A calculation can be mathematically correct and still be unsuitable for design if nobody can tell what conditions it represents. Knowing how to write design assumptions is therefore not a documentation exercise added at the end. It is part of defining the engineering problem before formulas, software outputs, or sizing decisions are allowed to carry weight.
For a reviewer, assumptions answer a direct question: what has to be true for this result to be valid? For the engineer who returns to the work six months later, they explain why a particular model, load case, material value, or boundary condition was selected. Good assumptions make a calculation sheet easier to check, revise, and reuse without silently extending its intended scope.
What design assumptions should do
A design assumption is a stated simplification, interpretation, or condition adopted where the real system cannot be modelled in full detail or where project information is incomplete. It should connect directly to the method and the result.
For example, a beam deflection check might assume that the member is prismatic, linearly elastic, laterally restrained, and subject to small deflections. These are not decorative statements. They determine whether elementary beam theory is an appropriate model and whether the calculated deflection can be used for the decision at hand.
Assumptions are often confused with inputs, requirements, and design criteria. Keep the distinction clear. An input is a known or selected value, such as a 6 m span or 10 kN/m imposed load. A criterion is the pass-fail rule, such as a deflection limit defined by the governing standard or project brief. An assumption explains the modelling position behind the calculation, such as treating a connection as pinned or neglecting secondary bending.
When those categories are mixed in one unlabelled note, checking becomes slow. A reviewer has to infer which values are fixed by the brief, which are taken from a standard, and which are engineering judgement.
Start with the decision, not a standard disclaimer
The fastest way to produce weak assumptions is to begin with generic wording such as “all work complies with relevant codes”. That statement is too broad to be checked and does not identify the actual calculation boundary.
Start by writing the decision the sheet is intended to support. It may be to establish the required bolt diameter, verify a steel beam, determine pump head, or check the capacity of a concrete footing. Then identify the conditions that materially affect that decision.
For a bolted flange calculation, the design assumptions might address whether preload is present, whether joint members remain in elastic compression, how external load is distributed, the friction condition, and whether fatigue is within scope. For a footing check, they may address soil stratification, groundwater level, load combinations, bearing pressure distribution, and whether long-term settlement is excluded.
This approach prevents an assumption register from becoming a copied list of boilerplate. It also creates a useful test: if changing a statement would not alter the method, input selection, applicability, or interpretation of results, it probably does not belong in the design assumptions section.
Use a structure a reviewer can verify
Write each assumption as a short, testable statement. It should state what is being assumed, where it applies, and why the assumption is acceptable or what limits it creates. Avoid vague qualifiers such as “reasonable”, “typical”, or “adequate” unless the calculation sheet also defines the basis for that judgement.
Compare these two statements:
“Connection effects have been considered.”
“Beam-to-column connections are modelled as pinned for global analysis. Connection rotational stiffness is excluded from the frame model; member forces are therefore suitable only where simple-connection behaviour is detailed.”
The second statement is longer, but it can be checked. It identifies the modelling choice and the consequence. It also alerts a later designer that the detailing must not contradict the analysis model.
A useful format for significant assumptions is:
Assumption: State the adopted condition.
Basis: Identify the drawing, survey information, standard clause, manufacturer data, precedent, or engineering reasoning that supports it.
Effect or limitation: State what the assumption permits, excludes, or requires elsewhere in the design.
Not every simple calculation needs all three labels. For high-consequence work, novel analysis, or calculations expected to be reused, this structure is worth the extra space.
Separate known information from engineering judgement
Design work frequently begins before all site and supplier information is available. That is normal. The risk arises when provisional values are presented as facts.
Label information according to its source. Site dimensions taken from an issued drawing are different from dimensions inferred from a photograph. A material grade specified in the contract is different from a grade selected for a preliminary check. Soil parameters from a geotechnical report are different from values adopted pending investigation.
A clear assumption can read: “Founding level is assumed to be 1.2 m below finished ground level, pending confirmation from the geotechnical investigation. Bearing-capacity results are not valid if made ground extends to the proposed founding level.” This statement is more useful than simply entering 1.2 m in a calculation cell, because it records the dependency that must be closed before issue.
Use dates and document revisions where they matter. “Based on architectural drawing A-104 Rev C” is easier to manage than “based on drawings”. If the design develops, readers can identify whether the calculation needs to be revisited.
Make the model boundaries explicit
Most design errors are not arithmetic errors. They are model-boundary errors: an omitted load path, a restraint assumed but not detailed, a simplified geometry applied outside its valid range, or a local effect hidden within a global model.
State the boundaries of the analysis. In structural work, this can include support conditions, member continuity, diaphragm action, second-order effects, composite action, load redistribution, and the treatment of openings. In mechanical work, it may include steady-state versus transient operation, temperature range, vibration, wear, fluid properties, and manufacturing tolerances.
There is no need to claim that every effect is negligible. Some effects may be excluded because they are checked separately. Say so. For example: “Global bending is assessed using a line-beam model. Local web bearing and stiffener requirements are checked independently.” That sentence prevents a reviewer from assuming a single model covers both behaviours.
Be particularly careful with phrases such as “fixed support” and “uniform load”. They may be valid idealisations, but real fixity and real loading rarely match the phrase perfectly. Define the physical detail or loading arrangement that justifies the idealisation, and identify the point at which it stops being credible.
Put assumptions beside the calculation they govern
A front-page basis-of-design note is valuable, but assumptions should not live only there. Place critical assumptions close to the formula, diagram, or result they affect. A note beside a buckling calculation can identify the effective length basis. A note beside a pressure-loss calculation can state the fluid temperature and roughness value used. A sketch can show the assumed load distribution more clearly than a paragraph.
This is where readable calculation documents outperform spreadsheet sprawl. In a tool such as Calculeaf, unit-aware equations, explanatory notes, images, and plots can sit in the same worksheet. The reviewer can see the assumption, calculation path, and output together rather than tracing references across multiple tabs.
Proximity also reduces revision risk. If a changed support condition modifies a moment equation, the related note is visible at the point of change. The author is less likely to update an input while leaving an old narrative statement behind.
Handle standards and units with precision
Do not write “designed to Eurocode” or “designed to the relevant standard” as a substitute for a calculation basis. State the standard edition, applicable National Annex where relevant, the design situation, and the specific method where interpretation is not obvious. Project specifications may override default standard values, so record that hierarchy.
Units deserve similar treatment. A unit-aware worksheet can prevent many conversion mistakes, but it cannot determine whether a pressure was supplied as gauge or absolute, whether a nominal pipe size was mistaken for an internal diameter, or whether a load is characteristic, factored, or service-level. These are design assumptions and should be stated in words.
Where SI and US customary source data are mixed, show the original source value and the converted value where this aids review. Avoid manual conversion factors embedded inside formulas. The calculation should reveal the quantity and unit being used, not just the final number.
Review assumptions as design inputs change
Assumptions are controlled design information, not a one-time preamble. Review them when the geometry, load basis, supplier selection, site data, governing standard, or calculation method changes. A revision may not require recalculating every result, but it does require a conscious check of whether the original model remains valid.
Before issuing a calculation, ask four practical questions: can another engineer identify every material assumption; is each one supported or clearly marked for confirmation; does each significant assumption have a visible effect on the method or result; and have exclusions been assigned to another check or discipline? If any answer is no, the document is not yet ready for efficient review.
The best design assumptions do not make an analysis look more cautious. They make its limits visible, allowing the next engineer to use the calculation with confidence and to recognise quickly when a different model is required.