The work, and what it changed.
Platforms, apps and AI systems we've shipped, each built around the people who use it and measured by the number it moved.
89 studies · 9 industries · design to operations
- UninterruptedService continuityHospitality POS that keeps working when th…
- −58%Time to settle a split tabHospitality POS that keeps working when th…
- +34%Chapter completion rateA reader people finish
- +61%Series resumed after a weekA reader people finish
- 125Brands in the catalogueOne number
- −94%Manual catalogue entryOne number
- 0Documents leaving the machineA writing desk where the manuscript never…
- 100%Usable without an accountA writing desk where the manuscript never…
- 3Extraction strategies comparedLong text becomes narrated audio
- 4Pipeline stages surfacedLong text becomes narrated audio
- −88%Cost to add a toolThirty small tools behind one account
- 1Billing implementationsThirty small tools behind one account
- Self-serveInstructor onboardingA course marketplace where the instructor'…
- 1Ledgers under the platformA course marketplace where the instructor'…
- 73%Domain shared across verticalsOne platform
- 1Codebases maintainedOne platform
- 1Codebases for two storesA back office on the desk
- 100%Print failures diagnosable on deviceA back office on the desk
- 0Orders created in an invalid stateA cart that won't check out until the pres…
- −64%Counter calls asking for statusA cart that won't check out until the pres…
- −91%Affiliate balance disputesA storefront in four languages
- 4Markets servedA storefront in four languages
- 1Entitlement systemsTwo billing models in one product
- 0Codes redeemed twiceTwo billing models in one product
- 3 levelsCatalogue depthA marketplace where the vendors list
- 0Merchandising changes needing a deployA marketplace where the vendors list
- 1Places an access decision is madeA bought console
- 0UI actions that fail server-sideA bought console
- CoverageMatching modelMatching by who covers the address
- 0Jobs offered outside a provider's areaMatching by who covers the address
- 4Record types modelledJobs
- 4+Sources replacedJobs
- NoneBehaviour changedThe same app
- 0Screens touching transportThe same app
- Group + areaMatching inputsThe right blood type isn't enough: it has…
- 3Area levels modelledThe right blood type isn't enough: it has…
- 3Content types modelledThree kinds of publication
- 0Comments auto-decidedThree kinds of publication
- 1Applications to runThe person who sells is the person who sto…
- 0Systems to syncThe person who sells is the person who sto…
- 1Steps to settle a marketMarking the winner has to settle every ope…
- 0Partially settled marketsMarking the winner has to settle every ope…
- 1Systems to raise an orderThe order started as a DM
- 1Actions to link a customerThe order started as a DM
- SharedProduct identityTwo shops selling the same thing were inco…
- Per shopPrice and stockTwo shops selling the same thing were inco…
- 0Taps to add a scanned lineThe counter doesn't have the seconds a sea…
- ScannerScreen the app opens onThe counter doesn't have the seconds a sea…
- 1Codebases maintainedPurchase and sale are the same ledger seen…
- 2Verticals servedPurchase and sale are the same ledger seen…
- 3Catalogue axesA medicine isn't a SKU
- 0Equivalence lists maintainedA medicine isn't a SKU
- 4Taps to dispense, first buildWe built the same pharmacy counter twice o…
- 1Taps to dispense, secondWe built the same pharmacy counter twice o…
- 43Models, full productThe small-shop edition is the same product…
- 17Models, small editionThe small-shop edition is the same product…
- 0Hand-typed joinsThe invoice and the trip it pays for were…
- AllBank accounts reconciledThe invoice and the trip it pays for were…
- 1Implementations of tenancyEvery SaaS they started rebuilt tenants
- NonePer-query tenant filtersEvery SaaS they started rebuilt tenants
- AllSpreadsheets replacedThe people doing the accounts are the peop…
- 6API modulesThe people doing the accounts are the peop…
- 1, alwaysChecks per writeHiding a button isn't access control
- 0Rows sent to hideHiding a button isn't access control
- NoDangling references possibleA branching story is a graph
- NoneMedia uploadedA branching story is a graph
- 0List reloads on filterFiltering navigated away from the list
- NeverScroll position lostFiltering navigated away from the list
- 3Tabs per recordOne long form is why the device records we…
- 1Tabs needed to createOne long form is why the device records we…
Tour OperatorsA trip assistant that still answers with roaming switched off
A companion app for a tour operator's guided trips, whose assistant answers questions about a traveler's own itinerary on-device. The moment people land abroad they turn data roaming off, and most apps go dark exactly when they're needed.
- -38%
- In-trip calls to guides
- All of them
- Itinerary answers available offline
- 4.7★
- App store rating
Mobile & WebAI Integration12 weeksRead the study
E-commerce & RetailPoint the camera at a broken part, find it on the shelf
A cross-platform app that identifies a plumbing or electrical part from a photo and points to the aisle it's in, running the recognition on the phone itself, because the stores have almost no signal past the front doors.
