A file drop is an integration contract
The operational and semantic rules that turn a delivered file into accepted business input.
4 min read
Explainer
A knowledge graph that can explain a proposed migration path
Representing migration candidates, prerequisites, evidence, and unknowns without turning graph paths into guarantees.
3 min read
Acceptance criteria for a system that does not repeat itself
A fixed seed makes a best effort, not a guarantee. Acceptance shifts from exact output to declared task properties, a predeclared sampling rule, and uncertainty measured at the unit actually sampled.
4 min read
What the accounting boundary does to your architecture
Capitalisation rewards crossing from exploration into development, so the boundary quietly rewards early commitment to a design. An incentive on technical decisions nobody describes as technical.
5 min read
Adverse action notices when the model is a language model
Regulation B requires the specific principal reasons the decision actually used. A fluent explanation of a decision the model did not make is the failure, and the CFPB circulars people cite were withdrawn in 2025.
4 min read
AI business-rule extraction needs a chain of evidence
Connecting generated rule descriptions to code, configuration, observed work, and confirmed business meaning.
3 min read
An archive and a backup preserve different things
Backups support restoration of operation. Archives support continuing access to records and their meaning after the original application is removed.
4 min read
API facades and the behavior behind the interface
A facade exposes a consumer contract over retained behavior. Translation and routing do not remove the operational obligations of the underlying application.
4 min read
Application outsourcing, managed services, and business-process outsourcing
Distinguishing outsourced application work, operational services, and responsibility for an entire business process.
3 min read
Batch checkpoints, event streams, and duplicate business actions
Connecting restart positions and messaging acknowledgments to the business effects that must not be repeated.
3 min read
The benefit decays before the transfer completes
Month-one and month-nine measurements are different quantities. A gate reading the early number transfers a system whose case has already expired, and time released is capacity rather than cash.
5 min read
Boring technology, and the debt you write down
Familiar means familiar to the receiving team, not old. And the ledger records what the code cannot do, including the debt nobody chose.
6 min read
Build-operate-transfer already means two other things
BOT names an infrastructure concession, an offshore entity handover, and a prototype transfer. The phrase does not say which, and the acronym family does not say when responsibility moves.
4 min read
Business capabilities and the applications that implement them
Capability mapping connects required business outcomes to processes, people, information, and applications. Shared labels do not establish interchangeable implementations.
3 min read
Change data capture from snapshot to ongoing replication
Change data capture combines an initial state with subsequent changes. Offsets, source-history retention, duplicate handling, and excluded objects determine continuity.
4 min read
Characterization tests: recording the behavior that exists
Characterization tests preserve selected observed behavior. Their baseline must distinguish required behavior from existing defects and comparison artifacts.
4 min read
Cheap code moved the bottleneck, it did not remove it
When generation cost falls toward zero, the constraint relocates to verification, approval, and integration — none of which got cheaper. A queue forms where nobody is measuring.
5 min read
Chunking is a retrieval decision, not a preprocessing detail
Where the boundary falls decides whether a rule and its exception arrive together. The pipeline that cannot delete a withdrawn document is the failure that survives handover.
5 min read
Citrix automation under session and credential constraints
Combining virtual-desktop automation prerequisites with identity, secret lifecycle, and safe recovery.
3 min read
Claim-level evidence for modernization decisions
Making consequential assertions traceable, scoped, dated, and open to correction.
3 min read
COM components across the 32-bit and 64-bit boundary
How COM process boundaries determine architecture compatibility and what activation tests must prove.
4 min read
Comprehension debt and the transfer that cannot happen
Understanding is measured on the receiving side, against named tasks. A team that passes acceptance and cannot explain why a rule change moved an outcome has not received the system.
5 min read
Container portability ends at the application’s external dependencies
Separating image reproducibility from host, storage, identity, and replacement dependencies.
3 min read
The credentials that outlive the engagement
Vendor accounts, service principals, API keys and cloud tenancy persist after the final invoice. Revocation is not one action, and absence from an inventory is not absence.
5 min read
Cutover reversibility under cheap generation
Cheap code lowers the cost of building the replacement. It does not lower the cost of being wrong about the legacy behaviour, and sent messages do not un-send.
