Case studies

Eight-figure innovation budgets; production traffic near zero

Boards demanded "blockchain strategy." Teams stood up Hyperledger fabric networks to track assets that already had SQL audit tables. Pilots demoed well at conferences; legal, GDPR, and throughput killed every path to production except press releases.

Estimated €500M+ over seven years before abandonment

Lidl bet that a standardized SAP Retail rollout would unify European operations. After years of customization wars, performance cliffs, and store-level reality colliding with vendor templates, the discounter killed the program and went back to the bespoke stack that actually understood €0.29 yogurt pricing.

Multi-year, multi-billion IT integration; branch and core systems lagged for a decade

The Postbank acquisition was sold as scale. In practice it was two retail banks, two core stacks, and one integration program that outlived executives, consultants, and several "target architectures." Customers saw outages, duplicate accounts, and branch chaos while slide decks promised a unified platform Real Soon Now.

Weekend cutover rollback; months of dual-run and incident debt

A clinical SaaS vendor with real revenue decided the monolith was "blocking scale." Leadership scheduled a flag-day migration to dozens of services, Kubernetes, and new data paths — because Netflix slides said that is what grown-ups do.

3× infra spend; velocity flat for two years

Eighty engineers, one product, four customers who cared about SSO and invoices. A new VP imported Netflix slides. Kubernetes arrived. So did forty microservices, distributed tracing bills, and standups about contract versioning instead of features the sales team could demo.

Hours of nationwide payment and branch downtime; remediation and audit costs in the tens of millions

Enterprise Service Bus was sold as loose coupling. In production it became the one throat to choke: every payment, account lookup, and branch terminal route flowed through a clustered middleware layer nobody fully understood until it fell over on a Tuesday.

18-month rebuild; doubled headcount for data consistency work

A Series B team ditched Postgres because "Facebook uses Cassandra." Eighteen months later they had three document stores, hand-rolled transactions, and an analytics team that exported CSVs to answer questions SQL would have answered in 1998.

Multi-year rewrites of line-of-business UIs; mobile miss cost market share

Banks and insurers shipped rich dashboards in Flash and Silverlight because browsers were weak and IT loved installer-based control. Apple killed Flash on mobile. Microsoft killed Silverlight. Customers were left with plugins that security teams blocked and developers refused to maintain.

Traffic collapse and sale; rewrite consumed a year of product oxygen

Digg v4 was a top-to-bottom rewrite: new stack, new UX, social graph ambitions, publisher deals — shipped as a big bang that alienated power users and broke core voting workflows. Reddit ate lunch while Digg argued about Mongo and Cassandra.

Six-figure LLM API and SI costs; production cutover: zero

Exec decks promised autonomous COBOL-to-microservices migration "by Q3." Teams fed repos to copilots, generated plausible diffs, and discovered the model hallucinated APIs their mainframe never had. Pilots looked great in demos with 500-line samples; production needed tests nobody had.

Seven-figure platform licenses; shadow IT integration debt

Leadership bought a low-code platform to "10× citizen developers" and shrink engineering headcount. Ops built brittle workflows. Integrations hit API limits. Governance panicked. Two years later the company hired senior engineers to untangle the graph citizen devs had painted over legacy APIs.

Years of JVM/Scala investment alongside Ruby; Fail Whale era ops pain, not a total rewrite bill

Twitter became the cautionary tale for Ruby on Rails at hyperscale. The Fail Whale was real. So was the response: selective extraction of hot paths, JVM services, and later JVM-first culture — while plenty of product surface stayed dynamic and shipped.