Software and growth systems for small teams
Fix broken API integrations
An integration used to work and now records are missing, duplicated or out of date. Restarting it may hide the symptom while creating more uncertainty about which transactions completed. I investigate the failure, establish the affected records and repair the connection with a reconciliation and recovery plan.
An agreed first improvement milestone starts from £1,200. Further work is scoped separately after reviewing the existing application.
What the finished solution should accomplish
Repair failing API integrations, identify missing or duplicated records and add the recovery checks needed to keep the connection usable.
What this could look like in practice
A token expiry could stop new orders reaching operations while the sales system continues accepting them. The repair needs to restore access and determine which orders were missed. Replaying everything without checking destination identifiers could create duplicates. Diagnosis and recovery must be treated as parts of the same job.
Decisions to make before building
Preserve logs and establish the last known successful run before clearing queues or rotating credentials. Check provider changes, permissions, payload differences and rate limits. Agree a safe replay window and record-matching strategy. Add monitoring that distinguishes an empty but healthy run from a failed run with no output.
When a simpler option is enough
If a native connector is failing, check its documented status and support route first. Reauthorising an account or correcting a mapping may solve the issue without a rebuild. Custom repair is useful where the fault is in owned code, data transformation or recovery behaviour the standard connector cannot provide.
An initial scope could include
- Failure diagnosis and impact assessment
- Scoped repair with safe replay
- Reconciliation report and monitoring improvements
How we would check the work
- Reproduce the fault before claiming it is fixed
- Replay a known sample without duplicate writes
- Reconcile the affected time window with the source
Can you guarantee every missed record can be recovered?
Recovery depends on source history, identifiers and available logs. I establish what can be verified and identify any remaining gaps. Where the source no longer retains a record, the repair report should say so rather than imply a complete recovery.
Scope, ownership and working together
You work directly with Joel from discovery through implementation. We agree the outcome, exclusions and acceptance checks before a build. You receive access to the code, deployment and handover documentation. Hosting, third-party services and ongoing support are identified separately; owning the code does not make operating costs disappear.
An agreed first improvement milestone starts from £1,200. Further work is scoped separately after reviewing the existing application. Permissions, migration, source-system access and the number of workflows affect scope. Hosting, maintenance and further milestones are agreed separately. Smaller extensions to existing infrastructure can receive a smaller quote.
Related next steps
Have a problem worth solving?
Tell me what you want to change, what is blocking it and the budget you have in mind. We can establish fit and agree the next step by email.
Send a project brief