Observational Method Eurocode 7: monitoring in project practice

The observational method under Eurocode 7 (PN-EN 1997-1, p. 2.7) is a geotechnical design approach in which the design is verified and, if needed, changed during construction on the basis of measurements. It may be used only if acceptable limits of behaviour, a monitoring plan and a contingency plan were set before work started, and if measurement results reach the decision makers faster than the structure can cross the limit.

In short

• The observational method is not "design by eye". It is a design with alternatives plus monitoring that shows which alternative applies during the works.
• Eurocode 7 sets conditions before construction (limits, expected range of behaviour, monitoring plan, response time, contingency plan) and during construction (measurement, assessment, action, serviceable equipment). Without any one of them, there is no observational method, only hazard.
• The condition most often forgotten: the response time for measurements and procedures must be shorter than the time the system needs to go from "within limits" to "outside the limit".
• The gain: fewer struts, anchors and construction stages at the conservative end. The price: ready contingency actions, people with decision authority and measurement that works without an operator.
• The method is not suitable for sudden phenomena that do not allow time for observation. Those mechanisms must be designed conservatively regardless of monitoring.

"Safe" design costs money, and nobody counts it

Soil is not steel. You take strength and deformation parameters from a few boreholes and a handful of tests, then extrapolate them across the whole excavation. A designer who selects cautious characteristic values and applies partial factors is acting rationally. Yet every margin for uncertainty has a material cost: an extra level of struts, a stiffer wall, more anchors, pre-stressing, longer construction stages under struts, and clashes with the foundation slab.

This cost is invisible when the design is being prepared, because nobody prepares a version for "what if the ground behaves as it most likely will". A conservative design assumes that the adverse scenario governs. If reality turns out better, nobody will know, because nobody measured it.

The observational method reverses that logic. Instead of locking uncertainty into a margin, it locks it into a procedure: you design for the most likely scenario, prepare a variant for the adverse scenario, and measure which one is unfolding. The idea is not new. Ralph Peck described it in his 1969 Rankine Lecture, and Eurocode 7 brought it into the standard with a specific list of requirements.

If you are a designer, you know the other side of that coin: your name is on the design. If you propose two levels of struts instead of three and something goes wrong, nobody will ask how much steel the investor saved. That fear is justified, and the observational method does not erase it. It moves it from intuition into procedure: the risk is named, calculated in two variants, backed by measurement and ready action. An observational design that meets the standard's conditions is better documented than a conservative design, because after construction you have not only calculations, but also a record of how the structure actually behaved.

From a decision maker's point of view, such as a contract director or investor, this is a shift from one type of cost to another: less steel and time in exchange for monitoring, organisation and readiness to intervene. Whether it pays depends on the spread of possible behaviours and the cost of keeping contingency actions on standby. That decision is made with eyes open, not because "we will save on struts". For you, it means one thing: before you calculate savings, calculate what it costs to secure them.

The observational method in 6 steps (PN-EN 1997-1, p. 2.7)

The observational method is a design approach in which the geotechnical design is reviewed during construction on the basis of measurement results, and pre-prepared contingency actions are triggered if the behaviour of the structure or ground goes beyond the set limits.

The standard allows it where forecasting geotechnical behaviour is difficult. Below is a paraphrase, not a quotation, of the requirements in clause 2.7, arranged into six steps. The first five concern the stage before construction begins, and the sixth concerns the construction period. Note on numbering: the second generation of Eurocodes is being implemented (EN 1997-1:2024). It keeps the observational method as one of the recognised design verification methods, but changes the clause numbering and gives monitoring a separate chapter on project implementation during execution and use. When referring to the standard in documentation, use the edition that applies to your project.

Step 1. Set acceptable limits of behaviour

The standard requires that acceptable limits of behaviour be defined before construction begins. For an excavation, these are, for example, the allowable horizontal wall displacement, settlement of an adjacent building, force in a strut or anchor, or water level behind the wall. The limits follow from ultimate and serviceability limit states of the support system and from the sensitivity of the surroundings. A limit is not an alarm threshold. The threshold comes earlier and gives time to react.

Step 2. Assess the range of possible behaviour and the probability

The standard requires an assessment of how widely the system may behave, and proof that there is an acceptable probability that the actual behaviour will remain within the defined limits. In practice, this means at least two calculation sets: the most likely scenario and the least favourable scenario that is still realistic. Design variants are born from that pair.

Step 3. Design the monitoring plan

The monitoring plan must reveal whether the actual behaviour stays within the limits, and do so early enough and at sufficiently short intervals to allow contingency actions to be taken effectively. The selection of instruments and measurement sections for an excavation is described in our guide to geotechnical monitoring of a deep excavation.

