Why engineering project management needs a new data model

Engineering organisations have never had more project data, yet most portfolio decisions still lean heavily on experience, intuition and spreadsheets.

Spreadsheets are flexible, but they don’t truly remember anything. When a project closes, the nuanced context behind its hours is largely lost:

• Why did design burn more time than expected?
• What changed between the original forecast and the allocations?
• Which characteristics made this project hard, and which made it straightforward?

ProjoLink’s Characteristic-Based Resourcing (CBR) and the upcoming Kaizen-4™ algorithm are designed to change that. Together, they turn your projects into a structured learning system where every month of every project contributes to better decisions next time.

This article outlines how that works, what is live in ProjoLink today, and what is coming next with Kaizen-4™ for machine-learning-driven suggested hours.

From tasks and phases to Characteristics

Traditional tools focus on tasks, phases and work breakdown structures. CBR adds another layer: project characteristics.

Instead of relying on free-text descriptions, ProjoLink asks teams to describe each project using configured, structured fields that capture complexity and context in a repeatable way. These characteristics live alongside the standard ProjoLink data model of:

• Budgeted hours
• Forecasted hours
• Allocated hours
• Actual hours

…for every role, for every month of the project.

Over time, that creates a rich dataset: not only “how many hours were spent”, but what the environment looked like when those hours were planned and delivered.

Three characteristic types engineered for real projects

Engineering work is rarely one-size-fits-all. CBR supports three characteristic types that can be configured per function, per discipline and per project type.

1. Selection characteristics

These represent categorical choices – defined sets of options where each value implies a different level of complexity or effort.

Examples of Selection characteristics include:

• Motion Platform type for a simulator (e.g. 3-DOF, 6-DOF, static rig)
• Development approach (e.g. new design, derivative, minor update)
• Testing strategy (e.g. pure simulation, hardware-in-the-loop, full vehicle test)

Selection characteristics help standardise language. When Mechanical Design chooses a Motion Platform value, or Controls selects a certain architecture pattern, the organisation can later analyse how those choices correlate with hours.

2. Numerical characteristics

These capture quantities and counts – the things that typically scale effort.

Examples of Numerical characteristics include:

• Number of driver-feedback scenarios to configure
• Number of ECU interfaces to integrate
• Number of new modules or components introduced
• Number of external suppliers involved

By recording these values as numbers rather than text, ProjoLink prepares them for analysis. The Kaizen-4™ algorithm can later see, for instance, how hours tend to scale as the count of interfaces or modules grows.

3. Boolean (True/False) characteristics

These capture binary decisions or conditions:

• New HMI elements required: True/False
• Safety-critical functionality involved: True/False
• New regulatory standard applied: True/False

Boolean characteristics are powerful flags. They let the system ask questions such as, “When this flag is True, how do the hours and variances behave compared with when it’s False?”

Function-aware and AI-ready

Characteristics are defined with the function in mind – Mechanical Design, Electronics & Control Systems, Simulation, and so on. Each function can maintain a library that reflects its own complexity drivers.

Characteristics can also be flagged for AI use, meaning they are included as inputs when training and running the Kaizen-4™ algorithm. This keeps noise to a minimum and focuses the model on the characteristics that the organisation agrees are meaningful.

CBR today: ready to train your organisation-specific model

In the current ProjoLink product, CBR is live and ready for engineering organisations to use in production:

• Teams can configure characteristic libraries across functions.
• Each project month can be tagged with characteristic values for each role or function.
• Budgeted, forecasted, allocated and actual hours are collected centrally, month by month.

That combination makes ProjoLink behave like a project feature store:

• Every month of every project becomes a structured data point.
• Each data point links: {project context, characteristics, role, month, hours in all four states}.

Organisations can already begin training their internal models and building up complex data retention and context. Even before automated suggestions appear, CBR provides value:

• It standardises how complexity is described.
• It enables consistent reporting on “projects with similar characteristics”.
• It preserves project knowledge beyond the people who happened to work on it.

The next step is to put this data to work with machine learning.

The Kaizen-4™ Algorithm: continuous improvement built on four hour signals

Kaizen-4™ is ProjoLink’s upcoming machine-learning engine, built specifically for multi-project engineering environments.

At a conceptual level:

Kaizen-4™ captures context on the hours budgeted, forecasted, allocated and actual for every role, every month. It observes how these hours move in relation to the project’s characteristics and uses that understanding to propose improved suggested hours for similar projects or upcoming months.

The “4” in Kaizen-4 refers to the four hour states that ProjoLink already tracks:

1. Budget
2. Forecast
3. Allocated
4. Actual

These are the four signals the algorithm learns from.

Conceptual view of how Kaizen-4™ works

Without diving into low-level model types, the lifecycle looks like this:

1. Capture context
   • For each project, month and role, ProjoLink records:
     • Characteristic values (Selection / Number / True-False)
     • Budget, Forecast, Allocation and Actual hours
   • This forms a structured dataset over time.
2. Observe relationships and variances
   • The algorithm looks for patterns:
     • How do hours typically relate to specific characteristic combinations?
     • Under what conditions does Forecast consistently undershoot Actual?
     • When characteristics change mid-project, how do the hours respond?
3. Infer causes and adjust internal understanding
   • If hours or characteristic values change, the model compares before/after behaviour.
   • It updates its internal understanding of which characteristics matter most for each function and role, and how strongly they influence the hour profile across the four states.
4. Generate suggested hours and guardrails (upcoming output features)
   • For a new project or future months of an existing one, Kaizen-4™ can estimate:
     • Suggested hours per role and month
     • How those suggestions might differ from an existing forecast
     • Where risk of variance is highest
   • These outputs are being developed in a way that stays explainable and reviewable by humans rather than acting as a black box.

