The four levels of analytics
Analytics is a progression from understanding past events to shaping future outcomes. Each level answers a specific question about your data: what happened, why it happened, what will happen next, and what to do about it. Each level builds on the one before, each costs more effort and integration, and each is worth more. Most construction reporting stops at level one.
What happened?
Hindsight. Summarises past and current events, where most reporting stops.
Why did it happen?
Insight. Probes deeper to find root causes, separating correlation from causation.
What will happen?
Foresight. Uses historical data and algorithms to forecast the likely future.
What should we do?
Action. Suggests specific interventions, augmenting human decision-making.
A00 Riverside Link Road: Project Controls Dashboard
The A00 Riverside Link Road is a fictional scheme built for training: every number below is illustrative, but the metrics, methods and data sources are standard project-controls practice. One project, one month-end dataset, four report pages, each one rung further up the ladder.
What happened?
Descriptive analytics summarises past and current events: the classic month-end pack of progress, spend, earned value and milestones. The A00 is six points behind plan and £3.4M over earned value. That is where a typical dashboard stops: a red RAG and a variance number, with nothing about why, where it is heading, or what to do about it.
Why did it happen?
Diagnostic analytics probes the reasons behind events, separating correlation from causation. Correlating the performance data with weather, telematics and design registers shows 87% of the variance sitting in two packages, and 62% of lost hours coming from two causes: rain stopping clay fill placement, and soft ground in one cutting. The fix is a production problem, not a commercial one.
What will happen?
Predictive analytics uses historical data and algorithms to forecast what comes next. Monte Carlo simulation of the risk-loaded schedule gives the baseline date a less than 10% chance of being met, a P50 outturn of £92.4M, and concentrates 80% of the uncertainty in two named drivers: exactly the root causes found at Level 2. Bad news, but cheap bad news. It arrived in June, not October.
What should we do?
Prescriptive analytics suggests specific actions, augmenting human decision-making. Each intervention option runs through the same risk model, so the decision is made on modelled evidence: spend £2.7M on a second crew and lime stabilisation to protect the opening date, avoid £2.3M of damages and prelims, and cut the P80 overrun by £4.3M. Doing nothing is a decision too, and it is the most expensive one.
The point of this page. The same month-end data sits behind all four reports. Level 1 told us we were red. Level 2 told us why. Level 3 told us it would cost £9.6M and 11 weeks if ignored. Level 4 turned it into a £2.7M decision with a deadline. Teams that stop at Level 1 hold a meeting about the colour red. Teams that climb the ladder buy back the opening date. Each level builds on the previous, each costs more effort and integration, and each is worth more. Don't get stuck at one level.
All project data is illustrative (the A00 Riverside Link Road is fictional) but the metrics, methods and data sources reflect standard project-controls practice: earned value management (PV/EV/AC, SPI/CPI, EAC), quantitative schedule risk analysis via Monte Carlo, Pareto root-cause analysis, and scenario-based decision support.
Ready to Make Best Practice Executable?
Book a discovery call to see how Synaform and Project:OS can make governance and delivery work for your programme.