Measuring transformation and the capstone roadmap

Larchmont's three-year plan began in lesson 1 with a baseline and a set of targets. A year later, the board asks the obvious questions: is it working, is it worth the money, and what should change? This final lesson covers how to measure a transformation honestly, how to combine measures of delivery, experience, operations, AI governance and people into a coherent picture, how to avoid the traps that make measurement misleading, and how to turn findings into the next stage of the roadmap. It ends with a capstone that pulls the whole course together.

Why transformation measurement is hard

Transformations are long, involve many changes at once and play out in a business that keeps moving. Benefits often lag effort by months. Some benefits, such as reduced risk or better culture, are hard to quantify. Several initiatives may affect the same outcome, so attributing results is difficult. And people have incentives to report progress, which can tilt measures toward activity ('teams trained', 'tools deployed') rather than results.

Good measurement responds with a clear chain from strategy to measures, a mix of leading and lagging indicators, honest baselines, regular review and a willingness to act on bad news. It also accepts that some judgments will be qualitative and says so rather than inventing false precision.

From goals to measures

The goal-question-metric (GQM) approach, developed in software engineering in the 1980s, is a simple discipline: state the goal, ask the questions that would show whether it is being achieved, and only then choose metrics that answer those questions. It prevents the common mistake of measuring whatever data is easy to collect. OKRs, introduced in ITIL-120, follow a similar logic, with objectives stating intent and key results measuring progress.

Leading indicators show whether the conditions for success are in place and move early: platform adoption, automation coverage, skills coverage, share of AI systems assessed. Lagging indicators show results and move later: online booking share, customer satisfaction, cost per booking. A transformation dashboard needs both, because lagging indicators alone arrive too late to steer and leading indicators alone can show activity without results.

Flow and value measures

Many digital organizations measure flow across whole value streams, not just within teams. Flow time is the total time from when work is accepted to when it delivers value to the customer. Flow efficiency is the share of that time when the work is actively being worked on rather than waiting; ITIL-130 introduced it in value stream mapping. Flow load is the amount of work in progress, and flow velocity the amount completed per period. Little's law links these: average flow time equals average work in progress divided by average throughput, so reducing work in progress shortens flow time.

Value measures connect flow to the business. For Larchmont these include cost per booking, revenue from digital channels, customer retention and claims cost. Benefits realization, covered in PMP-150 and ITIL-130, tracks whether the benefits promised in the business case actually arrive, who owns each benefit and when it is expected.

Traps in measurement

Goodhart's law, often paraphrased as 'when a measure becomes a target, it ceases to be a good measure', warns that people will optimize the number rather than the outcome. A service desk measured on first-contact resolution may close tickets prematurely; a virtual agent measured on containment may make it hard to reach a person; a product team measured on deployment frequency may split changes artificially. Using several balanced measures, looking at the experience behind them and talking to the people doing the work reduces gaming.

Larchmont after year one

One year in, LTS compiled results against the baseline and three-year targets. Some measures moved fast, some slowly, and one went the wrong way. The progress index below shows how much of the gap from baseline to target has been closed for each measure.

# Year-one progress toward three-year targets (invented results)
measures = [
    # name, baseline, year 1, target
    ("online booking share %",          38,  51,  75),
    ("minutes to book online",           14,   7,   4),
    ("lead time idea to prod (days)",    74,  31,  10),
    ("self-service resolution %",         9,  24,  40),
    ("incidents found by monitoring %",  41,  63,  80),
    ("digital CSAT (1-5)",              3.1, 3.6, 4.0),
    ("staff positive about tools %",     52,  58,  75),
    ("change failure rate %",            17,  19,  10),
]
print(f"{'measure':<34}{'base':>6}{'yr1':>6}{'target':>8}{'gap closed':>12}")
for name, b, y, t in measures:
    closed = (y - b) / (t - b) * 100
    note = "  <- moving the wrong way" if closed < 0 else ""
    print(f"{name:<34}{b:>6}{y:>6}{t:>8}{closed:>11.0f} %{note}")
avg = sum((y - b) / (t - b) for _, b, y, t in measures) / len(measures) * 100
print(f"average gap closed after one of three years: {avg:.0f} %")

measure                             base   yr1  target  gap closed
online booking share %                38    51      75         35 %
minutes to book online                14     7       4         70 %
lead time idea to prod (days)         74    31      10         67 %
self-service resolution %              9    24      40         48 %
incidents found by monitoring %       41    63      80         56 %
digital CSAT (1-5)                   3.1   3.6     4.0         56 %
staff positive about tools %          52    58      75         26 %
change failure rate %                 17    19      10        -29 %  <- moving the wrong way
average gap closed after one of three years: 41 %

The picture is encouraging but uneven. Booking time, lead time and incident detection have closed half or more of their gaps in a third of the time. Self-service and online share are close to schedule. Staff experience lags, consistent with the culture survey in lesson 11 and with the extra load of change. The change failure rate has worsened slightly as deployment frequency rose, which matches the early DORA data in lesson 4: speed has outpaced investment in testing and progressive release.

Reviewing and adapting the roadmap

Measurement is only useful if it changes decisions. Larchmont's steering group held a year-one review structured around three questions: what is working and should be scaled, what is not working and why, and what should change in the roadmap. It used the guiding principles from lesson 10 as a check and invited product teams, the service desk and two customers to present their views.

The review led to four decisions. First, wave two would include a stability push: every product team would spend a quarter improving automated tests and progressive release, with an explicit goal of bringing change failure rate below 13 % before increasing deployment frequency further. Second, staff experience would get a named executive owner and a funded program fixing the ten most complained-about internal tools. Third, the virtual agent would expand from tracking to proof-of-delivery and contact changes only after its true resolution rate passed 60 %. Fourth, the résumé screening feature would be permanently switched off, since the vendor could not explain its scoring.

Measuring a digital transformation increasingly includes its environmental footprint, as data centers, cloud services and AI models consume energy. In the United States, the SEC stopped defending its climate disclosure rule in 2025, but California's SB 253 requires large companies doing business in California to report greenhouse gas emissions, with first reports due in 2026, and many customers ask suppliers for emissions data. Larchmont added cloud energy use and emissions estimates from its providers to the transformation dashboard so that AI and platform growth would be visible in its sustainability reporting.

Capstone: Larchmont's years two and three

The capstone brings together every lesson. Imagine you are advising Larchmont's steering group on the plan for years two and three. A strong plan would include the following elements, each traceable to an earlier lesson.

A plan like this is not a fixed blueprint. It is the current best hypothesis, to be tested and revised at each review. That is perhaps the most important idea of the whole course: digital products, automation, AI and transformation all reward organizations that learn quickly from evidence, and punish those that commit to a plan and stop looking.

Preparing for advanced ITIL assessments

Advanced ITIL assessments on products, experience, strategy, AI governance and transformation tend to present scenarios and ask which action best applies the framework. Useful habits are to identify the outcome the scenario is really about, look for answers that focus on value and experience rather than activity, prefer iterative and evidence-based approaches over big-bang ones, check that the answer considers people, technology, suppliers and processes, and for AI questions look for accountability, transparency, human oversight and lifecycle risk management. Confirm the format, edition and rules of your specific exam with the exam provider before you book.