4 min read
Database-first migration: changing storage while retaining the client
Database-first migration replaces the data layer while retaining a client. The changed access path must preserve required queries, updates, and business meaning.
4 min read
DBF tables, memo files, and the meaning of a record
Why DBF migration must account for format variants, memo references, deleted records, and application interpretation.
4 min read
Decommissioning a system while keeping its records usable
Retiring execution while preserving record integrity, interpretation, access, and historical answers.
3 min read
Dependency mapping beyond the source code
Source analysis, runtime observation, configuration, and operator knowledge reveal different dependencies. Each relationship needs a direction, purpose, and evidence boundary.
4 min read
Developers were wrong about their own speed, and the measurement problem generalises
Forecast 24 percent faster, felt 20 percent faster, measured 19 percent slower. The finding is about self-reported productivity as an evidence class, and it disqualifies most of what pilots collect.
5 min read
Documentation-only handoff, and the three transitions that replace it
Shadowing, reverse shadowing, and pairing transfer different things in a required order. Delivering documents and confirming receipt are two records, and neither is evidence of use.
5 min read
Endowment effect meets parity-plus acceptance
The behavioural evidence is contested, which is the argument for checking the workflow first. A replacement judged on new capability gets judged on missing capability instead.
5 min read
Staff augmentation, managed services, fixed-bid, outcome-based: where the risk sits
Six engagement models fix four independent variables — who directs the work, how price is set, who owns continuing operation, and what transfers at the end. A label pins one and leaves three open.
5 min read
Escrow does not survive model deprecation
Source escrow assumes the artifact is the source. Where behaviour depends on a hosted model the provider retires, the deposited package no longer reproduces the system.
5 min read
The EU AI Act's risk tiers, and where a prototype sits
Obligations attach to role, risk class, and the act of placing on the market. Real-world testing is carved out of the pre-market exclusion, and the 2026 amendments moved the high-risk dates.
5 min read
The evaluation set is the asset that survives the kill
Killing a prototype destroys the system and preserves the encoded definition of correct behaviour. Salvage is a forward comparison, not a rescue justified by what was already spent.
4 min read
Excel refresh timing and the meaning of a financial total
How refresh completion, calculation order, and business cutoffs determine what a workbook total represents.
3 min read
Extend, kill, or transfer
The decision the operate phase exists to produce. Three named outcomes, three others that get hidden inside them, and a default that applies when the evidence is missing.
5 min read
Extracting a VB6 function across a COM boundary
Choosing a coherent VB6 extraction boundary while preserving COM consumers, transactions, and deployment.
3 min read
Falsifiable success criteria
A success criterion is falsifiable when a stated observation would end the project. Most criteria have two states and need three, because inconclusive is the result that actually occurs.
5 min read
Fixed-price modernization when discovery is incomplete
Defining priced obligations and acceptance evidence when legacy behavior and dependencies are still being discovered.
3 min read
From FoxPro DBF files to a relational database: preserving meaning
Combining FoxPro file interpretation, target schema design, client connectivity, and business reconciliation.
4 min read
Golden, adversarial, and regression sets
Three purposes, not three folders. One case can carry all three tags and still be one case family, and three passing tags are not three independent successes.
5 min read
Guardrails are an architecture, not a filter
Controls belong at the boundary that can enforce them. Authorisation happens in the downstream system, not in the model, and autonomy level sets how much each layer must carry.
5 min read
How a strangler migration replaces a system incrementally
A strangler migration routes selected responsibilities from an existing implementation to replacements. Completion requires removing the old dependencies, not merely adding a facade.
4 min read
Idempotence: what a repeated operation is allowed to do
Idempotence concerns the intended effect of repeated requests. It is distinct from identical responses, broker acceptance, and exactly-once business execution.
4 min read
Full-stack IP ownership, work-for-hire, and licence-back
Owning the copyright is not holding the file, and holding the file is not being able to run the system. Calling a contract work made for hire does not make it one.
5 min read
Kill criteria are a political instrument
A precommitted criterion works only where the sponsor survives triggering it. Escalation of commitment is the mechanism it has to defeat, and it was demonstrated in the people who made the original choice.
