Women founder funding service standards as a template

A service standard is only a promise until it can visibly fail. Written as a control, with a measure, an owner and a recovery step, it tells the team when a promise is slipping and what to do next. The wider service operations work sets the mandate; the controls below make its promises testable.

What to take away

Write each standard with five fieldspromise, measure, owner, evidence and recovery.
Avoid promises such as fast, inclusive or expert unless the team can define, test and report them.
Set warning levels before breach, so action is possible while the promise can still be kept.
Report the median and the slower tail for response times, not just an average.
Publish changes with an effective date and never rewrite past performance under a relaxed definition.

The five fields of a control

A control has five fields. The promise says what the founder can expect, the measure says how it is counted, the owner names who answers for it, the evidence says which record proves it and the recovery says what happens when it is missed.

A source standard shows the pattern. It can promise a recheck before communication, measure sampled compliance, name the evidence editor, store the source date and prescribe correction when a check is missed. Every other standard in the register follows the same shape.

Standard	Promise	Measure	Warning level	Recovery
Accuracy	Official URL and checked date on every record	Sampled compliance with the recheck	Record near its expiry	Correct the record and identify affected users
Response	Acknowledgement and substantive reply by channel	Median and slower tail of elapsed time	Ageing queue	Give an update and new date before the commitment expires
Access	The whole journey is usable	Failures found in end-to-end testing	Adjustment request with no owner	Offer a working alternative route
Complaints	Acknowledged and investigated fairly	Complaints acknowledged within the stated time	Complaint approaching its limit	Escalate to the named owner
Continuity	Owner and backup for each critical task	Result of an exercise with dummy records	Owner or supplier unavailable	Switch to the manual route

Accuracy as a control

Accuracy is easy to test because every record can be inspected. Each opportunity record needs an official URL, checked date, geographic scope, status and owner. High-risk fields such as eligibility, amount, cost, security, ownership and deadline receive an independent check.

Opportunity record required fields

Official URL
Checked date
Geographic scope
Status
Owner
Independent check for high-risk fields

Set the expiry by volatility. A deadline record may require same-day confirmation, while a stable explanation can have a scheduled review. The GOV.UK business finance finder is one discovery route, but the named provider's live terms settle a decision.

Named UK programmes are reference points, not evidence that a route is currently open. Innovate UK's Women in Innovation awards are women-focused, while the British Business Bank's Start Up Loans programme is a general start-up finance route. Check the current programme owner's terms and eligibility before listing either.

Response as a control

Publish acknowledgement and substantive-response targets by channel as elapsed-time commitments, each with an explicit warning threshold. Set them from service capacity and observed performance rather than borrowing an unsupported universal figure.

Response handling sequence

Publish acknowledgement and response targets
Measure median and slower tail
Triage urgent external deadlines separately
Update and new date before commitment expires

Measure the median and the slower tail, not just an average, because an average can hide the founders who waited longest. Give urgent external deadlines a separate triage path, and state that contacting the service does not pause a funder's clock.

Do not invent one universal deadline for every case. Complexity, adviser dependence and provider response vary. Give an update and new date before the original commitment expires. Clocks run across every step of a triage to handover workflow, so name the step each target covers.

Access as a control

Test access across the whole journey, not on one page. The GDS accessibility introduction calls for accessibility from the start, disabled people in research and both manual and automated testing for government services.

Access journey test points

Supported channels
Assisted completion
Adjustment handling
Mobile and low-bandwidth behaviour
Keyboard use
Readable errors
Save-and-return
Emails, documents and calls

Other operators can use those practices as a quality reference while obtaining advice on their own legal duties. The measure is the number of accessibility failures found when emails, documents and calls are tested as well as web pages. Put a warning level on any adjustment request that has no owner, because that is the point where a founder's deadline starts to depend on luck.

Complaints as a control

Complaints are a measurable promise because there is a clock. The ICO guidance on handling data-protection complaints states that you must acknowledge receipt within 30 days. Investigate fairly, and set an earlier internal target if the team can sustain it.

Handle personal data under the UK GDPR and Data Protection Act 2018, collect only what is needed, restrict access and apply a retention schedule. Publish routes for correction, rights and complaints, including how a founder can raise a data-protection concern with the service and when they may take it to the Information Commissioner's Office.

Regulation can add to this. Where the service is an FCA-regulated firm, the Financial Ombudsman Service may also apply to a complaint, so take advice before you promise a complaints route. Keep the internal standard at least as strict as any external route that applies.

Continuity as a control

Name an owner and a backup for source updates, enquiries, incidents and deadline submissions. Keep a manual route for critical tasks during supplier or platform outages. Exercise the plan with dummy records before a live cohort, so the first test of the backup is not a real founder's deadline.

Standards must also cover third parties. Put response, security, accessibility, source update and exit requirements into contracts where a supplier controls them. Maintain a founder-facing workaround if the vendor's timetable is slower than the service commitment. Otherwise a supplier's slow week becomes the founder's missed deadline.

Set warning levels before breach

An ageing queue, unassigned adjustment or expiring source should trigger action while the promise can still be kept. A dashboard that reports failure afterwards is useful for learning but not prevention. Each warning level should name the person who acts and the step they take.

Track source errors, completion, abandonment, response distribution, accessibility failures, adjustments, complaints, reversals and founder effort. Show the denominator and missing data, and segment only where lawful and statistically responsible.

Change and review the standards

Publish changes with an effective date. Do not rewrite historical performance under a newly relaxed definition. Keep the previous measure and explain breaks in the series, so users and managers can understand whether service quality actually changed.

Review standards quarterly and after serious failures or service changes. Tighten them when performance is stable; change the process when targets are repeatedly missed. The standards remain internal promises until ownership, capacity and independent testing are confirmed.

Before you act

Write each promise with a measure, owner, evidence and recovery.
Name a warning level for every standard before breach.
Publish response targets by channel and report the median and slower tail.
Test emails, documents and calls as well as web pages.
Give each critical task an owner and a backup.
Record every change to a standard with an effective date.

Common questions

What goes in the recovery field?

The action taken when a check or target is missed. For accuracy it is correcting the record and identifying affected users. For response it is giving an update and a new date before the commitment expires.

Why set a warning level before a breach?

So that action can happen while the promise can still be kept. An ageing queue, an unassigned adjustment or an expiring source are examples of signals that prompt a response before a founder is let down.

Can a standard be relaxed after a miss?

Only openly. Publish the change with an effective date, keep the previous measure and explain the break in the series. Historical performance is not rewritten under the new definition.