An anchor reaching 30kN at failure does not have a 30kN working load. The number recorded when an anchor finally pulls out, breaks or causes the surrounding base material to fail is a raw test result. It is not automatically a design resistance, allowable resistance or safe working load.
That distinction sits at the centre of the Construction Fixings Association guidance and BS 8539. Modern anchor design separates the load applied to the fixing from the resistance available to withstand that load, and it applies statistical treatment and safety factors before raw test data becomes usable engineering information.
The terminology matters because construction sites still routinely use expressions such as working load, safe working load, ultimate load, proof load and pull-out capacity as though they describe interchangeable numbers. They do not.
Post-installed anchor ultimate load test showing concrete cone failure
Illustrative concrete cone failure around a post-installed anchor. An ultimate failure load is raw test data and should not be treated directly as a working, allowable or design resistance.
The key message: ultimate resistance is the load measured at failure. Working, recommended, allowable and design resistances are values used at different stages of engineering assessment after appropriate reductions or safety factors have been applied. A raw ultimate test result should never simply be copied into the design as the allowable working load.

Start With the Basic Distinction: Action vs Resistance

The CFA guidance adopts the terminology used in BS 8539 and the European anchor-design framework. Loads imposed on an anchorage by the supported fixture are described as actions. The ability of the anchor and surrounding base material to withstand those actions is described as resistance.
Demand Side Resistance Side
Characteristic action NEk
The characteristic or unfactored load applied to the anchor.		 Characteristic resistance NRk
A statistically derived resistance associated with the anchor and relevant failure mode.
Design action NEd
The characteristic action after the applicable action partial factors have been applied.		 Design resistance NRd
The characteristic resistance after the applicable material partial factor has been applied.
Under the partial safety factor approach, the engineering check is therefore based on the relationship between design action and design resistance, not between a working load and an arbitrary pull-test result.
Demand: NEk → apply action factors → NEd

Resistance: NRk → apply material factors → NRd

Design check: NEd ≤ NRd

The Anchor Load Terms That Should Not Be Mixed Together

BS 8539 and current CFA guidance distinguish a series of values that can appear on drawings, calculations, manufacturer data sheets and site-test reports.
Term Notation What It Actually Represents
Ultimate resistance / failure load NRu The load recorded when an individual anchor reaches failure during a destructive test.
Mean ultimate resistance NRu,m The average failure load from a series of tests.
Characteristic resistance NRk A statistically derived resistance. Where obtained from testing, it reflects both the average strength and the variability of the results.
Design resistance NRd Characteristic resistance reduced using the applicable material partial safety factor.
Recommended resistance Nrec The maximum working resistance quoted by a manufacturer for a particular anchor and base-material condition.
Allowable resistance NR,all A maximum working resistance derived from site testing where no appropriate manufacturer recommended resistance is available for the actual base material.
Characteristic action NEk The characteristic load demand from the fixture. Often referred to informally as the unfactored load.
Design action NEd The factored action used in the partial safety factor design check.
Proof load Np The specified non-destructive test load applied to a working anchor during proof testing.
Preliminary test load Ntest The calculated load applied to dedicated test anchors under the preliminary-test regime.
Load at first movement N1st The load at which movement of the anchor or fixture relative to the base material is first recorded.
The hydraulic gauge may display all of these values in kilonewtons, but that does not make them equivalent engineering quantities.

Why “Working Load” Can Be Ambiguous

The expression working load remains common on construction sites, but it can describe two different ideas depending on who is speaking. A contractor may use working load to mean the ordinary unfactored load imposed on the fixing during service. In current CFA terminology, this is more clearly described as the characteristic action NEk.
The term can also appear on older manufacturer literature as part of expressions such as safe working load or allowable working load, which describe a resistance rather than an applied action. That is why specifications should preferably state the actual engineering quantity and notation rather than simply asking for a test at a multiple of “working load”.
Site Expression Better Engineering Description
Working load applied to anchor Characteristic action NEk, if that is what the designer intends.
Safe working load from manufacturer Manufacturer's recommended resistance Nrec, where that terminology is used.
Safe working load from site tests Allowable resistance NR,all, when derived using the relevant BS 8539/CFA test regime.

What Does an Ultimate Anchor Test Actually Tell You?

