How to Verify Progress on an Outsourced Software Build
How do you track outsourced software development progress?
Track outsourced software development progress by checking working user workflows against agreed acceptance criteria. Ask for a live demonstration, access to a test version and evidence showing which deliverables passed review. Use hours logged and task counts to understand effort, while keeping unfinished work and unresolved blockers visible.
Your central question is simple: what can a user successfully do now that they could not do before?
What should count as a completed deliverable?
Define deliverables around an outcome you can inspect. “Build the approval feature” leaves too much open. A clearer description is: “An authorised reviewer can approve a submitted request, and the requester can see the updated decision.”
Before work begins, agree on:
- The user and their starting point.
- The action they need to complete.
- The result they should see.
- The relevant permissions and error cases.
- The evidence required for acceptance.
- Any exclusions from this delivery.
Keep the criteria specific enough to produce a pass or fail. “Easy to use” needs interpretation. “A requester can correct a missing required field without losing the rest of their entry” gives you something to try.
Separate delivery from acceptance. The partner can submit work for review; your named reviewer then accepts it or records what remains unresolved. Silence should not become approval unless your agreement explicitly says so.
What should a useful demonstration show?
Ask the partner to demonstrate a complete workflow from its natural starting point. For a request system, that might mean creating a request, submitting it, reviewing it under another account and checking the final result.
Choose scenarios from the acceptance criteria. Include ordinary use and a relevant failure case, such as missing information or an action attempted without permission.
During the demonstration, ask:
- Is this the version available for our review?
- Which steps use working connections to other services?
- Which results are simulated or entered in advance?
- What still requires a manual intervention?
- Can we repeat this with a fresh entry?
A demonstration can legitimately include unfinished pieces. The important distinction is whether those pieces are disclosed and excluded from acceptance.
Screenshots can support a visual review, but they do not establish that a workflow works. A recording provides a reference; access to the demonstrated version lets you check it yourself.
What staging access should you request?
Request access to staging: a test version where you can try the software without affecting live business activity.
Ask for a short guide covering the available workflows, suitable sample data, accounts for each relevant role and known limitations. You should also know whether changes you make will remain available for later review or be cleared.
Check that the version you can access matches the demonstration. If it does not, ask which version is being submitted for acceptance and arrange a fresh review.
Use fictional or approved test data. Agree on any restrictions before uploading business records, and avoid using real customer information merely to make a demonstration look realistic. Include access and data handling expectations in your software partner selection questions.
Staging access is useful evidence of behaviour. It does not, by itself, establish readiness for public use.
What acceptance evidence should accompany each delivery?
Ask for a compact record linking each acceptance criterion to an observed result. You need enough detail to understand what was checked and repeat the important checks.
| Acceptance criterion | Evidence to request | Acceptance decision |
|---|---|---|
| A user can submit a valid request | Repeatable steps and the saved request | Accept if the expected information persists |
| Missing required information is explained | Example showing the message and correction | Accept if the user can recover as agreed |
| Only an authorised reviewer can approve | Checks with permitted and restricted accounts | Accept if both behave as specified |
| The requester can see the decision | Review from the requester's account | Accept if the correct decision appears |
| A failed submission preserves entered information | Demonstration of the agreed failure case | Accept if recovery meets the criterion |
Record exceptions beside the affected criterion. If submission works but recovery fails, describe that result directly.
Treat the partner's testing and your acceptance review as separate responsibilities. Ask what checks the partner completed before presenting the work. For broader context, see software testing and quality assurance.
How should you handle blockers and dependencies?
Ask for each blocker to include the affected deliverable, the missing input, who owns the next action and what happens if it remains unresolved.
“Waiting for feedback” is incomplete. A useful statement is: “The reviewer workflow needs your decision on whether rejected requests can be edited. You own that decision; acceptance of the resubmission path remains blocked.”
Make your own obligations visible too. Delayed decisions, missing sample material and unavailable reviewers belong in the same record.
For each blocker, distinguish between work that can continue and work that cannot. Ask whether a temporary substitute is being used, what it proves and what must be checked again once the real dependency is available.
A simulated response can support an early demonstration. Keep the actual connection listed as unverified until it has been tested.
What does an evidence-based progress report look like?
Use a short report organised by deliverable, with a link or reference to evidence and a clear next action.
Illustrative example: a request approval workflow
| Deliverable | Current state | Evidence | Remaining work |
|---|---|---|---|
| Submit a request | Accepted | Owner repeated the agreed staging checks | None within the agreed scope |
| Review and approve | Ready for review | Demonstration and reviewer access available | Owner acceptance check |
| Return a rejected request for editing | Blocked | Decision needed on permitted changes | Owner supplies the rule |
| Notify the requester | In progress | Message preview only | Delivery and failure handling remain unverified |
This separates accepted work from work that is merely visible.
Keep hours and spending alongside the report when relevant to your agreement. They answer how much effort or budget has been used. The acceptance record answers what that effort has delivered.
Which status reports deserve closer inspection?
Treat these as reasons to request evidence, rather than proof that a partner is acting badly:
- “Almost done” appears repeatedly. Ask for the remaining acceptance criteria and the next reviewable result.
- Many tasks close, but workflows stay unavailable. Ask which user outcome those tasks enable.
- Every demonstration follows the same prepared route. Request a fresh entry and a relevant alternative path.
- “Tested” has no supporting detail. Ask what was checked, what failed and what remains unchecked.
- Staging repeatedly differs from the demonstration. Establish which version is being submitted.
- Blockers have no owner or next action. Turn each into a specific decision or dependency.
Also inspect statements such as “only integration remains.” Ask which user actions depend on that integration and what evidence will establish completion. Judge the remaining work by its effect on the outcome.
What common mistakes should you avoid?
Accepting screens instead of workflows. A form may look finished while saving, permissions or error recovery remain incomplete. Review the agreed journey through to its result.
Changing the criteria during review. Distinguish a missed requirement from a new request. Record new scope separately so both parties can assess its effect.
Requiring every delivery to look publicly launchable. An early delivery can be useful without being ready for customers. Label its purpose and limits.
Equating visible progress with the whole build. Some necessary work has little visual output. Ask what it enables and how completion will be demonstrated.
Approving without recording exceptions. If you accept a limitation, write down its effect, agreed follow-up and whether it prevents launch.
When does workflow-based tracking need adapting?
Discovery, design exploration and technical investigations may finish before there is a working user journey.
For discovery, agree on decisions, requirements and unresolved questions. For a design review, inspect the proposed journey and agreed interaction states. For a technical investigation, request the question tested, evidence gathered, limitations and recommendation.
The principle remains the same: define an observable result before work starts. Do not require a polished demonstration from an investigation, or accept an investigation report as proof that a feature works.
What questions do founders often ask?
How often should we review progress?
Agree on a regular review interval that leaves time to act on feedback. Tie acceptance to reviewable deliverables, and request earlier notice when a blocker threatens an agreed outcome.
Can we verify progress without reading code?
Yes. You can check user behaviour, saved results, permissions and acceptance evidence without reading code. Arrange a separate technical review for questions about maintainability or implementation quality.
Should payment depend on accepted deliverables?
Where your commercial agreement uses delivery milestones, define the evidence and acceptance process alongside them. For time-based work, still track accepted outcomes so spending and delivery remain visible.
What if the partner cannot provide staging access?
Ask for the reason and an alternative way to repeat the acceptance checks. A supervised session may suit a constrained situation. Record what you could verify, what remained inaccessible and how the gap will be closed.