Kaizen-4™ is not intended to replace engineering judgement. It is designed to augment it with evidence from the organisation’s own history.

Current status: in private beta with selected design partners

The underlying Kaizen-4™ engine is currently in private beta with selected design partners.

For those partners:

• ProjoLink is using their historical and live CBR data to train early versions of the model.
• Output formats such as suggested hours, variance risk flags and confidence indicators are being tested and refined.
• The focus is on getting the interaction right: clear explanations, sensible default behaviour and guardrails that fit real project workflows.

For all ProjoLink customers:

• CBR is fully available today for configuring characteristics and collecting high-quality data.
• Organisations can start building the dataset that Kaizen-4™ will learn from once outputs are progressively rolled out beyond the private beta.

This staged approach keeps the algorithm close to reality while allowing leadership teams to prepare their processes and data practices.

How this changes the game for Heads of Project Management

For a Head of Project Management, the combination of CBR + Kaizen-4™ aims to deliver several strategic advantages.

Benefits

• Consistent forecasting language
  • Teams describe complexity in the same structured way, making cross-project comparisons meaningful.
• Data-driven hour baselines
  • Forecasts reference what actually happened on projects with similar characteristics, not just how people feel about the work.
• Improved capacity and portfolio planning
  • When suggested hours and variance patterns are understood, portfolio-level loading and hiring decisions become more grounded.
• Knowledge retention beyond individuals
  • The reasons past projects were hard or easy are captured as characteristics and hours, not just in people’s memories.
• Structured path to continuous improvement
  • Month after month, Kaizen-4™ refines its understanding. This builds a feedback loop between planning, execution and learning.

Considerations and trade-offs

• Data quality and discipline
  • The value of Kaizen-4™ depends on characteristics being filled in consistently and hours being recorded accurately. Organisations need clear ownership and governance.
• Change management
  • Moving from spreadsheet-centric approaches to structured characteristics requires buy-in from project managers and function leads. Training and internal communication matter.
• Model maturity over time
  • Early models are less confident; the strongest insights come after a critical mass of project months have accumulated. Leaders need to view this as a compounding asset, not a one-off feature.
• Complement, not replacement
  • Machine-learning suggestions should be treated as a second opinion that can be challenged and refined, not an automatic replacement for expert judgement.

Leaders who plan for these considerations can extract significantly more value once Kaizen-4™ outputs become generally available.

Preparing your organisation for Kaizen-4™

If you want to be ready to flip the switch on ML-driven suggestions as soon as they are available, there are practical steps you can take now:

1. Define and align on characteristics
   • Work with your functions to define Selection, Numerical and Boolean characteristics that genuinely drive effort.
   • Decide which ones should be flagged for AI use.
2. Adopt a disciplined monthly cycle
   • Ensure that Budget, Forecast, Allocation and Actual hours are being maintained in ProjoLink each month.
   • Treat variance analysis as a standard management routine.
3. Embed CBR in project setup
   • Make characteristic tagging part of how new projects are initiated, not an optional extra.
4. Monitor data completeness
   • Use ProjoLink’s views and dashboards to watch for missing or inconsistent characteristics and hours.
5. Plan how you will use suggestions
   • Decide in advance: when Kaizen-4™ provides suggested hours, who reviews them, how they are accepted or adjusted, and where that feedback is captured.

These steps create an environment where machine learning can offer meaningful, high-trust recommendations instead of noisy outputs.

Where this sits among other approaches

Project leaders typically have three broad options:

1. Stay with spreadsheets and ad-hoc judgement
   • Maximum flexibility, minimal upfront change.
   • Little organisational learning; context is hard to reuse.
2. Standardise templates without ML
   • Improves consistency and governance.
   • Still heavily manual, and still limited in how it learns from history.
3. Adopt CBR with an ML layer like Kaizen-4™
   • Requires structured data and change management.
   • Builds a reusable asset: a project-specific learning system driven by your own history.

ProjoLink is deliberately designed to support the third path while still feeling familiar to teams used to traditional project planning tools.

Next steps

If you’re leading a portfolio of engineering projects and want your forecasting and resourcing decisions to be informed by more than instinct and isolated spreadsheets, now is the right time to explore CBR and Kaizen-4™.

• CBR is available in ProjoLink today for configuring characteristics and capturing rich project context.
• Kaizen-4™ is in private beta with selected design partners, and its outputs will be rolled out progressively as the models mature.

You can book a demo through the EfficiaFlow website to see how CBR works in your environment and discuss how your organisation could participate in, or benefit from, the Kaizen-4™ rollout.

Question to reflect on:
If every project month your organisation delivers from today onwards contributed to a smarter, more accurate forecast for the next one, how different would your portfolio decisions look two years from now?

‍