Certified Technical Consultant cheat sheet

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.

Scope baseline = signed statement of work (SOW) plus approved change requests; any deliverable not traceable to the baseline is out of scope until a change order is executed.

Fixed-price contracts transfer cost risk to the consultant; time-and-materials (T&M) transfers effort risk to the client; a not-to-exceed (NTE) cap converts T&M into a ceiling with a renegotiation trigger.

RACI assigns exactly one Accountable per activity; Responsible does the work, Consulted is two-way before the fact, Informed is one-way after the fact.

Earned value: CV = EV - AC; SV = EV - PV; CPI = EV/AC; SPI = EV/PV. CPI or SPI below 1.0 signals cost or schedule overrun.

Estimate at completion: EAC = BAC/CPI when current cost performance persists; EAC = AC + ETC when a bottom-up re-estimate is used.

Change control sequence: log request, assess impact on scope/schedule/cost/risk, obtain written approval, rebaseline, then execute.

A risk register entry carries probability, impact, exposure (P x I), owner, response strategy (avoid, mitigate, transfer, accept), and trigger.

Engagement close requires deliverable acceptance sign-off, lessons-learned capture, resource release, and archival of configuration and decision records.

Requirement types: business (why), stakeholder (who benefits), functional (what the system does), non-functional (quality attributes such as latency, availability, throughput), and transition (data migration, cutover).

A requirement is testable only if it states a measurable condition and outcome; 'fast' and 'user-friendly' are defects, not requirements.

MoSCoW prioritization ranks Must, Should, Could, Won't; Must items must together form a viable, deliverable solution.

Use open questions to elicit, closed questions to confirm; a five-whys chain moves from a stated want to the underlying business need.

Traceability matrix links requirement ID to design element, build component, and test case; an orphan requirement has no downstream test and no upstream business justification.

Fit-gap analysis classifies each requirement as standard fit, configuration fit, customization, or process change; customization count is the primary driver of total cost of ownership.

Interview, workshop, observation, document analysis, and prototype review are the five core elicitation techniques; prototypes are the strongest tool for resolving ambiguous UI requirements.

Design sign-off freezes the functional specification; post-sign-off changes re-enter the change control process from domain 1.

Integration styles: point-to-point (simple, N-squared connections), hub-and-spoke/ESB (central mediation), and event-driven publish/subscribe (loose coupling, eventual consistency).

Synchronous request/response blocks the caller and fails fast; asynchronous messaging decouples producer and consumer and requires idempotent consumers plus retry with dead-letter queues.

Idempotency means repeated delivery of the same message yields the same end state; achieve it with a natural key or deduplication store.

REST uses HTTP verbs on resources and is stateless; SOAP carries a WSDL contract and WS-Security; choose SOAP when formal contract and transport-level security are mandated.

CAP theorem: under a network partition a distributed system must choose consistency or availability; the choice is a design decision, not an implementation detail.

Authentication proves identity (SAML, OIDC); authorization grants permission (OAuth 2.0 scopes, RBAC). Never conflate the two in a design review.

Single sign-on via SAML assertion or OIDC ID token; token lifetime and refresh strategy determine session behavior and revocation latency.

Non-functional architecture targets must be quantified as SLOs (e.g., p95 latency, RPO/RTO for disaster recovery) and validated by load or failover test.

Configuration changes delivered through the platform's supported settings; customization requires code or schema change. Prefer configuration because it survives upgrades.

Sandbox-to-production promotion path: development, integration/test, staging/UAT, production; each promotion is a versioned, repeatable deployment artifact.

A data migration run has five stages: extract, profile, cleanse/transform, load, reconcile; reconciliation compares source and target record counts and key field totals.

Custom objects and fields should carry a naming convention and a description; undocumented custom metadata is the leading cause of upgrade breakage.

Bulk data loads require batching and API rate-limit awareness; parallel loads against the same records cause lock contention and partial failures.

Unit tests cover logic in isolation; integration tests cover the configured end-to-end path; a customization without an automated test is a regression risk.

Rollback plan must exist before deployment: either a reverse migration script or a restore point verified in a non-production environment.

Test levels: unit, integration, system, UAT, regression, and smoke; smoke runs first on a new build, regression runs after any change to guard existing behavior.

UAT exit criteria are defined in advance: all critical and high defects closed, agreed defect tolerance met, and business sign-off recorded.

Defect severity measures impact on the system; priority measures urgency of fix. They are independent fields and can disagree.

Entry criteria for a test phase are a stable build, test data ready, and environment available; exit criteria are pass rate, defect thresholds, and sign-off.

Deployment strategies: big-bang (single cutover, highest risk), phased (by module or geography), parallel run (both systems live, costly but safest), and pilot.

A go/no-go decision uses a readiness checklist: test completion, data reconciliation, training completion, rollback plan, and support rota confirmed.

Post-deployment verification (hypercare) monitors error rates, integration queue depth, and user-reported issues for a defined window before handover to steady-state support.

Status reporting cadence is set in the SOW; a RAG status must be backed by a stated criterion (e.g., red = milestone slip over 5 working days or unresolved critical issue).

Escalation path is defined at kickoff with named contacts on both sides and time thresholds per level; escalation is a process, not a punishment.

Training design follows audience analysis, learning objectives, materials, delivery, and assessment; role-based curricula outperform one-size-fits-all sessions.

Handover package: as-built documentation, configuration workbook, runbook, support model, credentials transfer, and warranty terms.

A decision log records date, decision, rationale, alternatives rejected, and owner; it is the defense against scope and memory disputes.

Meeting minutes are actionable only when they list owner and due date per action item and are circulated within one business day.

Knowledge transfer is verified by the client performing the task unaided, not by attendance at a training session.

Personal data protection principles: lawful basis, purpose limitation, data minimization, accuracy, storage limitation, integrity/confidentiality, and accountability.

A data processing agreement is required when a processor handles personal data on behalf of a controller; cross-border transfer needs a documented legal mechanism.

Confidentiality survives engagement end; client data used in demos, sandboxes, or case studies requires explicit written permission and de-identification.

Conflict of interest must be disclosed in writing as soon as identified; the client decides whether the engagement continues with mitigation.

Accepting gifts or hospitality beyond a nominal threshold creates a perceived conflict; disclose and decline per the client's policy where one exists.

License compliance is a consultant obligation: verify entitlement before enabling features, and never use production data in a non-production environment without masking.

Audit trail requirements (who changed what, when) are often regulatory; disabling logging to improve performance can be a compliance breach.

Whistleblowing and reporting obligations override client confidentiality when law or imminent harm is involved; escalate through the firm's compliance channel first.