An ultimate-load test deliberately continues until the anchor, anchorage or surrounding base material reaches failure. Depending on the fixing and substrate, the failure mode may include pull-out, bond failure, concrete cone breakout, splitting, steel failure, excessive movement or a combination of mechanisms.
The peak force measured during that event is the ultimate resistance NRu.
If five anchors fail at 18kN, 21kN, 23kN, 24kN and 25kN, the results tell the engineer considerably more than simply:
“The anchors take about 22kN.”
The results show the average strength, the lowest observed failure, the scatter between test locations and potentially the influence of variable material or installation conditions. Those characteristics affect the usable resistance that may eventually be assigned. CFA guidance therefore treats ultimate-load results as input data for an engineering assessment rather than as the final working capacity.

How Raw Ultimate Loads Become Engineering Resistance Values

For the statistical allowable-resistance route for anchors without a relevant ETA, current CFA guidance based on BS 8539 Annex B.2.3.2 requires between five and fifteen anchors to be loaded carefully to failure.
For each anchor the tester records:
  • load at first movement;
  • ultimate failure load;
  • mode of failure;
  • relevant condition of the anchor and base material.
The mean ultimate resistance is then statistically reduced. Current CFA guidance expresses the characteristic resistance for this route as:
NRk1 = αdist × NRu,m × (1 − K × v) × Î©
The factors serve different purposes:
Parameter Purpose
NRu,m Mean ultimate failure resistance from the test series.
K Statistical factor reflecting sample size.
v Coefficient of variation reflecting scatter between the test results.
αdist Modification factor used where reaction-bridge spacing could beneficially influence the measured result.
Ω Adjustment for application or environmental conditions relevant to the eventual anchorage.
The CFA guidance states that the coefficient of variation should not exceed 30% for this statistical route.

Sample Size Changes the Statistical Penalty

Number of Tests K Factor
5 3.40
10 2.57
15 2.33
A larger sample generally provides greater confidence in the measured population and therefore attracts a smaller statistical penalty. The result still depends strongly on how consistent the measured failures are.

Worked Example: Why a 25kN Failure Does Not Mean a 25kN Working Load

Consider five dedicated test anchors with the following ultimate failure results:
18kN | 21kN | 23kN | 24kN | 25kN

Mean ultimate resistance NRu,m = 22.2kN
Approximate sample standard deviation = 2.77kN
Coefficient of variation v ≈ 12.5%
The highest result is 25kN and the mean is 22.2kN. Neither number is the allowable working resistance.
For a five-test statistical series, K = 3.40. If, purely for illustration, αdist = 1.0 and Ω = 1.0, the statistical term reduces the mean substantially:
NRk1 ≈ 22.2 × (1 − 3.40 × 0.125)

NRk1 ≈ 12.8kN
That is already almost half the highest measured failure load, and it is still not the final allowable resistance. The next stage applies the global safety factor selected for the application, and the resulting allowable resistance must also not exceed the mean load at first movement N1st,m.
This example is illustrative only. The actual calculation must use the applicable BS 8539/CFA regime and project-specific factors selected by the responsible engineer or specifier. It demonstrates why dividing one failure load by an arbitrary number is not a substitute for the prescribed assessment.

Allowable Resistance Is Not Simply Ultimate Load Divided by Two

A common shortcut is to take a site failure load and divide it by two:
“Anchor failed at 30kN.
Divide by two.
Therefore safe load = 15kN.”
That approach bypasses the factors that the site-testing procedure is specifically intended to capture.
It ignores:
  • how many anchors were tested;
  • variation between the results;
  • the lowest-performing test location;
  • load at first movement;
  • reaction-bridge influence;
  • substrate condition;
  • environmental or long-term loading factors;
  • the actual purpose of the test;
  • whether the anchor has an ETA or published design data.
The global factor used in an allowable-resistance calculation is part of a defined engineering route, not a number selected after looking at one test result.
Manufacturer literature can add another layer of confusion because both modern partial-factor values and traditional working-load values may appear in technical documentation. BS 8539 defines recommended resistance Nrec as the maximum working load recommended by the manufacturer. CFA guidance notes that there is no directly equivalent term in Eurocode 2 design.
Under the partial safety factor framework, the principal resistance value used in the structural verification is design resistance NRd. BS 8539 explains that some manufacturers provide a recommended resistance to help users relate Eurocode-derived design resistance back to the traditional global safety-factor approach.
Recommended resistance ≠ design resistance ≠ ultimate test load.
That is why a manufacturer stating a recommended resistance of 12kN and a site anchor failing at 28kN does not automatically justify increasing the design load to 28kN. The two numbers were generated for different purposes and contain different levels of statistical and safety treatment.

