Skip to content

03 · Digital Transformation Programs

A digital transformation program is not a large IT project wearing a bigger title — it's a program whose primary deliverable is organisational change, where the technology is often the easier half. Programs in this category fail overwhelmingly on adoption, not on delivery: the new system ships on time, and eighteen months later half the organisation is still running the old process in parallel because nobody made the old way actually stop working.

Why transformation programs are structurally different

Typical IT project Digital transformation program
Primary deliverable Working software Changed ways of working, enabled by software
Success measure System is live Old process is retired and new one is the default
Main risk Technical delivery Adoption, change fatigue, competing incentives
Timeline Fixed, bounded Often multi-year, benefit-driven (Level 3, module 01)
Sponsor's job Approve budget Actively model and mandate the new behaviour

The transformation benefits realisation curve

Technology go-live and benefit realisation are not the same event, and the gap between them is where most transformation programs lose executive patience.

Phase What happens Typical duration post-go-live
Go-live System is technically available Day 0
Adoption trough Productivity often dips below the old-process baseline as people relearn 4–12 weeks
Parallel running Some teams use old + new simultaneously "just in case" 1–3 months if not actively shut down
Old process retirement Old system/process access is formally removed Must be a deliberate decision, not a drift
Benefit realisation Measured gains appear 3–12 months post-go-live, depending on benefit type

The adoption trough is not a sign the program failed — it's an expected, temporary productivity dip while people relearn a new way of working, and a sponsor who panics and reverts at week 3 of the trough guarantees the program never reaches realisation. The program's job is forecasting the trough explicitly, in the plan, so it isn't mistaken for failure when it arrives on schedule.

Worked example: adoption curve with a forced cutover decision

A finance transformation program replaces a manual approval process with an automated workflow. Three months post-go-live, 60% of approvals still go through the old manual path in parallel.

Option Action Trade-off
Continue parallel running indefinitely Let usage migrate "naturally" Never happens — data from comparable programs shows parallel systems plateau, they don't self-resolve; the old path stays a permanent crutch
Set a hard cutover date, communicated 6 weeks ahead Old system access revoked on a fixed date Short-term pain spike, but forces the adoption the parallel period was supposed to produce
Incentivise the new path Recognition/targets tied to new-process usage Slower, but lower-friction; often combined with a cutover date as a backstop

The programs that actually retire the old process combine both of the last two: a communicated deadline (removes the option to defer indefinitely) plus incentives in the weeks before it (reduces the deadline's felt harshness). A cutover date alone, with no support, produces a spike in help-desk tickets and a credibility hit; incentives alone, with no deadline, produce the indefinite-parallel outcome.

Change management workstream (parallel to delivery)

Workstream Delivery-side owner Change-side owner Key artefact
Technical build Program manager / engineering lead Architecture, test plan
Communication Change lead Comms calendar, key messages by audience
Training Change lead Role-based training plan, completion tracking
Resistance management Change lead + sponsor Stakeholder resistance log (Level 3, module 07 patterns, applied at scale)
Incentive/mandate design Sponsor Change lead Cutover plan, incentive structure

Treating change management as a separate, adequately-resourced workstream — not a communications afterthought added in the final sprint — is the single structural decision that most correlates with transformation programs that actually realise their benefits map.

Measuring adoption, not just usage

Metric What it actually tells you Common mistake
Login count System is being opened Doesn't mean the process changed — people log in and still do the old workaround
% of approvals through new workflow True adoption of the new process The metric that actually matters
Help-desk ticket volume trend Whether friction is decreasing over time Rising tickets in week 1–2 is normal; rising in month 3 is a real signal
Manager survey: "which process do you actually use day to day" Self-reported reality, catches shadow processes Only as honest as psychological safety allows — anonymise it

How It Actually Works

The benefits-realization curve's characteristic dip before the eventual gain is the J-curve, and it has a precise mechanical cause, not just a mood: during cutover, the organization is paying the full cost of the change (training time, temporary productivity loss, parallel-running old and new systems) while capturing none of the benefit yet, because adoption — and therefore benefit — is a function of cumulative usage, which starts at zero regardless of how good the new system is. Modeling adoption as an S-curve diffusion process (again Rogers' curve from Level 2 Module 7) rather than a step function is what allows a forced-cutover decision to be made rigorously: the total value destroyed by the J-curve dip is roughly the integral of (cost − benefit) over the adoption ramp, and a forced cutover is justified specifically when running two systems in parallel for longer would cost more than accepting a steeper, shorter dip — a calculation, not a leadership preference.

Exercise

A transformation program replacing a legacy CRM goes live on schedule. Three months later, login counts to the new system are high, but sales reps are reportedly still exporting data to spreadsheets to do their actual forecasting work — a shadow process the login metric doesn't catch.

  1. Identify which phase of the benefits realisation curve this program is actually in, and explain why login count alone would have masked the real problem.
  2. Design one better adoption metric (using the "measuring adoption, not usage" table's logic) that would have caught the shadow spreadsheet process earlier.
  3. Propose a cutover plan (deadline + incentive, per the worked example) to retire the shadow process, including what you'd communicate to sales reps and how far ahead of the deadline.