METIS LABS
From 5 to 8% export errors to near zero, across 20 clients.
A scaling Dutch accountancy firm ran 20 e-commerce clients off manual Excel lists, re-keyed every tax period. One self-serve pipeline, opened for inspection below.
Twenty e-commerce clients, every tax period, exported by hand.
The firm handled bookkeeping for 20 e-commerce clients selling on Shopify and bol.com. Each period, the numbers were pulled into Excel lists and switched over by hand into the firm's own format. The work was repetitive, and it leaned on the client doing their part correctly every time.
Manual switching produced an error rate of roughly 5 to 8%. On top of that, each client cost 30 minutes to an hour of back-and-forth per tax period to get the data right. Across 20 clients, that added up to a recurring drag on every close.
Near-zero export errors, and under a minute per export.
We agreed two targets in writing before building: cut the manual-switching error rate to effectively zero, and bring the time per export under a minute. Both mattered because the firm was scaling, and the manual approach got worse with every new e-commerce client added.
A self-serve link, and a pipeline that pulls the same way every time.
We built a two-layer setup. The client links their own Shopify or bol.com environment, with no user account to manage, and the data is pulled out consistently in the format the firm actually uses. That removes the manual switching step, and the errors that came with it.
Two-layer setup: a client-facing linking layer, separate from the firm's processing layer
Why: clients link their own environment without an account, so onboarding a new client does not add manual work.
Pull and normalise the data to the firm's internal format automatically
Why: the manual switching step was where the 5 to 8% errors came from, so we removed the step rather than checking it.
Built so it can integrate directly with the accountant's internal system
Why: exports should land where the work already happens, not in another spreadsheet to be re-imported.
Checked four weeks after handover.
We set out to get under one minute per export and landed at about 1.5 minutes. Still down from up to an hour, and reported as measured rather than rounded to the target.
Because an opened box includes the invoice.
A fixed-scope build, priced and agreed before we started.
Drowning in a process like this?
Tell us the problem and what it's costing you. We'll tell you, plainly, whether software is the answer, and what it takes if it is.
Open a conversation