Step 4. Ensure a sufficiently short response time for measurements and procedures

The standard requires the response time of the instruments and the procedure for analysing results to be sufficiently short relative to the possible evolution of the system. This is the sentence most often overlooked in the whole of clause 2.7, and the one that decides whether the observational method is possible at all.

Step 5. Prepare a contingency plan

If monitoring reveals behaviour outside the limits, there must be a contingency plan prepared in advance, ready to be implemented. "Prepared" means designed, calculated, with material and equipment available in a time shorter than the system evolution time, and with named decision makers.

Step 6. During construction: measure, assess, act, repair the equipment

During the works, measurements are taken in line with the plan, results are assessed at appropriate stages, and when limits are exceeded, the planned contingency actions are triggered. The standard adds a requirement that is rarely remembered: equipment that does not provide reliable data of the right type or in sufficient quantity must be replaced or supplemented. A sensor that has gone silent is not neutral. In an observational project, it is a hole in the protection.

National rules: geotechnical categories and monitoring scope

The Regulation of the Ministry of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the founding of building structures (§ 4) introduces three geotechnical categories. Excavations, retaining structures, embankments and ground anchors are classified at least as category II. Category III includes, among others, structures in complex ground conditions, high-rise buildings designed within existing urban development, and tall structures founded deeper than 5 m or with more than one underground storey. The category is set by the designer.

For categories II and III, the regulation requires a geotechnical design (§ 7(2)), in accordance with PN-EN 1997-1 and PN-EN 1997-2, which must define the scope of necessary monitoring of the constructed structure, neighbouring structures and surrounding ground (§ 10(10)). A deep excavation for a multi-storey car park in dense development often meets the criteria for category III, and the requirement to define the monitoring scope naturally complements the observational method.

The necessary condition: data must arrive before the system has changed

In the observational method, two clocks are running. The first is the system clock: how long the wall needs to move from the warning value to the limit of acceptable behaviour. The second is the organisation clock: how much time passes from the moment the structure crosses the threshold to the moment a decision maker knows about it and acts. The observational method is allowed only when the second clock is clearly faster than the first.

Let us write that as a simple condition, illustratively, not as a standard formula:

t_measurement + t_transmission + t_analysis + t_decision + t_action < t_evolution

where t_evolution is the time the system needs to go from the warning threshold to the limit. Each component on the left can be shortened, but only to a point. A human decision takes hours, and installing an extra strut takes days. Only measurement, transmission and initial analysis can be reduced from days to minutes. That is exactly why fast data is a condition, not an add-on: it is the only term in the sum that you can reduce by two orders of magnitude.

Periodic measurement and the 15-minute loop

As an illustration, assume that at the current rate of excavation a diaphragm wall can move from the warning threshold to the limit in two days. Periodic measurement with an inclinometer probe once a week, with office processing and an email the next day, gives a loop counted in days and a blind window counted in a week. The same excavation with a chain inclinometer (in-place) in continuous monitoring mode, read at the default 15-minute interval, with thresholds on the channels and an SMS to the on-duty engineer, has an event-to-knowledge loop of around 15 minutes. The arithmetic is simple: a weekly reading equals 672 windows of 15 minutes, and a daily reading equals 96 such windows. We show how to calculate your own multiplier for a specific structure in the post "100x faster" - from event to decision.

Not every measurement has to be every 15 minutes by default, or more often. The standard says "sufficiently short relative to the system evolution". In calm phases, such as a completed slab and all levels of struts installed, a daily reading may be enough. In critical phases, between reaching the level of the struts and installing them, or while lowering the groundwater table, the system can change from hour to hour. The monitoring plan should provide different frequencies for different phases. The simplest way to do that is automation that measures continuously while a person decides what to look at.

Speed without procedure does nothing

Data that arrives every 15 minutes by default, or more often, but is read on Monday morning, does not shorten any loop. That is why the monitoring plan must include a procedure alongside frequency: who receives a warning threshold notification, who receives an alarm, who confirms receipt, and who has the authority to stop the works. How to set thresholds so that an SMS means something is described in our post on warning and alarm thresholds. What matters here is one thing: the warning threshold must be early enough for the organisation loop to fit before the limit.

The organisation clock must also tick when data does not arrive. NO_DATA, meaning no reading from a channel for longer than the set time, is an event requiring response in an observational design, because step 6 of the standard requires the equipment to be repaired or supplemented, and until that happens the excavation is uncontrolled. The conclusion for you is simple: in the monitoring plan, record how often you measure, and next to that, who reacts to each status and within what time, including silence.