Why Manufacturer Software Matters

An isolated catalogue resistance rarely tells the complete story of a real anchorage.
Anchor resistance can be influenced by:
  • concrete strength;
  • cracked or non-cracked concrete assumptions;
  • effective embedment depth;
  • edge distance;
  • anchor spacing;
  • group geometry;
  • member thickness;
  • load direction;
  • combined tension and shear;
  • failure mode;
  • anchor material and durability requirements.
BS 8539 notes that manufacturer design software can be advantageous because it requires the specifier to enter the application parameters and assesses the relevant design conditions systematically. A single number lifted from a simplified table should therefore not be assumed to represent every possible installation of that anchor.

Where Proof Loads Fit Into the Picture

Proof testing belongs to a different branch of the anchor-testing decision tree. A proof load is applied to working anchors principally to validate installation quality. It is intentionally not an ultimate-capacity test.
STRUCTinspect has covered this distinction in detail in The Engineer’s Guide to BS 8539: Proof Testing vs. Allowable Load Verification. A fixing that sustains a 15kN proof load has demonstrated that it can sustain the specified 15kN test demand under the test conditions. The test has not measured where the fixing would ultimately fail.
Likewise, an ultimate test to failure does not retrospectively become a proof test simply because the engineer compares the failure result with the required working action.

Proof Test, Preliminary Test and Ultimate Test Are Different Tools

Test Typical Purpose Working or Dedicated Test Anchor? Failure?
Proof test Validate installation quality. Working anchor. Not intended.
Preliminary test Establish allowable resistance using the simplified route where applicable. Dedicated test anchor. Normally loaded to calculated test load; regime may escalate if that load is not reached.
Ultimate / statistical failure test Generate failure data for determination of site-specific resistance. Dedicated test anchor. Yes.
For a broader introduction to anchor site-test terminology see Anchor Proof Testing vs Pull-Out Testing: What Is the Difference?.

The Test Rig Can Change the Apparent Ultimate Load

Failure testing is not simply a matter of connecting a hydraulic tester and increasing the pressure until something moves. The reaction bridge transfers load back into the surrounding substrate. If the bridge feet are positioned too close to the test anchor, they can restrain the concrete or masonry around the anchor and interfere with the natural failure mechanism.
For ultimate-load tests and preliminary tests in concrete, CFA guidance states that the nearest reaction support should ideally be at least 1.5 × effective embedment depth from the anchor centreline where concrete cone breakout may govern. If this cannot be achieved, the influence of the reaction arrangement has to be considered and reported.
The highest measured failure load is not necessarily the best test result if the testing apparatus artificially strengthened the failure zone.

Edge Distance and Anchor Groups Also Affect Real Capacity

A laboratory or manufacturer resistance value is normally associated with defined geometric conditions. Real installations may place anchors closer to edges or closer to one another. Where anchor spacing is reduced, failure zones can interact. Where an anchor is close to a free edge, the available base material for developing the full resistance can be reduced.
It is therefore unsafe to assume:
One anchor = 20kN
Therefore four anchors = 80kN
Anchor groups have to be assessed as groups under the relevant design method. The resistance of four closely spaced anchors cannot simply be obtained by multiplying the isolated capacity of one anchor by four.

Numbers That Should Not Be Compared Directly

Number A Number B Why Direct Comparison Is Misleading
Single ultimate failure load Allowable resistance One is raw destructive-test data; the other is a reduced working resistance obtained through a defined assessment.
Ultimate failure load Manufacturer recommended resistance Recommended resistance already represents a working resistance rather than physical failure capacity.
Design resistance NRd Characteristic action NEk Partial-factor design normally compares design action NEd with design resistance NRd.
Proof load Ultimate resistance A proof test is deliberately non-destructive and does not locate the failure point.
Short-term site test Long-term bonded-anchor performance A brief site test does not reproduce long-term creep, durability or environmental qualification.
Single-anchor test result Group resistance Spacing, group effects and shared failure mechanisms can control the anchorage.

Five Common Site Scenarios

1. “The Anchor Failed at 30kN, So Its Safe Working Load Is 30kN”

Incorrect. Thirty kilonewtons is the failure load of that individual test anchor at that location. It contains no working safety margin and does not account for statistical scatter across the relevant test population.

2. “Let’s Divide the Failure Load by Two”

A generic factor of two is not a substitute for the applicable CFA/BS 8539 procedure. Sample size, variation, reaction spacing, application factors and first movement may all affect the allowable resistance.

3. Manufacturer Recommended Resistance = 12kN, Site Failure = 28kN

The 28kN result does not automatically supersede the manufacturer's design information. Unless the testing has been undertaken and assessed through the appropriate site-specific resistance regime, it remains a test result rather than a revised design value.

