Book a call
Book a discovery session contact@pinnaclecodelabs.com

</> Proof of work

Evidence, not adjectives.

Every engagement gets written up the same way: the context, the disconnect we found, what we built, and what actually changed. No adjectives — just what moved.

The pattern

What connected
systems produce.

Across engagements, the same four things move. They are second-order effects of removing friction, not features we sell.

Lower infrastructure run cost
53%
Faster release cadence
3×
Disconnected systems onto one spine
11
Post-launch cover, included
30 days

Selected work

Three systems,
joined up.

Anonymised at the client's request. Each follows the same shape — systems that were correct on their own, and silent with each other.

Multi-location operator

POS, bookings and CRM on one data spine

Reporting cycle
4 days → 20 min
Systems integrated
7
Manual reconciliation
−90%
Engagement
5 months

Context

A venue business running dozens of sites, each with a point-of-sale system, a booking flow, a waiver platform and a marketing stack — all bought separately, all correct in isolation.

The disconnect

Finance rebuilt the same numbers by hand every month because no two systems agreed on what a booking was. Marketing could not see who actually turned up. Nobody trusted the dashboard, so decisions waited for the spreadsheet.

What we built

A canonical event model and an integration spine: POS, bookings, waivers and CRM publishing to one warehouse, with reconciliation reports that prove the totals match. Financial reporting and the booking application were rebuilt on top of it.

Result

Month-end reporting lands each morning instead of four days later. The booking flow finally reflects reality, and the marketing team works from the same numbers as finance.

Marketplace platform

Re-architected for ten times the demand

Availability
99.98%
Run cost
−41%
p95 latency
−62%
Engagement
4 months

Context

A two-sided marketplace whose traffic had grown faster than its original monolith was designed for, with peaks driven by campaigns nobody could predict a week out.

The disconnect

Every spike became an incident. Scaling meant buying larger machines, so growth and cost rose together — which made the commercial team avoid the campaigns that worked.

What we built

Queue-backed workers for everything that did not need to be synchronous, autoscaling behind clear SLOs, a caching layer with honest invalidation, and a cost model the finance team could forecast from.

Result

Peaks are now uneventful. Cost per transaction fell as volume rose, which changed what the business was willing to try.

Manufacturer

Legacy ERP wrapped, not replaced

Time to launch
6 mo → 7 wks
ERP changes
Zero
Integrations
5
Engagement
3 months

Context

A manufacturer with a decade-old ERP at the centre of everything — heavily customised, poorly documented, and the system of record for stock, pricing and orders.

The disconnect

Every new product idea started with “we would have to change the ERP”, so most of them stopped there. A full replacement had been quoted at two years and was quietly dreaded.

What we built

An API façade over the ERP with a clear contract and read models for the things the business needed most, plus a small ordering application built entirely against the façade. The ERP was never modified.

Result

The new channel launched in seven weeks. The replacement decision is still open — but it is now a choice rather than a prerequisite.

What would you
measure?

Tell us the number that is stuck and we will tell you which connection is holding it. Free, and useful either way.

Book a discovery session contact@pinnaclecodelabs.com