The release note should describe what changes for the person using the form, such as which group a request reaches first. Describing the rule itself tells the reader little. The service desk gets the same note, so agents stop explaining the old routing to callers.
Measurement is a step that is easy to skip. Choose the measure before the release, and make it the complaint itself: reassignments in that queue over the same period before and after the change. Read a sample of the records that still move as well. A falling count with unchanged records could mean work is being routed around the rule rather than fixed by it. For how value is tracked after go-live, see the value realization operating model.
If the count does not move, review whether the change addressed the suspected cause and whether other factors affected the result. Reopen the observation stage with the new evidence before deciding on another change.
Guides from the Oblytech ServiceNow library that explain the practice behind this page.