5 min read
Write the kill criteria before the demo
Recorded conditions for stopping, agreed before anyone sees the evidence they govern. Six trigger families, and the rule that a stop decision and a finding about value are separate statements.
5 min read
Liability after the last invoice
Which obligations survive handover, which insurance responds to a model's output, and why detection after transfer establishes nothing about when the defect arose.
5 min read
LLM-as-judge, and what calibrating it against humans actually requires
A judge is a measurement component with order sensitivity and no established authority. Swap the presentation order, and a judge that picks the first answer both times has produced two conflicting verdicts.
4 min read
Lotus Notes migration must preserve execution authority
Mapping Notes workflow state together with agent identity, authority, triggers, and in-flight work.
3 min read
Low-code applications still have a deployment lifecycle
Low-code packaging, dependencies, query limits, and restoration remain software lifecycle concerns.
3 min read
Masked data, real conclusions: the fidelity gap in your evidence
Every de-identification transform removes signal the model was going to use. What a pilot on treated data can and cannot establish, stated as a rule rather than a caveat.
5 min read
Microsoft Access front ends and back ends
Separating Access application objects from shared data without mistaking a storage change for a rewrite.
4 min read
Migrating workflows that already have work in progress
Mapping active workflow state, approval semantics, and pending external actions during a platform transition.
3 min read
Migration sequencing across technical and business dependencies
Turning portfolio priorities into feasible waves with explicit intermediate states and recovery boundaries.
3 min read
Model risk management after SR 11-7, and where generative AI sits
SR 26-2 replaced SR 11-7 in April 2026 and its footnote 3 puts generative and agentic AI outside scope. Being out of scope routes the work elsewhere; it does not remove it.
4 min read
Moving Access tables to SQL Server while keeping the forms
Tracing retained Access behavior through SQL Server data access, conflict detection, and controlled write ownership.
4 min read
MVP against a working baseline
A restricted replacement justified as an MVP while failing the operational needs of the users required to adopt it. The diagnosis is the removed work, not the feature count.
4 min read
NIST AI RMF and ISO/IEC 42001: a framework and a management system
One is a voluntary risk framework applied to a system, the other a certifiable management system applied to an organisation. Cited as alternatives, they answer different questions and neither certifies an application.
4 min read
No capacity, no skills, no budget, no interest
Four handover failures with four diagnostics and four remedies. A stalled transfer is an outcome, and it does not identify its own cause.
5 min read
Nobody plans the end of the chain
Transferred systems get retired by a team that did not build them, using documentation written for a handover years earlier. It is the least-instrumented stage and the one where the original evidence is worth most.
5 min read
ODBC: the application, driver manager, driver, and data source
How ODBC connection layers, process architecture, and execution identity affect compatibility.
3 min read
Parallel prototyping changes selection, not construction
When three attempts cost what one used to, the gain is in comparison. The scarce resource becomes the ability to choose between working options, which is an evaluation problem.
5 min read
Parallel replacement and comparable results
Parallel replacement compares implementations over matched inputs and state before changing operational authority. Shadow execution must control external effects.
4 min read
Comparability, not documentation: how pharmaceutical technology transfer proves a handover
A regulated transfer completes when the receiving site independently reproduces the specified result against criteria agreed in advance. Assisted and unassisted execution answer different questions.
4 min read
Phoenix pay, HealthCare.gov, and NPfIT read through a transfer lens
Three heavily documented programmes, re-examined for one question: what was handed over, to whom, and against what evidence. Plus the limits of reasoning from cases at all.
5 min read
Your pilot cannot detect the effect you are looking for
Cohorts are sized for access, not for power. Where the minimum detectable effect exceeds the claimed benefit, the pilot cannot produce the evidence its business case assumes.
5 min read
Portfolio prioritization: risk, value, and treatment choice
Using portfolio evidence to choose investigation and treatment priorities without letting a score choose the architecture.
3 min read
PowerBuilder replacement and the transaction hidden in the client
Preserving PowerBuilder save semantics across DataWindows, database transactions, client buffers, and replacement APIs.
