Our API integration services connect the software your business already pays for, so data moves between systems by itself instead of being exported from one and typed into the next. We have connected accounting packages, CRMs, payment providers, online shops and AI services for UK businesses since 2021.
The Cost You Are Removing
Somebody exports a file, opens another system and types it in again. Every week, sometimes every day. Nobody budgets for it, and it grows with the business: more orders mean more retyping, and every repetition is another chance to put a wrong figure in the wrong place.
An integration has the systems talk to each other directly. A typical first project is an order system and an accounting package that share nothing, where someone in finance spends Monday morning re-entering last week’s invoices by hand.
What You Get
- A working connection between your systems, in production, with every message logged
- Webhooks where the provider offers them, plus a scheduled reconciliation behind them so a missed update is caught by a check before a customer notices
- Monitoring that alerts us when the integration stops or falls behind
- Written field mapping covering which field goes where and in what format, down to units and time zones
- Duplicate and retry handling, so the same event arriving twice never creates two orders
- Documentation your team, or another developer, can work from later
The APIs We Work With
We have built against the Xero, HubSpot and Stripe APIs, along with Sage, Microsoft 365, Google Workspace, Shopify, WooCommerce, Klaviyo, Twilio and OpenAI. Our API integrations overview lists every platform with its own page, including what each API can and can’t do. If yours isn’t there, its documentation matters more than our familiarity with it: an unfamiliar API with good docs is quicker work than a familiar one without them.
What Decides Whether This Is Easy or Hard
The worst API in the chain sets the pace. A well-documented API with good coverage is predictable work. One that is undocumented, read-only, rate-limited or changes without notice can turn a fortnight into a month, and no amount of skill on our side can use data the provider doesn’t expose.
So before any estimate is worth taking seriously, we establish what each system can do. Can it be read from, can it be written to, does it send a notification when something changes, and how often can it be called? A supplier who quotes without checking those four things is guessing.
Push or Poll
There are two ways for one system to learn that something happened in another, and your customers can tell the difference.
Polling asks repeatedly whether anything has changed. It is simple, works with almost any system, and is always slightly behind. Check every fifteen minutes and an order can sit unnoticed for fifteen minutes.
Webhooks push a message the moment something happens. They are immediate, and they can arrive twice, out of order or not at all.
We use both. Webhooks make things happen straight away, and a periodic reconciliation compares the two systems and picks up whatever went missing, since some messages sent over the internet never arrive.
Connector or Custom
Connector tools such as Zapier or Make are quick to set up and need no developer. They suit simple flows between systems that already have ready-made integrations. They charge per task, which adds up at volume, and complex logic built inside them gets hard to follow and harder to maintain.
A custom integration costs more up front, has no per-task charge, handles the awkward cases properly and lives alongside the rest of your system.
Our usual advice is to prove the flow with a connector, learn what it has to handle, and build it properly once the volume justifies the cost. Zapier vs a custom integration sets out where each one holds up. If a connector is doing the job well at your volume, keep it; you don’t need us for that.
What Moves the Price
API quality on both sides is the largest variable, as above.
Direction comes next. A one-way sync is roughly half the work of a two-way one, and two-way raises a question only you can answer: when both systems change the same record, which version is right?
Then the data models. Matching fields is quick, but reconciling two systems that disagree about what a customer is takes design time. Volume mostly affects error handling and retries, since hundreds of records a day behave differently from hundreds of thousands. Your tolerance for failure matters as well. An integration that can be down for a day without harm costs less to build than one that can’t.
An Integration We Run
For a lead-generation business we send the outcome of each lead back to the ad platforms that sourced the click. When an enquiry becomes a quote or a sale, the platform hears about it and bids on what turned into revenue, where before it only saw form fills.
Each platform wants the event in its own format. A send that fails gets retried, and a retry mustn’t be counted as a second sale, so every event carries a key the platform uses to discard duplicates. Each platform has its own adapter behind one shared interface, and adding another platform means writing one more adapter.
Where Integrations Break
By a distance, the most common failure is that nobody watches them. An integration that has stopped looks exactly like one with nothing to do, and without monitoring you hear about it from a customer.
Providers change things, sometimes without notice, so anything critical needs a periodic check against the source. Undecided duplicate handling turns one event into two orders; it is the most common integration bug and entirely preventable. Field mapping assumptions (dates in the wrong order, pounds versus pence, a field optional on one side and required on the other) produce wrong data instead of errors, which takes far longer to spot.
Questions to Ask Whoever Builds It
- Have you read both APIs, or are you assuming? Ask what each one cannot do.
- How will we know if it stops working? “Somebody notices” is not an answer.
- What happens if the same message arrives twice, or one side is down?
- Where are the field mapping assumptions written down?
- At our volume, what will this cost per month to run?
A Brief You Can Send Anyone
Whoever you ask to quote, these eight points will get you a better answer than a general description:
- The systems to connect, with names and versions
- What you do by hand today, step by step
- How often it happens and how long it takes each time
- Whether data needs to flow one way or both
- What “a record” means in each system, and where they disagree
- Volume, in records per day or month
- The data that must never be lost
- How quickly a change needs to appear on the other side
How We Build Them
Every integration we build logs each message it sends and receives, so investigating a problem months later means reading a record instead of guessing. For a growing UK retailer we connected Shopify orders, warehouse stock and Xero, which ended the double entry and the risk of selling stock they didn’t have (the retailer case study). For another client we pulled five separate platforms into a single operations dashboard through their APIs, replacing a morning routine of five browser tabs.
Talk to Us About an Integration
Send us the export you keep retyping and the names of the systems involved, and we will come back with what each API allows and roughly what the work involves. Get in touch, or browse our full list of services if the job is bigger than a single connection.
