Back to all articlesBusiness software decisions

Turn Software Requests into Requirements You Can Actually Score

Alactic Editorial9 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

“Better reporting” can put a product on a shortlist without explaining which decision its reports must support. “Easy to use” can eliminate another product without identifying the task that was difficult. When requests stay at that level, the shortlist reflects interpretations rather than agreed business needs.

To write business software requirements you can actually score, connect each request to a task or decision, name its owner, state the operating constraint, and define observable acceptance evidence. Then record how often the need arises and what happens if it is not met. A request without an identifiable owner or consequence should not influence the shortlist until someone resolves those gaps.

The worksheet below turns those principles into a usable requirements register. Its purpose is to settle what fit means before vendor evaluation—not to design fields, build dashboards, or specify an implementation.

1. Start with the work hidden inside the request

A feature request is useful input, but it is not yet a requirement. It usually names a possible solution while leaving the underlying work unexplained.

Ask the requester to describe the last occasion when the missing capability caused difficulty. What were they trying to do? What information did they have? What could they not decide or complete? What happened next?

These questions separate needs that otherwise sound identical. One manager asking for better reporting may need to assign follow-up work. Another may need to explain changes in a monthly forecast. A shared label does not make those the same requirement.

Use this drafting pattern:

When [trigger], [role] must be able to [complete a task or make a decision], within [operating constraint]. We will judge this using [observable evidence]. [Owner] confirms whether the result is acceptable. If it fails, [business consequence].

Keep the statement focused on one outcome. If it contains several independently testable tasks, split it. Otherwise, a product might satisfy the easiest part while receiving credit for the whole request.

2. Give every requirement an owner and a consequence

The owner is the person accountable for accepting the outcome, not necessarily the person who submitted the request. A sponsor may ask for a reporting capability, while the manager using the report is better placed to judge whether it supports the decision.

Choose one accountable owner for each requirement. Other people can contribute constraints or evidence, but a committee of unnamed stakeholders cannot resolve a disputed result.

The consequence should describe what breaks, slows down, or becomes unreliable. Avoid circular statements such as “we will not have better reporting.” Useful consequences include an assignment decision being delayed, a handover losing essential context, or a recurring manual reconciliation remaining necessary.

Frequency provides context, not an automatic priority. A frequent inconvenience may be tolerable. An infrequent task may still be essential if failure blocks a critical operating process.

  • No identifiable owner: return the request for clarification; do not score it.
  • No identifiable consequence: ask what changes if the capability is absent. If nothing meaningful changes, remove it from the decision criteria.
  • Owner or consequence disputed: hold the requirement outside the approved set until the disagreement is resolved.

This is not a reason to discard inconvenient concerns. It is a way to stop unsupported requests from carrying the same weight as understood needs.

3. Filled worksheet: a fictional team translates its requests

Fictional worked example: The following worksheet describes a hypothetical service business replacing its customer-management software. All roles, thresholds, timings, and consequences are illustrative, not Alactic customer results. The evidence describes what the team would later ask a candidate system to demonstrate; it does not claim that any product has these capabilities.

R1 — “Better reporting”

  • Decision supported: The service manager decides which overdue customer follow-ups to reassign at the daily workload meeting.
  • Testable requirement: The manager can identify open follow-ups overdue as of a chosen date, grouped by current owner, with access to the underlying records.
  • Acceptance evidence: Using a prepared fictional dataset containing seven overdue follow-ups, the manager identifies all seven, distinguishes their owners, and opens a selected record without reconciling a separate spreadsheet.
  • Frequency: Every working day.
  • Owner: Service manager.
  • Consequence of failure: Reassignment depends on manual reconciliation, and overdue work may remain with an unavailable colleague.
  • Initial classification: Necessary; the current spreadsheet workaround is scheduled to end.

R2 — “Easy to use”

  • Task supported: A rotating coordinator records an incoming customer request and assigns responsibility.
  • Testable requirement: After the same agreed orientation, the coordinator can record the essential request details and assign an owner without step-by-step coaching.
  • Acceptance evidence: In a fictional trial, the coordinator completes three representative scenarios within four minutes each, with no missing essential details or incorrect assignments. The team defines those details before the trial.
  • Frequency: Repeated throughout the working day.
  • Owner: Coordination lead.
  • Consequence of failure: Requests require correction before work can begin, increasing the coordinator's queue.
  • Initial classification: Necessary, subject to confirming that the time threshold reflects actual operating needs.

R3 — “Everything in one place”

  • Task supported: A colleague taking over an open case decides the next action without asking the previous owner to reconstruct events.
  • Testable requirement: The colleague can locate the latest commitment, current owner, and next agreed action from the case context.
  • Acceptance evidence: Given a fictional handover case, the colleague correctly states those three items and identifies where each was recorded. Merely displaying a long activity history does not pass.
  • Frequency: During scheduled handovers and unplanned absences.
  • Owner: Service operations lead.
  • Consequence of failure: The next action is delayed or contradicts an existing customer commitment.
  • Initial classification: Necessary; the team does not require every type of business information to be stored together.