3 min read
Process elimination as a modernization outcome
Process elimination ends work whose outcome is no longer required. Application shutdown and transferring work elsewhere are different outcomes.
4 min read
Prompt injection is a tool-permission problem
The injection is not the vulnerability; the tool the injected instruction can reach is. Retrieval and fine-tuning do not mitigate it, and prototypes grant permissions generously.
6 min read
Prototype, pilot, proof of concept: seven labels that do not line up
The seven instruments are drawn from four independent dimensions of the same work, so they overlap instead of partitioning. Each settles one thing and leaves an untested remainder.
6 min read
When the prototype quietly became a regulated model
The transition is caused by use, not by declaration. A real decision, a real customer, a real record — and the obligations attach retroactively to work already done.
4 min read
The prototype-to-production multiple
Two different ratios travel under one name, and a higher multiple can describe a cheaper project. No benchmark value is established, and the denominator is where the argument lives.
4 min read
Publishing your failure rate is a procurement instrument
A reported failure rate converts an admission into a screening filter. It only works if the denominator is disclosed, because a zero-failure portfolio is a claim about reporting.
5 min read
Recovery time, recovery point, and a demonstrated restore
Recovery time and recovery point are requirements. A restore exercise measures the duration and recovered state actually achieved under defined conditions.
4 min read
Rehost, replatform, refactor, or rewrite: what changes in each?
Rehosting changes execution location, replatforming changes supporting services, refactoring changes internal structure, and rewriting creates another implementation.
4 min read
Replacing a mainframe file extract with a service interface
Preserving a host extract’s encoding, cutoff, completeness, and consumer semantics when introducing an API.
3 min read
Replacing Crystal Reports means validating queries and rendering
Preserving a report’s data selection, calculations, dependencies, and delivered output across replacement.
3 min read
Reproducible builds and recoverable source
A reproducible build produces identical specified artifacts from declared inputs. Recoverable source and artifact provenance answer different questions.
4 min read
Retain and govern: an explicit decision to keep a system
Governed retention preserves a useful system with accountable operation, evidence of recovery, and explicit triggers for reconsideration.
4 min read
Retrieval, fine-tuning, prompting: which problem each one solves
Three intervention points, not three exclusive strategies — the original RAG paper already combined two of them. Diagnose the failure before choosing the technique.
5 min read
Rollback after the new database has accepted writes
How target-side transactions change reversal from a routing operation into a data-recovery decision.
3 min read
Run cost is the number the prototype hides
Build cost is paid once; inference, retrieval, review and monitoring scale with use. A cheaper model bill can raise cost per accepted outcome, and the run-cost owner is absent from the approval.
4 min read
SaaS portability beyond the data export
The difference between retrieving SaaS records and carrying a business process into another environment.
3 min read
SaaS replacement or process outsourcing: where does responsibility sit?
Comparing a hosted application with provider-executed business work through responsibilities, exceptions, and exit.
3 min read
Shadow mode
Running a candidate against live inputs with its outputs withheld. The word does not establish the boundary — shared caches, writable destinations and quotas escape it unless someone checked.
5 min read
Shortest path to value
Waters' name for a legacy-replacement roadmap that delivers useful capability while limiting disruption. It answers a different question from an MVP, and the two coexist in one programme.
4 min read
Snagging lists, defects liability, and readiness ladders
Construction separates practical completion from final acceptance and holds the builder liable across the gap. Four disciplines stage readiness; software handover has no equivalent structure.
5 min read
Sometimes the correct answer is never to transfer
The structure assumes handover is the goal. Where the receiving organisation has no path to the capability, forcing a transfer produces an orphan with an owner's name on it.
5 min read
Stabilization: reducing immediate operational risk
Stabilization addresses a named weakness in current operation. Isolation, recovery, and dependency repair reduce different risks and leave different constraints.
4 min read
Synthetic, masked, sampled, production: what each data strategy costs you in evidence
Not four rungs of one ladder. Three are construction choices that overlap and one is an operating context, and synthetic does not mean anonymous.
5 min read
The application has moved, but the data center cannot close
How shared infrastructure and recovery dependencies can keep a source facility in the operating path.
