How to run a trial with a delivery partner

A useful trial with an accounting delivery partner has a manageable scope and clear measures. Pick similar files in one software, or one recurring task across several clients. Agree the review standard, query workflow, fee and billing basis, exit terms and delivery dates in writing. Track quality, review time and handovers across the agreed batches before deciding whether to widen the scope.

Why a trial, and why this shape

Every promise a delivery partner makes about quality, turnaround and fit is unverifiable until it has done your work in your systems to your standard. A trial converts the promise into evidence at a known cost. The shape matters: a trial that is too big cannot be reviewed properly; one that mixes file types produces lessons that do not transfer; one without a written standard produces review points that are really disagreements about what the standard should have been; and one without a measure ends in a feeling rather than a decision.

Week zero: set it up

Choose the batch

• For example, choose ten to twenty clients of one type in one software, if your reviewers can assess that volume: Xero sole traders, small companies with March year ends or monthly BAS clients.
• Or one recurring task across many clients: VAT workings for every quarterly client this quarter, or the bank reconciliations for the whole bookkeeping desk this month.
• Clean records, not the shoeboxes. The trial tests the partner's process, not its patience.

Write the standard

One or two pages: which reconciliations, which schedules, how queries are presented, what the cover note contains. Use your existing template if you have one. If not, ask the partner for its default and edit it. This document is the single biggest predictor of a good trial.

Agree the workflow and the dates

• Query workflow: to your reviewer by default; or via a shared mailbox in your name.
• Delivery date for each batch, and a review-point response time that both sides can meet.
• Named contacts: one on each side.
• Access: named logins on your systems with two-factor authentication, minimum roles.
• Terms: the trial fee and billing basis, confidentiality, client-contact restrictions for firms, and what happens to access and working papers if you stop.

Weeks one to two: the first batch

The first batch tests the written standard against real files. Log each review point and categorise it as an error, a preference or something missing from the standard. Update the standard where it was incomplete. The number of points will depend on the files and the review; record the actual baseline rather than assuming a normal range.

Also note the things that are not points: did the batch land on the date? Was the cover note useful? Were the queries the ones you would have raised? Did the partner ask sensible questions during the batch rather than guessing?

Weeks three to four: the second batch

Compare the second batch with the first. Are the same errors or preferences recurring? Are new points caused by new file types or an incomplete standard? Discuss the pattern with the partner and update the checklist where needed. Improvement matters more than an arbitrary target number.

Check the calendar again. Two batches on their dates is a pattern. One on and one off is a conversation.

Weeks five to six: the third batch and the decision

At the agreed decision point, compare delivered dates, error and preference points, review time, fee and unresolved queries with the criteria set before the trial. If the arrangement works, consider more clients of the same type before adding harder files. If it does not, decide whether to change the scope, run another measured batch or stop under the agreed terms.

What to measure

Use the table as a blank scorecard. It does not predict a partner's results.

What a trial is not

• Not an undefined commercial promise. Agree the trial fee, billing basis and exact batch in writing, whether the trial is paid or discounted.
• Not a test of the partner's patience with your worst files. Those come later, once the process is proven.
• Not a substitute for the security and confidentiality terms. Sign them before batch one, even for a trial. Client data is client data.
• Not a way to avoid writing the standard. If the standard does not exist, the trial's first job is to produce it.

What to investigate when a trial stalls

• The standard was not written down. Record the expected reconciliations, open-point format and handover before judging a batch.
• The batch was too big or too mixed. Narrow the file types so the reviewer can assess the work consistently.
• Feedback did not carry forward. Compare repeated points with the prior batch and ask what changed in the checklist or process.
• The review was too light. A trial needs enough review to distinguish errors from preferences.
• Dates slipped without an explanation. Record the cause, agree a correction and check whether later handovers improve.

After the trial

If you widen, keep measuring. Record review points, reviewer time and batches delivered on their agreed dates. Continue reviewing each file to the standard your firm requires; summary measures help you spot trends but do not replace that review. Add file types one at a time, update the standard for each, and revisit capacity against the filing calendar. Keep the exit terms agreed at the start.