Contingency plan: trigger -> action -> who -> within what time

A contingency plan is not a paragraph saying that "appropriate steps will be taken in the event of an exceedance". It is a matrix in which each row has a trigger, a specific action, the decision maker and the time within which the action must start. The table below is a fill-in template: the ellipses are fields for values from your project, and the times in brackets illustrate the order of magnitude, not a recommendation.

Three notes to the template. The structural action must be feasible within the stated time: strut sections on site, details completed, equipment available. If bringing in an extra strut takes a week, the warning threshold must lie accordingly earlier, or the strut must already be on site. The decision to stop the works belongs to the people present on site, and the structural decision belongs to the designer. The plan separates these roles so that nobody waits for the other. Record every activation of an action: when the notification arrived, who confirmed it, what was done. That log is proof that the method was applied, not just declared. We write about how such records stand up in disputes in the post on monitoring data as evidence.

Illustrative example: a braced excavation, "conservative" versus "observational" design

The example below is hypothetical. The numbers illustrate the decision mechanism. They do not come from any implementation and are not a design recommendation.

Imagine an excavation about 10 m deep with a diaphragm wall, in dense urban development, with a several-metre layer of clay of uncertain parameters: tests gave a wide spread of results that cannot be narrowed cheaply. Calculations in two variants show that, for the most likely parameters, two levels of struts are enough, while for the least favourable parameters three levels plus pre-stressing are needed.

In the observational variant, the monitoring plan includes inclinometers in the wall (in critical phases, chain inclinometers, read every 15 minutes by default, or more often), force sensors or vibrating wire strain gauges on the struts with temperature compensation, piezometers behind the wall, and measurement of adjacent buildings. Behaviour limits follow from calculations for the realised variant. Thresholds are set below the limits. As an illustration, the warning threshold is around 70% of the limit value and the alarm threshold is around 90%, although threshold selection always belongs to the designer and depends on the system's rate of evolution.

If the wall displacement profile after reaching the level of the second strut lies clearly below the warning threshold, and the forces in the struts stay within the forecast for the likely variant, construction continues without the third level. If the curve moves toward the threshold or the increment in displacements rises from phase to phase, you trigger the row in the table: stop excavation, install the third level. Then the cost is the same as in the conservative design, plus monitoring and delay. You take that risk knowingly, and it must be counted in the decision, not just the savings on steel.

There is also a third possibility: the observations show behaviour better than in the likely variant. Then the designer has an argument to, for example, omit pre-stressing of the second level. Without measurements, that information does not exist. What does that mean for you? In each of the three outcomes, you have a record of what you based the decision on, and that is something a conservative design does not provide.

When savings on struts are safe: six conditions

The investor wants to save on struts and anchors. The designer is afraid to take on the risk. Both are right, and both need the same thing: a list of conditions under which a less conservative variant is an engineering decision, not a gamble. They follow directly from clause 2.7 of the standard and from practice:

1. The failure mechanism is progressive. The adverse scenario appears as increasing displacement, force or water pressure that can be measured before the limit is reached. Sudden mechanisms are excluded from the method and must be designed conservatively.
2. Limits and thresholds are calculated, not assumed. The limit comes from calculations for the realised variant; the warning threshold lies early enough for the whole organisation loop to fit.
3. The data loop is shorter than the system evolution. In critical phases this means continuous monitoring with automatic alerting, not a manual weekly reading.
4. The contingency action is physically ready. Calculated, with material on site and equipment available within a time shorter than t_evolution.
5. Decision authority is defined before construction. Who stops the works, who decides on the structure, who pays for the contingency action, how they can be reached outside working hours. Design supervision is in the plan, not called in after the fact.
6. Every reading, status and decision leaves a trace. The log of notifications and confirmations, raw data and configuration change history are available after construction.

Fast data does not reduce risk on its own. It shortens the window in which the risk is invisible, and that is exactly what makes conditions 2, 3 and 6 feasible. If one of the six conditions cannot be met on your site, the answer is: conservative design. That is also a good decision, provided it is conscious. A bad decision is a "nearly observational" design: savings on struts without the protection that was meant to justify them.

Risks and limits: when not to use the observational method

The observational method has boundary conditions that no good sensor can bypass.

Sudden phenomena. Base instability of an excavation, hydraulic heave, brittle anchor failure, soil liquefaction: these are mechanisms that do not give warning in measured quantities, or give it too late. If the variant analysis shows that the adverse scenario is sudden, the system clock is shorter than any organisational loop, and the observational method is not allowed for that mechanism. Design it conservatively, and monitor the rest.

Lack of organisation and decision authority. The observational method is also a contract. Who pays for the contingency action if it has to be triggered: the contractor, the investor, the insurer? Who has the right to stop excavation at 6 a.m. on Saturday without calling three people? Without written answers before the works begin, the warning will arrive on time and the decision will not. Any design change during construction requires the designer's involvement, so design supervision must be in the plan, not called in after the fact.