3 min read
The evaluation harness is the deliverable
The system can be rebuilt; the encoded definition of working cannot. Ship the harness first, and test it by making a deliberately wrong answer fail.
4 min read
The full cost of retaining or replacing an application
Comparing whole-life costs against a consistent baseline, including coexistence and the conditions for realizing savings.
3 min read
The pilot as a sales tactic
A cheap pilot is a qualification instrument for the vendor. The diagnostic is what evidence and permitted artifacts remain if the buyer purchases nothing further.
5 min read
The receiving team is the deliverable
Invert the frame: the software is the byproduct and the capability is the product. Every practice that eases transfer is a training programme with an artifact attached.
5 min read
The run-cost handoff nobody scheduled
A transferred system arrives with a bill and no budget line, because the fiscal year that funded the build did not fund the operation. Six things move on six different dates.
5 min read
The separate support lifecycles inside one application
A deployment combines product, runtime, development-tool, operating-system, and component support commitments. A single supported label cannot describe them all.
4 min read
Third-party risk management is the gate your prototype actually has to pass
A business arrangement can exist with no contract and no payment, so the free pilot is already inside scope. And some obligations survive the relationship ending.
5 min read
Throwaway or evolutionary: decide before the first commit
Whether prototype code survives is a decision per component, made in advance, and separate from whether the learning survives. Neither always-rewrite nor demo-implies-retain holds.
5 min read
Transition assistance clauses: the exit you negotiate at the start
Termination for convenience is a settlement mechanism, not a free exit. And staff are participants with rights, not artifacts that transfer with the repository.
4 min read
TSB’s 2018 migration: data transfer and service acceptance
What the documented TSB case establishes about the boundary between migrated data and working banking services.
4 min read
Virtualizing equipment-control software and preserving physical behavior
Distinguishing a recoverable virtual machine from validated behavior at the equipment boundary.
3 min read
Walking skeleton, tracer bullet, strangler fig
Three patterns that build a thin end-to-end path before breadth. Two of them overlap and the third is a different kind of thing, because it has to keep an incumbent running alongside.
5 min read
What an unattended RPA bot depends on
The identities, sessions, selectors, and recovery rules behind unattended desktop automation.
3 min read
What automated source conversion actually changes
Separating syntax transformation, runtime support, behavioral equivalence, and future maintenance.
3 min read
What belongs in an application inventory?
An application inventory records deployed systems, their business roles, dependencies, and the evidence behind each attribute. Unknown values remain distinct from observed facts.
4 min read
What makes a system legacy?
Legacy status describes a constraint on continued operation and change. Age, support status, and treatment choice are separate assessments.
4 min read
When does a prototype become a capitalisable asset?
IFRS and US GAAP draw the line differently, neither draws it at the word prototype, and the US test is changing. Costs expensed before the criteria are met are not reinstated.
5 min read
When an Excel workbook is a production application
Recognizing the code, dependencies, and operating responsibilities inside a business workbook.
3 min read
When is an application actually decommissioned?
Decommissioning removes active execution after required consumers and historical access have been resolved. A quiet monitoring window is bounded evidence.
4 min read
Where Oracle Forms business logic ends and an API begins
Recovering Forms event logic and execution context before exposing database operations through an API.
3 min read
Which assumption would reverse the modernization decision?
Using sensitivity analysis to expose fragile assumptions and choose the next useful investigation.
3 min read
Who checks the expected result in an AI-generated test?
Evaluating generated assertions through independent examples, approved baselines, and meaningful fault detection.
3 min read
Who owns the embeddings?
Indexes, embeddings, logs and fine-tuning corpora are created during the engagement and left out of the contract that governs it. Possession, ownership, licensed use and portability are four separate answers.
5 min read
Why matching row counts cannot validate a data migration
Matching row counts establish quantity, not identity, meaning, or readiness for new writes. Migration validation checks an explicit transformation contract and accounts for what remains untested.
5 min read
Writing for retrieval changes which claims survive
When the primary reader chunks and recombines the text, prose style is discarded and structure is not. A constraint on argument construction rather than on formatting.
5 min read