Small business grants product comparison: testing discovery, portals and advisory packages

Grant products fall into distinct categories: official search services, application portals, research subscriptions, collaborative software and advisory packages. Compare products that solve the same task. A single score across all five categories would hide more than it explains.

What to take away

Compare grant products only within the same task category, because a single score across all five hides more than it explains.
For discovery services, measure whether results lead to the controlling funder page and verify each scheme.
For research subscriptions, count true relevant results, stale records and missed matches rather than total listings.
For advisory packages, score public claims and sales responses separately from paid sample work and observed delivery.
Use not tested for account-only features, support response, renewal negotiation, live submission and long-term reliability.

Discovery services

Compare coverage, geography, filters, source links, update dates, alerts and export. Check a sample of known live and closed opportunities. Measure whether the result leads to the controlling funder page.

Find a grant and the Department for Business and Trade finance and support finder are official discovery services with different coverage and interfaces. Neither proves eligibility, so verify each scheme.

Application portals

The funder normally chooses the portal. Compare usability only where the applicant has a genuine choice; otherwise prepare to use the required service. Test access, collaborator roles, saving, attachments, validation, export and acknowledgement.

UKRI's Innovate UK guidance says relevant applications must use the Innovation Funding Service. It also assigns duties to lead applicants and partners. An internal product must reconcile with that final record.

Research subscriptions

Test opportunity relevance, source accuracy, update speed, exclusions and cancellation. Give each provider the same applicant profile and known opportunities. Count true relevant results, stale records and missed matches rather than total listings.

Ask whether a human or automated method checks deadlines and eligibility. Open the official source before acting.

Collaborative software

Use a realistic application with long text, figures, attachments, several reviewers and a late change. Compare permissions, history, approvals, export fidelity, backup, deletion and support. Record the exact plan and configuration tested.

Review security through the NCSC small organisations guidance and personal-data handling through the ICO guidance hub.

Advisory packages

Compare defined outputs, named team, applicant work, revisions, timeline, fee, success trigger, post-award support and contract. Use the same project brief and ask providers to identify poor fit and missing evidence.

A proposal is not a completed service test. Score public claims and sales responses separately from paid sample work, references and observed delivery.

Total cost and exit

Calculate setup, licences, users, services, expenses, VAT, renewal and transition. Model no award, reduced award, delay and full award. Check export formats, notice periods and deletion.

Build a reproducible test

Create a sample applicant with fixed facts, budget, deadline, collaborators and missing evidence. Give each comparable product the same inputs. Record the plan, configuration, browser or device, date, tester and time spent.

For discovery, measure relevant current opportunities, false matches, closed records and source accuracy. For drafting, test permissions, conflicting edits, long answers, calculations and export. For advisers, assess whether the response challenges poor fit, identifies missing facts and keeps the applicant in control of declarations.

Do not enter real personal or confidential data into a trial without an approved basis and suitable terms. Preserve screenshots or exports where permitted and redact sensitive material before editorial review.

Report unknowns

Use "not tested" for account-only features, support response, renewal negotiation, live submission and long-term reliability that the method did not observe. Do not turn a missing public price into a cost score.

Separate a defect from user error and configuration. Ask the provider to explain a reproducible failure, then record whether it was fixed during the test period. A factual correction can change the evidence, but a provider cannot purchase a higher judgement.

Recheck before buying

Obtain a current quotation, contract, privacy terms, subprocessor list and service commitment for the selected plan. Confirm the contracting entity and bank details. Repeat any mandatory test after material product or scope changes.

Set mandatory conditions before scoring and disclose payments or free access. This draft compares product categories, not named commercial products. It has no live internal links and needs direct like-for-like tests before any product verdict is published.

Before you act

Compare products that solve the same task.
Test access, saving, attachments, validation and acknowledgement.
Give each provider the same applicant profile and known opportunities.
Record the exact plan and configuration tested.
Calculate setup, licences, users, services, expenses, VAT, renewal and transition.
Obtain a current quotation, contract, privacy terms and subprocessor list.

Common questions

How should I test discovery services?

What should I check for collaborative software?

How do I build a reproducible test?