Identifying a weak point in a business workflow is an important achievement, but improvement does not follow automatically. Removing an approval level or assigning a task automatically can speed up work while also creating new opportunities for errors. Implementing process change is therefore more than a technical adjustment: it is a business decision that requires accountable owners, trained staff, measurement rules and a contingency plan.
In this installment of our series on business process optimization, we walk through implementation using an illustrative complaint handling example that does not represent actual customer results. The starting point: customer complaints are assigned manually, several cases are returned for missing information, and responses await management approval. The goal is not to automate every step, but to reduce unnecessary waiting while retaining essential controls.
1. Formulate a testable change hypothesis
“Make complaint handling faster” provides direction, but is not precise enough to support a decision. First, separate the symptom from the suspected cause. Long resolution times may stem from incomplete data capture, unclear responsibility, an overloaded specialist or unnecessary approval. These require different interventions.
In our example, the hypothesis could be: if fully documented, low-risk complaints are routed directly to the appropriate case handler according to a predefined rule, assignment waiting times will decrease without increasing the rates of incorrect responses and reopened cases. For now, we leave the approval process unchanged: this makes it easier to identify what caused the results to change.
Before testing, document the baseline, the scope of cases to be examined and the acceptance criteria. Also define what qualifies as low risk; this should not be left to individual case handlers to interpret.
- When does the assignment waiting period begin and end?
- Which data must be present before a case can proceed?
- What counts as an incorrect assignment, rework or a quality issue?
- What results would justify a wider rollout of the new process?
2. Define responsibilities and decision-making authority
Before the pilot, clarify who can authorize the change, who will operate the new process and who can stop the test. If these roles overlap without clear boundaries, an exceptional complaint can easily be left without an owner. Notifying everyone involved is no substitute for accountability.
- Process owner: is accountable for the business objective, the rules and the rollout decision.
- Subject matter lead: defines complaint classifications, exceptions and quality requirements.
- Technical lead: manages configuration, testing, access permissions and the version transition.
- Operational contact: collects user issues and ensures they are passed on to the appropriate people.
Roles can be combined in a smaller organization, but decision-making authority must still be documented. Ensure backup coverage, a designated escalation path and clear authority to stop the pilot. In the complaint handling example, explicitly document who takes over a case when automatic assignment cannot identify an appropriate owner.
Approval cannot be purely an IT matter. The process owner must also accept the tasks, workload and responsibilities that the new process will create for the teams involved.
3. Design a limited pilot that still produces meaningful findings
The purpose of a pilot is not to prove the idea right at all costs, but to test the assumption. Choose a clearly defined scope, such as one complaint type handled by a designated team. High-risk or exceptional cases, or those requiring legal interpretation, can remain on the existing route, provided this complies with the organization’s rules.
Align the test duration with case volumes and turnaround times. A few quickly resolved cases will not necessarily reveal problems that emerge later. The observation period must be long enough for reopened cases and exceptional situations to become visible. Low case volumes may require longer observation.
The pilot plan should include entry, exit and stop criteria. Entry criteria might include a tested assignment rule and a trained team. Closing the pilot requires enough cases for meaningful evaluation and documented findings. One reason to stop could be complaints being left without an owner.
Where possible, compare results with a previous period or a parallel team handling a similar mix of cases. Also record staffing changes, campaigns and exceptional workloads: without this context, it is easy to attribute an improvement to the change when it was caused by something else.
4. Prepare staff for changing situations
Training is useful when it goes beyond demonstrating the new interface. Staff need to understand why the process is changing, which decisions the system makes and when human intervention is required. Using the old approach out of habit is particularly risky if it leads to parallel case handling that cannot be tracked.
Before the complaint handling pilot, practice both typical and problematic scenarios: correct automatic assignment, an incorrect category, missing data, an absent case handler and a case nearing its deadline. Every situation should have a clear next step and accessible support.
- Prepare brief, role-specific work instructions.
- Provide a separate practice environment without putting live cases at risk.
- Document where errors and improvement suggestions can be reported.
- Use a practical exercise to check whether the new way of working is clear.
Do not automatically treat feedback as resistance. If several case handlers bypass the same step, this may indicate inadequate training, but it could also point to a poorly designed process. One of the key benefits of a pilot is uncovering these operational details.
5. Implement a controlled version transition
When implementing a process change, everyone involved must know which rules apply and when they take effect. Make sure the old and new process versions, the details of the change, its approver and its go-live date are clearly identifiable. The change log will later serve as the basis for troubleshooting and impact assessment.
Cases already in progress require a separate decision. In the complaint management example, we can choose to route only newly received complaints through the new process, while existing ones follow the old process through to completion. If we do migrate open cases, we must verify that assigned owners, deadlines, statuses and case histories are preserved.
Before going live, test permissions, notifications, substitute assignments and any system integrations. Successfully saving the configuration does not, by itself, prove that the entire business process works. Track a test complaint all the way from registration to closure.
For xFLOWer-based improvements, the business transition plan should also be the starting point: first document the desired way of working and the checkpoints, then align the specific system settings with them.
6. Measure speed, quality and workload together
Success is not simply a matter of launching the new workflow. After the pilot, compare metrics calculated using the same definitions against the baseline. Alongside faster assignment, check the total turnaround time: the waiting time may simply have shifted to the next stage of work.
- Assignment waiting time: how much time passes between intake and assigning an owner?
- Total turnaround time: when does the case reach the defined closure status?
- Quality: how do the rates of incorrect assignments, reopened cases and rework change?
- Workload: are open cases accumulating within a single team?
In addition to the average, examine the median and the cases with the longest turnaround times. Identify missing timestamps and document how paused cases are handled. Do not let unresolved complaints become invisible in the assessment: report their number and age separately.
If the results are mixed, do not force a wider rollout. The hypothesis may be only partially confirmed: for example, assignment may become faster while the categorization rule still needs refinement.
7. Prepare for rollback and continuous refinement
Do not wait until a failure occurs to write the rollback plan. Define in advance which event will trigger it, who will authorize it and what will happen to cases already in progress. Switching back to the old configuration does not, by itself, resolve changes made to data and responsibilities in the meantime.
The plan should cover pending tasks, notifications already sent and actions already completed. For example, a response sent to a customer cannot be undone by simply restoring the system; a separate corrective action may be necessary. Case-by-case reconciliation may therefore also be needed after the transition.
If the pilot succeeds, gradually expand the range of cases covered and conduct regular reviews. The next improvement cycle could explore simplifying the approval path, but with a separate hypothesis and controls.
In summary: lasting improvement comes from combining a measurable objective, clear accountability, a controlled pilot and consistent follow-up. What is needed is not a one-off transition, but a way of working that supports learning and improvement.
Would you like to improve your existing workflows safely? Consult xFLOWer experts on which process step to test first and what measurement, implementation and rollback plan would support the change.
This article is part of our complete guide: Business Process Optimization: A Complete Guide to Improving Existing Workflows
xFLOWer.ai
Design the process behind this article – for free!
AI reads the processes hidden in this article; open any of them in xFLOWer.ai with one click.
