The setup
A garment printing and embroidery firm, around fifteen staff, selling through a set of branded web stores. Orders arrived around the clock; the accounts lived in QuickBooks. Between the two sat a person: every order's confirmation was printed or pulled up on screen, then re-typed into QuickBooks line by line. Customer, products, quantities, VAT, delivery.
I'm not naming the storefront platform, because in this trade doing so would shortlist the client. It's a white-label system built for garment decorators, and the shape of the fix applies to any store-plus-accounts pairing: Shopify to Xero, WooCommerce to QuickBooks, all the usual suspects.
What the day looked like
Re-keying an order took roughly 18 minutes when it went smoothly. With a double-figure count of orders on a normal day, that was most of somebody's job: not a task on the side, a role. And hand-copying has a failure mode that doesn't announce itself. A digit slips in a quantity or a price, nothing complains, and weeks later an invoice doesn't match a payment and someone spends an afternoon working out why.
What I built
A sync that picks each order up the moment it's placed, maps the products, customer and VAT treatment onto the right QuickBooks records, and posts it. No printing, no re-typing, no waiting for a person to be free.
The design decision that matters: the sync refuses to guess. An order it can't map cleanly, an unusual product, a customer that matches two records, a price that disagrees with the store, gets set aside for a person with a note saying exactly what's odd, rather than posted wrong and found at month end.
The numbers
| Measure | Before | After | How it was measured |
|---|---|---|---|
| Handling time per order | ~18 minutes | Under 3 | Timed across a sample of orders before the build; afterwards, the only handling left is the exceptions queue |
| Orders posting with no human touch | 0% | ~83% | The sync's own logs, steady month on month |
| Transcription errors | The odd one, found late | None | Month-end reconciliation stopped turning up mismatches from typing; what reaches QuickBooks is what the store said |
Figures are from the first months after switch-on, and the sync has kept those numbers since. The freed hours didn't leave the building; they went into artwork approvals and chasing quotes, which is where a print firm actually makes its money.
The 17% that still sees a human
I could have chased 100% and I didn't, on purpose. The last slice is the genuinely odd stuff: bespoke line items, part-shipments, customers with duplicate histories. Handling those automatically means teaching the sync to guess, and a wrong guess in your accounts costs more than the three minutes a person spends resolving it properly. That trade-off, where automation should stop, is most of the real design work in a job like this.
What a job like this costs
As a fixed-price milestone this shape of build typically lands between £1,500 and £3,000, agreed as one number before any work starts. Third-party running costs for a sync like this are usually small or nothing. Here's how pricing works.
Can you check any of this?
Fair question, and the honest answer: not from this page alone. Anonymising the client cuts both ways. What I can offer: the method above says exactly what was measured and how, and if we end up talking seriously, I'll arrange what verification the client is comfortable with.
Got orders being typed twice, or any two systems that don't talk? The free half-hour review is where we work out what your version of this looks like.