Monitoring treated as a cost to be cut. Fewer sections, manual reading instead of automatic, no NO_DATA states, no duty coverage. The design stays less conservative, and the protection that was meant to compensate disappears. That is worse than a conservative design without monitoring.

Observation "after the fact". Since Peck's time, the literature has distinguished between applying the method from the start and turning to it as a way out when the structure behaves unexpectedly. The standard requires limits, a monitoring plan and a contingency plan before construction begins. Monitoring added in panic after the first cracks appear in the neighbour is not the observational method. It is firefighting.

Thresholds without increments. A displacement level alone says little if you do not know how fast it is growing. The monitoring plan should also define the allowable increment between phases or per day, and specify who assesses it. Otherwise the limit may be crossed between two "green" readings. For you, this means that each of these limits can be checked at the design stage, before you order steel, and that is the right moment to decide.

What this looks like in Inclify

In an observational project, the platform handles the left-hand side of the condition from the response-time section: measurement, transmission, initial analysis and notification. Sensors on the excavation, such as chain inclinometers, vibrating wire strain gauges on struts, piezometers and tiltmeters on adjacent buildings, report through loggers every 15 minutes by default, or more often. The platform accepts any interval and checks data completeness against the actual cadence. For each channel you set a warning threshold and an alarm threshold with hysteresis, and the OK / WARNING / ALARM / NO_DATA states correspond to the rows in the contingency plan. SMS, email and in-app alerts go to project users according to their preferences. The person taking over the ticket confirms it in the platform, with who and when recorded, and silencing an alarm always has a deadline, so nothing disappears quietly.

For chain inclinometers, the platform calculates the displacement profile from the anchor in axes A and B and the resultant, gives the maximum resultant displacement (you read the depth where it occurs from the profile), and after the first valid frame it creates a dashboard with the vertical profile and trend. You therefore compare it with the design limit on a chart, not in a spreadsheet. The chart panel with two axes shows strut force against temperature, which lets you separate daily steel "breathing" from a lasting trend before sending someone to site. The temperature compensation report performs that assessment for each channel.

One caveat to keep the plan realistic: platform alarms work on value and on missing data, not on rate of change. You assess the increment between phases on the trend chart and in the 7/30-day risk assessment report, so in the plan table assign the row "increment > ... mm/day" to the on-duty engineer, not to automation. The trace remains: the alarm history records each status change and each confirmation, the audit log records every threshold and configuration change (value before and after), and you export table data to CSV for construction documentation. The communication log with raw frames from devices (default 7 days) decides whether the reading actually arrived, and the data SLA report shows gaps and freshness on each channel. That is your tool for enforcing the requirement in step 6 for equipment that provides reliable data.

We explain how a chain inclinometer differs from a probe and how to read a displacement profile in the post chain inclinometer or inclinometer probe, and the overall system picture in the complete guide to structure monitoring.

Sources and further reading

• PN-EN 1997-1:2008 Eurocode 7: Geotechnical design. Part 1: General rules, p. 2.7 "Observational method" and Chapter 4 - Polish Committee for Standardization, sklep.pkn.pl.
• EN 1997-1:2024 Eurocode 7: Geotechnical design. Part 1: General rules (second generation of Eurocodes) - CEN; national implementation in progress.
• Regulation of the Minister of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the founding of building structures (Journal of Laws 2012 item 463) - ISAP.
• Peck R.B., "Advantages and Limitations of the Observational Method in Applied Soil Mechanics", Géotechnique 1969, vol. 19, no. 2, pp. 171-187 (Ninth Rankine Lecture; classic source article on the method).
• Nicholson D., Tse C., Penny C., "The Observational Method in ground engineering: principles and applications", CIRIA Report R185, London 1999.
• Borecka A., Stopkowicz A., Sekuła K., "Metoda obserwacyjna i monitoring geotechniczny w świetle przepisów prawa...", Przegląd Geologiczny 2017, vol. 65, no. 10/2, pp. 685-691.

What next

If you are designing or procuring an excavation in geotechnical category II or III and are considering the observational method, you will gain the most by defining the monitoring plan and contingency plan together, at the detailed design stage, before ordering the struts. After that, the observational variant is only theory. Send us the construction phases and a draft monitoring plan, and we will show on the dashboard, for a project similar to yours, how thresholds, states, notifications and the inclinometer profile translate into rows in the contingency action table. If you already have sensors or loggers, they can be connected without replacing the hardware. Book a call with the Inclify team - we respond within 24 hours.