4. Anchor Passes a 15kN Proof Test

The anchor has demonstrated satisfactory performance against the specified 15kN proof-test objective. Its ultimate resistance has not been determined.

5. Tester Reports “Anchor Capacity = 25kN” After One Failure Test

That wording is potentially misleading. A better factual statement is:
“The test anchor reached an ultimate failure load of 25.0kN at the tested location. This individual result does not constitute a design, recommended or allowable resistance for the anchorage.”

Why Load at First Movement Matters

The maximum load reached is not always the only important part of a destructive test. BS 8539 requires the load at first movement to be recorded under the relevant ultimate testing regimes. The mean load at first movement can then limit the allowable resistance even where the ultimate failure loads are substantially higher. This prevents a high collapse load from disguising unacceptable movement at a much lower service load.
An anchor that eventually fails at a high load may still begin moving at a level that governs the usable resistance.

Site Conditions Can Reduce the Usable Resistance

Current CFA guidance includes adjustment factors for conditions that may reduce the resistance inferred from site testing.
Examples identified in BS 8539 include:
  • wet substrate;
  • elevated temperature outside the normal service range;
  • long-term loading;
  • masonry where the position relative to mortar joints cannot be guaranteed.
Where more than one condition applies, the effects may need to be combined. CFA guidance also requires an additional cracked-concrete adjustment where site testing has demonstrated performance only in uncracked concrete but the final application needs cracked-concrete performance.

Do Not Use a Short-Term Pull Test to Prove Long-Term Resin Performance

Bonded anchors provide another example of why ultimate and working loads cannot be interpreted from the gauge alone.
Before testing, the specified resin curing time has to be satisfied for the actual installation temperature and conditions. Testing an under-cured anchor can produce premature bond failure or disturb the installation before the adhesive has developed the required performance.
Conversely, a bonded anchor surviving a short-term tensile test does not by itself verify long-term creep behaviour under sustained loading. That performance depends on the anchor's qualification, ETA, design conditions and manufacturer data.

What an Ultimate-Load Test Report Should Say

The CFA guidance recommends recording enough information for the responsible engineer to understand the result in context.
Report Item Why It Matters
Test objective Separates ultimate, preliminary and proof testing.
Anchor manufacturer, type and size Identifies exactly what was tested.
Base material Anchor performance is substrate-dependent.
Effective embedment Influences both anchorage behaviour and appropriate reaction spacing.
Reaction-frame spacing Allows the engineer to assess whether the bridge may have influenced failure.
Maximum load Records the actual peak test result.
Load at first movement May govern the allowable resistance.
Failure mode Explains whether failure occurred in the anchor, bond or surrounding substrate.
Equipment and calibration Provides traceability for the measured test force.
For broader anchor-testing requirements, STRUCTinspect has also published BS 8539 Anchor Testing: What Contractors Need to Know.

Before Accepting an Anchor Test Result

  • Confirm the objective: proof, preliminary or ultimate/failure test?
  • Confirm the population: how many anchors were tested?
  • Check the substrate: is it representative of the intended installation?
  • Check the product: anchor type, size, embedment and installation method.
  • Check ETA/manufacturer data: does published performance already cover the application?
  • Review the raw results: individual failure loads, not just the average.
  • Review scatter: standard deviation and coefficient of variation where required.
  • Check first movement: was N1st recorded where the test regime requires it?
  • Check reaction spacing: could the bridge have artificially enhanced the measured failure load?
  • Check failure mode: what actually failed?
  • Check geometry: edge distance, spacing, group effects and structural thickness.
  • Check bonded anchors: curing time and substrate temperature.
  • Check safety factors: which design or allowable-resistance framework has been applied?
  • Check responsibility: who is interpreting the results and confirming the final resistance?

Ten Rules Engineers Should Remember

1. A failure load is raw test data, not a working resistance.
2. Do not divide one ultimate test result by an arbitrary safety factor and call it allowable.
3. Characteristic resistance accounts for statistical uncertainty.
4. Design action should be checked against design resistance under the partial-factor method.
5. Recommended resistance and design resistance are not the same quantity.
6. Proof testing verifies installation quality; it does not measure ultimate capacity.
7. Load at first movement can control the allowable resistance.
8. Reaction-frame geometry can influence ultimate test results.
9. Single-anchor results cannot simply be multiplied to obtain group capacity.
10. The tester records the evidence; the responsible engineer/specifier interprets the result for the application.

