Skip to content

All best practices

BP-001Grade B31 supporting cases (28 counted), 0 counter-examples

Gate every go-live on measured readiness, not on the calendar

The practice

Make every go-live a gated decision based on measured readiness (data accuracy, defects, training, a parallel or pilot run), and give the sponsor the authority to delay it.

Why it works

Go-live turns a project risk into an operating and customer risk. After cutover, a defect stops being rework; it becomes unpaid staff, empty shelves or locked-out customers. Dates are usually set by something outside the programme: a lease, a legacy end-of-support date, a competitor's launch or a public announcement. A gate that can only pass is therefore not a gate. The gate works when its criteria are fixed in advance, measured in the production configuration and at the largest volume the system will carry, and owned by someone with the authority to stop the clock. Master data deserves its own gate, because inaccurate data passes system tests and fails in operation. Flyvbjerg and Budzier's dataset shows why the downside is fat-tailed [R4], though it measures overrun risk, not the effect of gating.

KB synthesis: the mechanism is the knowledgebase's own reasoning, drawn from the member lessons. Factual support is the case claims under evidence.cases and the research under evidence.research (R-refs).

How to apply it

  • Write the go-live criteria into the programme charter before build starts: data accuracy threshold, open-defect ceiling by severity, training completion, parallel-run reconciliation.
  • Measure readiness in the configuration and at the volume of the largest site or market, not the first one.
  • Treat master data as a workstream with a named owner, an accuracy metric and its own gate.
  • Separate the operating go-live date from asset, lease or legacy-support dates; if they collide, escalate to the board rather than compress testing.
  • Give the people who see the readiness evidence a formal route to stop the clock, and record every gate decision and its evidence.
  • Never relax defect definitions or skip the end-to-end test to protect a date.

Warning signs

  • The go-live date was announced before the re-plan that justifies it was finished.
  • Defects are being reclassified to lower severity shortly before the gate.
  • The parallel run or pilot has been shortened or dropped.
  • Readiness is reported as vendor milestones or early small go-lives rather than operating measures.
  • Data accuracy is unmeasured, or measured on a sample.

When it applies

  • Big-bang cutovers of ERP, payroll, billing, core banking or order systems.
  • Simultaneous launches of sites, systems and supply chains.
  • Any change where an error reaches customers or staff on day one.

When it does not

  • Reversible, low-blast-radius releases, where continuous delivery with fast rollback is the safer control.
  • Hard external deadlines (for example a regulation) that force the date; there the control is a fallback process, not a delay.

Evidence grade: Grade B

At least 6 counted supporting cases for every 100 cases in the knowledgebase (6 × cases ÷ 100, rounded to the nearest whole number), whose primary regions span at least 4 regions.

Why this grade

Grade B by schema 4.10 route: >=14 counted cases across >=4 primary regions (the case bar is 6 per 100 cases in the knowledgebase, 235 cases). A needs a systematic, peer-reviewed study that tests this practice. None of the research cited in the KB does (R4 supports the mechanism only, not the practice). The practitioner surveys cited (R1) are self-reported and published by firms that sell transformation services; they are reported but do not count toward the grade. 3 supporting case(s) rest only on company-reported claims and are not counted toward the grade.

Where it applies

Transformation type
  • T05 ERP and core systems
  • T07 Supply chain
Topic (business problem)
  • Legacy ERP and fragmented systems block execution
  • Cross-functional processes are slow and broken
  • Cyber and technology risks threaten continuity
  • Technology spending is growing without business returns
  • Inventory is simultaneously excessive and unavailable
  • Service delivery is too slow, unreliable or expensive
Function
  • Supply Chain & Logistics
  • IT, ERP & Digital
  • Customer Experience & Support
  • Procurement & Strategic Sourcing
  • Risk, Compliance & Internal Audit
Starting situation
  • New initiative
  • In-flight initiative
  • Stalled or failing (turnaround)
Industry
  • Commercial Banking
  • Discrete Manufacturing
  • Electric Power Utilities
  • Food and Beverage Manufacturing
  • Grocery and Food Retail
  • Hospitals and Health Systems
  • Industrial Distribution and Wholesale
  • Logistics, Warehousing and 3PL
  • Other (cross-industry)
  • Specialty Retail and Omnichannel Commerce
Region
  • Africa
  • Central and Eastern Europe
  • East Asia
  • Global
  • Latin America
  • North America
  • Oceania
  • South Asia
  • Southeast Asia
  • Western Europe

Supporting and counter cases

Research

  • Among 1,471 IT projects, the average cost overrun was 27%, but about one in six had an average cost overrun of 200% and a schedule overrun of almost 70%.

    Dataset of 1,471 IT projects with a stated method (academic preprint). It measures cost and schedule overruns, not the effect of any practice.

    Why Your IT Project Might Be Riskier Than You Think · Flyvbjerg and Budzier, arXiv, 2013

  • BCG lists effective monitoring of progress toward defined outcomes among six factors associated with transformation success.

    BCG data on 70 companies plus a survey of 825 executives; success self-rated against targets; BCG sells transformation services.

    Flipping the Odds of Digital Transformation Success · Forth, Reichert, de Laubier, Chakraborty, Boston Consulting Group, 2020-10-29

Researched and fact-checked knowledgebase · Knowledgebase built 5 November 2026

Practices, their evidence grades and case summaries are public. Full case studies are available in AXIOM to subscribers and AXIOM Discover participants.