Collaborative Minds Blog
  1. CMWLab
  2. Collaborative Minds Blog
  3. Workflow
  4. Parallel vs Sequential Approvals: Dependencies, Rejections, and Timing

Parallel vs Sequential Approvals: Dependencies, Rejections, and Timing

Sequential approvals run in a defined order: a later stage starts after the required earlier result. Parallel approvals let independent reviews begin from the same complete submission. Choose the route by dependency, then define how approval, rejection, correction, and cancellation affect the whole request.

The important design decision is whether a reviewer needs another reviewer’s output. Starting tasks together can reduce waiting between independent checks, but the workflow still needs an explicit rule for reaching a final decision.

Parallel vs Sequential Approvals at a Glance

Choose the pattern from the dependencies
Design questionSequentialParallel
When do reviews start?Each stage starts after the required preceding outcomeIndependent reviews can start from the same ready submission
What is the main prerequisite?The dependency and order are explicitEach reviewer has the information needed without waiting for another review
How does the route finish?The required stages reach their permitted outcomesThe completion policy evaluates the required branch results
What needs special attention?Unnecessary stages and queues between themRejection, cancellation, changed versions, and synchronization

A hybrid route is also possible. A preliminary decision may be followed by several independent reviews, with a final decision after those reviews finish.

Ask What Each Reviewer Needs to Know

Write down the input and decision scope for each review. If finance must evaluate terms that legal may change, finance may need the legal result before completing its decision. If an IT reviewer can assess the same submitted specification independently, that review may be able to run alongside the legal review.

Distinguish a necessary dependency from a habit. “Finance has always gone second” does not explain whether the finance decision actually requires an earlier result. Conversely, moving finance earlier without checking its information needs can create rework.

Give every branch an owner, a defined input version, and a recorded outcome. The overall workflow owner needs a clear way to resolve exceptions between branches.

Compare Timing with an Explicit Example

Illustrative calculation: suppose legal needs two hours of active review, finance needs three, and IT needs one. The durations are fictional. First assume all three can review the same complete submission independently.

  • Sequential active review: 2 + 3 + 1 = 6 hours.
  • Independent concurrent reviews: the longest branch takes 3 hours.
  • Hybrid dependency: if finance must follow legal while IT runs independently, the dependent path takes 2 + 3 = 5 hours.

Real elapsed time also includes waiting for people, task creation, transfers, revisions, and completion processing. The arithmetic explains the dependency structure; measure your own request histories to assess the actual result.

Parallel routing does not remove the slowest required review. It also creates little benefit if reviewers lack the information they need or repeatedly ask the requester for different versions.

Define the Completion Policy Separately

Distributing a request to several people leaves the decision policy open. Specify which outcomes are sufficient and which outcomes stop the request.

  • All required approvers: every required role must approve the relevant submission. The rejection policy decides whether to stop immediately or collect the remaining feedback first.
  • Any authorized approval: one eligible positive decision may be sufficient. State how a rejection from another participant affects the result.
  • First response: the first recorded decision determines the outcome, whether it is approval or rejection.
  • Quorum: the request needs a defined threshold of eligible decisions, possibly with additional mandatory roles.

Microsoft’s approval documentation distinguishes several response policies. Use the documented behavior of your own system, especially when combining group assignments with separate workflow branches.

If the same person holds more than one role, establish whether one action satisfies both obligations. Decide this explicitly rather than counting duplicate assignments as independent decisions.

Specify Rejection, Rework, and Cancellation

A rejection needs a request-level outcome and a rule for the other tasks. One policy may stop the request immediately and cancel outstanding reviews. Another may collect all feedback before returning the request. Make the chosen behavior clear to reviewers and requesters.

Keep a return for correction distinct from a final rejection. If the requester submits a revised document or changes an amount, record the new submission version and apply the agreed policy for earlier decisions. Reviewers should be able to tell whether their task refers to the original or revised information.

For a withdrawn request, close or cancel the outstanding tasks through the defined route. Prevent late responses from accidentally moving an obsolete request back toward approval.

Check How the Branches Join

A join cannot wait for work that was never created. If a rule conditionally omits a reviewer, the completion logic must account for that path. Similarly, decide how a cancelled branch contributes to the request-level outcome.

In BPMN-based designs, the gateway type matters. Camunda’s parallel gateway documentation describes synchronization across incoming flows. Use that as a concrete example of why a split and its join must be designed together; verify the semantics of the engine and configuration you use.

At the business level, you should be able to explain the final outcome from the history: which reviews were required, which were created, what each returned, and why the workflow continued or stopped.

Test the Route Before Release

Cases that expose common routing mistakes
CaseWhat to verify
Every required review approvesThe request completes once, after the required results
One reviewer rejects while others are still workingThe documented rejection and remaining-task policy is applied
A condition omits one branchThe workflow does not wait indefinitely for that omitted task
The requester changes the submissionOpen tasks and earlier decisions follow the version-change policy
A required approver is unavailableAn authorized delegate or escalation route is used and recorded
A response arrives after cancellationIt cannot silently reopen or approve the obsolete request

Choose and Implement the Smallest Clear Route

Begin with the required decisions, the information each reviewer needs, and the authority rules. Draw the dependencies, then decide which work can happen independently. Define the request-level outcomes and exception paths before configuring notifications.

CMW Lab describes parallel routing, role-based authority checks, and configurable transition rules in its workflow engine. Validate the route using representative submissions in your product version and deployment. The wider approval process also needs intake, clear ownership, and a defined action after approval.

Frequently Asked Questions

Are parallel approvals always faster?

The potential saving comes from overlapping independent work. Queues, missing information, rework, and the slowest required review still affect elapsed time. Compare similar completed requests using the same start and finish definitions.

Can a workflow combine sequential and parallel approvals?

Yes. For example, a preliminary completeness check can precede several independent reviews, followed by a final authorized decision. Document the dependencies and completion rules at each boundary.

Does parallel mean that the first approver wins?

No. Concurrent task execution and the decision policy are separate choices. The workflow may require all approvals, a quorum, or another explicit rule.

Explore CMW approval management software with your dependency map and the exception cases you need to test.

Get a 30-day trial!

Author: PavelL

Browse more articles on workflows and business processes from this author.