Evidence-Based Summary

Ultimate resistance is the measured failure load of an anchor during a destructive test.
Mean ultimate resistance is the average failure load from a test series.
Characteristic resistance is statistically reduced and accounts for uncertainty and scatter.
Design resistance applies the relevant material partial safety factors to characteristic resistance.
Recommended resistance is a manufacturer-quoted working resistance and is not directly equivalent to Eurocode design resistance.
Allowable resistance is a working resistance that may be derived from site testing where appropriate published resistance data is unavailable.
A proof load is a non-destructive test demand used for a different purpose and should not be confused with ultimate capacity.
Sample size, result variability, first movement, reaction geometry and site conditions can all reduce the usable resistance derived from raw failure data.
A high number on a hydraulic gauge is therefore only the beginning of the engineering assessment, not the conclusion.

FAQ: Ultimate and Working Loads for Post-Installed Anchors

What is an anchor ultimate load?
It is the load recorded when the tested anchor, its bond or the surrounding base material reaches failure. BS 8539 describes this as ultimate resistance or failure load.
Is ultimate load the same as safe working load?
No. Ultimate load is a failure value. A working resistance incorporates the appropriate statistical treatment and safety factors for the relevant design or site-test regime.
Can I divide an anchor failure load by two to get a working load?
Not as a substitute for the BS 8539/CFA procedure. Site-derived allowable resistance may depend on sample size, scatter, adjustment factors, global safety factors and load at first movement.
What is mean ultimate resistance?
NRu,m is the average ultimate failure load from a series of anchor tests.
What is characteristic resistance?
It is a statistically derived resistance. BS 8539 defines it as a 5% fractile of ultimate resistance, based on the required level of confidence, or as derived by an applicable empirical assessment method.
What is design resistance?
Design resistance is obtained by applying the relevant material partial safety factor to characteristic resistance. It is compared with the factored design action under the partial safety factor design approach.
What is recommended resistance?
It is the maximum working resistance recommended by the anchor manufacturer for the stated anchor and base-material conditions. CFA guidance notes that there is no directly equivalent term in Eurocode 2 design.
What is allowable resistance?
It is a maximum working resistance derived from site testing where the anchor is approved by the manufacturer for the type of substrate but no suitable recommended resistance is available for that particular base material.
Why does the lowest anchor test matter?
A weak individual result may reveal variability in the substrate or installation. Statistical assessment therefore considers scatter across the whole test series rather than relying on the best or average value alone.
Does a larger test sample increase confidence?
Yes. BS 8539 uses different statistical K factors for five, ten and fifteen-test series, reflecting the greater confidence available from larger samples.
Can a high site failure load override the ETA?
Not simply because one or several anchors performed well on site. The appropriate ETA/design basis remains relevant unless a competent engineer uses an applicable BS 8539/EAD site-test regime to establish a justified site-specific resistance.
Can one anchor test be multiplied by the number of anchors in a group?
No. Anchor-group resistance depends on spacing, geometry, load distribution and interacting failure mechanisms and should be assessed using the applicable design method.
Does passing a proof test establish ultimate capacity?
No. A proof test validates performance against the specified proof-load objective without intentionally finding the failure point.

Primary Technical References

The principal UK reference for anchor terminology, design responsibility and site-testing regimes is BS 8539:2012+A1:2021 — Code of practice for the selection and installation of post-installed anchors in concrete and masonry.
Testing procedures and current industry interpretation have been reviewed against the Construction Fixings Association Procedure for Site Testing Construction Fixings 2024, including its terminology, allowable-resistance regimes, proof testing, reaction-frame requirements and reporting guidance.
For anchor design in concrete, the wider design framework includes BS EN 1992-4:2018 — Eurocode 2: Design of fastenings for use in concrete, together with the applicable ETA/EAD and manufacturer documentation for the anchor system.

Source Context and Technical Note

This article is a STRUCTinspect technical explainer intended to help engineers, contractors, project managers and testing teams distinguish between the raw loads measured during anchor testing and the resistance values used in structural assessment. Ultimate resistance, characteristic resistance, design resistance, recommended resistance, allowable resistance, proof load and characteristic action each describe different engineering quantities. They should not be substituted for one another simply because all are expressed in kilonewtons.
The correct value for any project depends on the anchor system, applicable ETA/EAD, base material, geometry, failure mode, design method, site conditions and the purpose of any testing undertaken. This article does not provide anchor design, structural design, legal or contractual advice. Test regimes, load levels, resistance calculations, acceptance criteria and final anchor suitability should be confirmed by the competent engineer or specifier responsible for the anchorage.