Executives are under pressure to ship faster without breaking teams. Too often the response is a ritualized rollout of agile ceremonies, new boards, and rebranded roles. The theater is impressive, the outcomes are not. The research base offers a more candid reading: treat agile as an operating hypothesis to test, not a system to install.
What the evidence actually says
Large enterprises keep asking for the scaling recipe. The literature does not have one. A systematic review of 52 publications covering 42 industrial transformations found that almost 90 percent of the evidence came from practitioner reports rather than academic studies, a weak foundation for universal prescriptions (Dikert, Paasivaara and Lassenius, 2016). The same review catalogued 35 distinct challenges and 29 success factors. The items that surface most often are managerial: visible support from leadership, aligning the organization around a shared direction, investing in training and coaching, and customizing the method to fit context rather than copying it wholesale (Dikert, Paasivaara and Lassenius, 2016). That profile looks nothing like a checklist of meetings. It looks like change management.
If you are tempted to read the catalogue as a framework in disguise, resist. The heterogeneity of cases and the preponderance of experience reports mean we should be cautious about generalizing. The implication is not that agile is hollow, but that copying ceremonies without addressing leadership posture, alignment, and local design is unlikely to move the needle. The review’s strongest signal is cultural and managerial, not procedural.
The 2024 signal: context and leadership beat ceremony
The largest contemporary dataset in this space, the DORA State of DevOps 2024 report, surveyed roughly 39,000 respondents and reports clear associations that should shape how leaders act. Organizational priority instability correlates with lower productivity and higher burnout, and the negative effect persists even in environments with good leadership and good documentation (DORA, 2024). In other words, you can run tidy standups and keep wikis immaculate, yet if the portfolio churns every week the system will still grind teams down. The authors are explicit that their survey is cross-sectional, so these are correlations, not proofs of causation (DORA, 2024). The pattern is still hard to ignore.
The same report links transformational leadership with better outcomes. Leaders who set a compelling direction, stimulate teams intellectually, and provide support are associated with higher productivity and organizational performance, higher satisfaction, and lower burnout (DORA, 2024). Again, the caveat holds. These are associations, not experiments. For executives deciding what to fund, the message is pragmatic: coaching leaders and stabilizing the portfolio are safer bets than buying another methodology rollout.
Measure what you change, then change it again
If scaling recipes are shaky and leadership context dominates, how should firms steer? By instrumenting their own experiments. The strongest proponents of measurement in software delivery, Nicole Forsgren, Jez Humble, and Gene Kim, argue that delivery team performance is measurable and that specific organizational capabilities drive higher performance (Forsgren, Humble and Kim, 2018). Their book, based on four years of State of DevOps research from 2014 to 2017 and analyzed with statistical methods, positions software delivery as a genuine source of competitive advantage and sets out a metrics-led path to improve it (Forsgren, Humble and Kim, 2018). The authors make causal claims, which you should read as their claims, not settled fact. That said, the thrust is directly useful: build a measurement spine and use it to run controlled changes.
Put the strands together and a pattern emerges. The review literature says customization and leadership backing matter more than mimicry (Dikert, Paasivaara and Lassenius, 2016). The 2024 survey says unstable priorities and weak leadership climate correlate with worse outcomes, regardless of how tidy your process looks on paper (DORA, 2024). The measurement camp says pick clear delivery metrics and use them to test and refine operating practices in your context (Forsgren, Humble and Kim, 2018). That is not a process installation. It is a continuous, instrumented redesign of how work flows.
A practical cadence for evidence, not ceremony
Here is a four-step cadence senior teams can run in ninety-day cycles.
-
Define the outcome that matters. Choose one outcome the business will feel, such as reducing employee burnout risk, improving perceived productivity, or increasing organizational delivery performance. Use your existing proxies, then lock them for the quarter. This reflects the DORA focus on productivity, performance, satisfaction, and burnout as management-relevant outcomes (DORA, 2024).
-
Identify one managerial variable to test. Stabilize priorities in a pilot portfolio for three months. Or invest in transformational leadership behaviors for a cohort of managers. The literature flags both priority stability and leadership climate as levers worth testing, with the correlational caveat noted (DORA, 2024).
-
Customize locally and write down the hypothesis. Document how a given unit will adapt roles, cadences, and decision rights to its context. The systematic review highlights customization, coaching, and alignment as frequent success factors in large-scale settings (Dikert, Paasivaara and Lassenius, 2016). Make the expected effect explicit: “If we stabilize priorities to a monthly intake, perceived productivity will rise.”
-
Instrument and compare. Establish a baseline, then compare pilot and control groups on your chosen outcomes. The measurement stance set out by Forsgren, Humble and Kim encourages teams to treat delivery as something you can measure and improve with targeted capability changes, even as you remain clear that the causal map in your own context must be discovered, not assumed (Forsgren, Humble and Kim, 2018).
This cadence is intentionally light on ceremony. You can run any standup you like. What matters is that leadership picks a variable the literature implicates, treats the intervention as a hypothesis, and funds the comparison. Most of all, executives must protect teams from mid-flight priority churn or the signal will drown in noise, a point underscored by the persistence of the instability effect even when other conditions look good on paper (DORA, 2024).
The practical read is stark. Stop installing agile. Start running it like a series of small, well-instrumented bets that fit your context, backed by leaders who mean it. The evidence we have points to culture, alignment, and leadership as the hard parts, not the ceremonies. Treat frameworks as drafts. Put your energy into stabilizing what teams are asked to do, building managerial habits that people experience as supportive, and measuring the outcomes you care about. That is how agile moves from theater to throughput.
Sources
- Dikert, Paasivaara and Lassenius (2016). Challenges and success factors for large-scale agile transformations. A systematic review of 52 publications and 42 cases that finds a weak evidence base and highlights management support, customization, training, mindset, and alignment as frequent success factors. https://www.sciencedirect.com/science/article/pii/S0164121216300826
- DORA (2024). Accelerate State of DevOps Report 2024. A large cross-sectional survey that finds priority instability correlates with lower productivity and higher burnout, and that transformational leadership correlates with better outcomes, with clear caveats about causality. https://dora.dev/research/2024/dora-report/
- Forsgren, Humble and Kim (2018). Accelerate: The Science of Lean Software and DevOps. The authors argue, based on 2014–2017 State of DevOps data, that measurable delivery performance and specific organizational capabilities drive improvement and can confer advantage. https://itrevolution.com/product/accelerate/