R4 — “Integrate with finance”

  • Task supported: A finance analyst reconciles closed service cases with the finance team's reference list.
  • Testable requirement: The analyst can obtain a structured extract containing the agreed case identifier, closure date, and billing reference without retyping records.
  • Acceptance evidence: A sample extract preserves identifiers, handles a missing billing reference visibly, and matches the agreed fictional reconciliation set. A finance representative confirms usability.
  • Frequency: Weekly.
  • Owner: Finance operations lead.
  • Consequence of failure: Manual transcription remains part of reconciliation.
  • Initial classification: The usable extract is necessary; a direct connection is a preference unless its absence creates an unacceptable workload.

R5 — “Custom dashboards”

  • Decision supported: The team lead chooses the order in which to review workload categories.
  • Testable requirement: The lead can save a preferred ordering of the agreed workload summary without changing colleagues' views.
  • Acceptance evidence: The lead changes and saves the ordering, returns to it later, and confirms another user's view remains unchanged.
  • Frequency: Viewed daily; changed occasionally.
  • Owner: Team lead.
  • Consequence of failure: The lead uses the standard ordering, with some additional navigation but no blocked decision.
  • Initial classification: Preference, not a shortlist gate.

R6 — “It needs AI”

  • Decision or task supported: Not identified.
  • Acceptance evidence: None agreed; a generated output alone would not establish business usefulness.
  • Frequency: Unknown.
  • Owner: No accountable user identified.
  • Consequence of failure: No specific consequence identified.
  • Disposition: Excluded from scoring. Reopen only when someone identifies a task, owner, and consequence. Do not award points merely for the presence of an AI label.

4. Separate necessities from preferred solutions

A necessity is an outcome the business cannot reasonably operate without under the agreed conditions. A preference improves the experience or reduces effort but has an acceptable alternative.

Ask: “If this were absent, what would we do instead, and would that be acceptable?” Record the answer. A workaround is not automatically acceptable because it is technically possible; it may consume scarce time, create another handoff, or depend on one person.

In the fictional worksheet, “integrate with finance” initially sounds like a demand for a direct connection. The actual necessary outcome is a usable reconciliation extract. If a controlled weekly export meets the need, insisting on a direct connection would narrow the shortlist without a demonstrated reason.

Conversely, do not downgrade an essential requirement because a workaround exists on paper. Identify who would perform it, its expected frequency, and who accepts the burden. If nobody owns the workaround, it is not a credible fallback.

Keep necessities as explicit gates rather than allowing a collection of attractive preferences to compensate for a failed essential task.

5. Specify evidence before assigning scores

Acceptance evidence should let different reviewers reach a similar conclusion. “The report looks useful” is weak evidence. Identifying the expected records in an agreed dataset is stronger because the result can be checked.

For each requirement, prepare an evidence packet: the scenario, sample inputs, expected outcome, permitted assistance, and failure condition. Include an exception where it matters, such as a missing reference or an unassigned record. This remains requirement definition; detailed test execution belongs later.

A recommended scoring convention is:

  • Meets: The agreed outcome is demonstrated under the stated conditions.
  • Partially meets: A specific gap or workaround remains. Record it and obtain the owner's assessment.
  • Does not meet: The outcome cannot be completed within the agreed constraints.
  • Unverified: Sufficient evidence has not been provided.

Keep unverified separate from failure: it describes an evidence gap, not a proven inability. But do not count it as a pass. For a necessary requirement, unresolved evidence means the decision is not ready.

If preferences later receive numerical weights, agree them before seeing vendor scores. Otherwise, the team can unconsciously adjust importance to favour a product it already likes.

6. Repair requirements that fail review

A requirement review should expose weak statements before they become evaluation disputes. Read each item aloud and ask another person to describe what would count as failure.

  • Too broad: If “manage customer relationships” requires several unrelated demonstrations, split it into specific tasks.
  • Too prescriptive: If a requirement names a button, screen arrangement, or technical mechanism, ask whether a different method could deliver the same acceptable outcome.
  • Too permissive: If almost any demonstration could pass, add the expected result and relevant constraint.
  • Arbitrary threshold: If nobody can explain a time or volume limit, mark it provisional and validate it with the task owner before approval.
  • Conflicting needs: If one team wants fewer inputs and another needs more complete records, identify the minimum information required for the downstream decision rather than combining both requests unchanged.

When a requirement changes, retain the reason and obtain owner approval again. If evaluation has already started, apply the revised requirement consistently to every candidate. Do not quietly relax it for one product or tighten it for another.

7. Finish with a small, approved register

For your next requirements session, bring the requests currently influencing the shortlist. Rewrite each using the worksheet fields: task or decision, testable statement, acceptance evidence, frequency, owner, consequence, and classification.

End the session with an approved set and a separate clarification queue. Count how many proposed necessities have both an accountable owner and agreed acceptance evidence. Every necessity should have both before it becomes an evaluation gate; unresolved items stay visibly pending.

The useful output is not the longest feature list. It is a register that makes the next decision defensible: what the business needs to accomplish, what evidence would establish fit, and which compromises someone is prepared to own.

Software requirementsBusiness softwareVendor evaluationDecision-making

Put the next conversation to work.

Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.

Explore Communications