- -46%
- Wrong-part returns
- Works
- Identification with signal off
- Under 4s
- Median time to name a part
Mobile & WebAI Integration14 weeksRead the study
Professional ServicesThe research that told a roadmap 'not that' before six months were spent
Jobs-to-be-done interviews and demand signals redirected a planned vendor-risk module toward the problem customers actually had, expense reconciliation, before a single sprint was spent building the wrong thing.
- 10 of 12, unprompted
- Interviews that surfaced the real pain point
- ~2 quarters
- Engineering time redirected before being spent
- Expense reconciliation, underserved
- Market gap identified
AI Consulting5 weeksRead the study
Professional ServicesThe access-control flaw that would have let one client see another's payroll
A pre-launch security audit found a broken-access-control flaw that let a crafted request pull another tenant's payroll records, rated and fixed before the product ever reached a customer.
- 1 (cross-tenant data exposure)
- Critical findings surfaced
- 14
- Total findings rated and prioritized
- Yes, 4 days before go-live
- Critical finding closed before launch
Managed ServicesEnterprise2 weeksRead the study
Media & PublishingThe inbox full of technical requests nobody could prioritize
A shared inbox of technical requests, with no visibility and no priority order, was replaced by a tracked, triaged queue. The first urgent production fix jumped the line the same day it was submitted.
- Same day
- Time to triage a new request
- Weeks → zero
- Average time a request stays untracked
- Resolved same day, jumped 11-item backlog
- First urgent fix under the new process
Managed ServicesMobile & WebOngoing retainerRead the study
E-commerce & RetailThe outage where the homepage looked fine and checkout was dead
A checkout-specific health check caught a payment-provider integration failure within two minutes, while a shallow homepage ping would have shown everything as healthy for hours.
- Under 2 minutes
- Time to detect the outage
- Under 5 minutes
- Time to reach on-call
- ~3 hours of full checkout downtime avoided
- Estimated sales saved during the sale
Managed ServicesMobile & WebOngoing retainerRead the study
Professional ServicesThe offboarded account that tried to log in eleven months later
Anomaly monitoring flagged an access attempt from a departed employee's still-active account before anything was touched, and the access review that followed closed nine other stale permissions nobody had noticed.
- Under 4 minutes
- Time to flag the anomaly
- 10 accounts
- Stale permissions closed
- 100% triaged weekly
- Known CVEs now on patch cadence
Managed ServicesEnterpriseOngoing retainerRead the study
Professional ServicesThe double-charge bug that only ever showed up in production
A rare double-charge complaint traced back to a race condition between two webhook deliveries: reproduced from production logs, fixed, and locked down with a regression test.
- Recurring → zero since fix
- Double-charge reports
- 18 months of intermittent reports → 4 days to isolate
- Root cause found
- Concurrent webhook race now tested
- Regression coverage
Managed ServicesEnterprise3 weeksRead the study
Professional ServicesThe release gate that turned a green pipeline into a real promise
A layered automated test suite and a CI release gate replaced a shipping process built on crossed fingers, and a flake-cleanup pass made a red result mean something again.
- 0 → 14 in first month
- Regressions caught pre-release
- 31% → under 2%
- Flaky test rate
- Retired
- Manual pre-release click-through
Managed ServicesEnterprise6 weeksRead the study
Professional ServicesFrom a Thursday-evening deploy ritual to shipping 40+ times a week
A containerized build and a gated CI/CD pipeline replaced a manual, once-a-week deploy ritual, with a one-click rollback that turns a bad release into a non-event.
- 1 → 40+
- Deploys per week
- Hours → under a minute
- Rollback time
- 0
- Weekend on-call deploys
Managed ServicesEnterprise7 weeksRead the study
Healthcare & Dental PracticesA content hub that keeps 40 franchise sites on-brand without a bottleneck
A headless CMS where corporate manages shared brand content centrally while each of 40 franchise locations publishes its own class schedule and local promotions without touching code.
- Days → minutes
- Time to publish a local update
- 40 of 40
- Locations self-publishing
- −85%
- Corporate webmaster queue
Mobile & Web9 weeksRead the study
News & Digital PublishersOne content model, three brand sites, zero duplicate entries
A headless CMS with one structured content model feeding three separately branded publications through the same API, so an article is written once and syndicated everywhere it belongs.
- 0
- Duplicate content entries
- −90%
- Cross-site correction time
- 14
- Editor hours saved weekly
Mobile & WebEnterprise10 weeksRead the study
Media & PublishingA desktop tool that keeps client footage off the network entirely
A native desktop tool for organizing and pre-processing raw client footage entirely on-device, with a signed installer for Windows and macOS so editors never have to upload sensitive footage to a server first.
- −4 hrs avg
- Project start time
- 0
- Footage leaving studio machines pre-delivery
- 100%
- Editor workstations covered
Mobile & Web8 weeksRead the study