Why a Greasy Floor Tests Better on the Seventh Swing Than the First
By Walkway Management South Florida
We tested a restaurant dining floor in the walkway that runs past the kitchen door. The first pendulum swing came back at 41. The seventh swing, on the same five inches of floor about ninety seconds later, came back at 59.
Nothing about the floor changed in that ninety seconds. No one cleaned it, no one treated it, the temperature didn’t move. The only thing that happened in between was that we tested it six more times.
That gap is the whole subject of this article. On a floor with grease on it, the pendulum slider lifts the contaminant off its own contact path a little more with every pass. The surface under the slider gets cleaner as the test proceeds, so the readings climb. If you report the averaged number the way you would on an ordinary floor, you are reporting the friction of a strip of floor that your own test just cleaned, and no customer ever walked on it in that condition.
This is why we run the Pendulum Slip Tester and the BOT-3000E on the same job instead of picking one. Below is what that looked like on a real floor.
The floor and the numbers
The surface was a resinous cementitious topping over concrete, a matte pigmented finish with fine texture and light exposed aggregate. Good material, in fair to good condition, no cracking or delamination. It was tested as-found, early in the morning, before the restaurant opened for the day.
We ran the pendulum with the Slider 96 (4S) hard rubber, which is the slider used for shod pedestrians. Because a pendulum reading depends on which way you point the instrument, we tested four orientations at the station, seven swings each, under dry and then wet conditions. We kept every individual swing instead of only the average.
Every orientation did the same thing, dry and wet.
| Orientation | Condition | Swing 1 (as-found) | Swings 2 through 7 | Rise | Average of last five |
|---|---|---|---|---|---|
| 1 | Dry | 41 | 44, 50, 54, 55, 58, 59 | +18 | 55 |
| 1 | Wet | 32 | 35, 37, 38, 41, 42, 42 | +10 | 40 |
| 2 | Dry | 51 | 53, 57, 58, 59, 59, 60 | +9 | 58 |
| 2 | Wet | 31 | 36, 38, 40, 40, 40, 42 | +11 | 40 |
| 3 | Dry | 47 | 53, 54, 55, 57, 60, 61 | +14 | 57 |
| 3 | Wet | 35 | 38, 40, 43, 44, 45, 44 | +9 | 43 |
| 4 | Dry | 44 | 45, 49, 49, 54, 58, 59 | +15 | 53 |
| 4 | Wet | 31 | 33, 34, 36, 38, 38, 39 | +8 | 37 |
Eight test series. Eight rising staircases. Not one of them settled down.
The part that decides the answer
Look at the wet rows, because wet is the condition that matters in a restaurant dining room.
The as-found first-contact values were 32, 31, 35 and 31. Under the slip-potential model published by the UK Health and Safety Executive and the UK Slip Resistance Group, which reads 0 to 24 as high slip potential, 25 to 35 as moderate, and 36 and above as low, all four of those readings sit in the moderate band.
The averages of the last five swings were 40, 40, 43 and 37. Every one of those sits in the low band.
So the same instrument, on the same four spots, in the same ten minutes, puts the floor in two different classifications depending on which number you write down. Average the swings and the floor looks fine. Report what the slider found on first contact and it doesn’t.
The averaged number isn’t wrong arithmetic. It is an honest average of a surface that was getting cleaner while we measured it, which is not the surface anyone walks on.
The objection we expect, and the answer
Anyone who knows the pendulum will push back here, and they should: readings are supposed to rise over the first couple of swings. That is normal. The rubber slider needs to bed in against the surface, the contact strip stabilizes, and the first swing or two run low for that reason alone. It is exactly why the procedure records seven swings, throws away the first two, and averages the last five.
So how do you tell slider bed-in from the test scrubbing a contaminant off the floor?
You watch for the plateau. On a sound, clean surface the readings come up over the first two or three swings and then flatten out, because there is nothing left to change. The whole point of discarding the first two is that swings three through seven should be measuring a stable surface.
That is not what these did. They rose monotonically all the way through swing seven and never converged. Orientation 1 dry was still climbing at the end, 58 then 59, after already gaining 18 points. A slider that has bedded in does not keep gaining ground for six consecutive swings.
Two other things point the same direction. The rise showed up in all four orientations, so it isn’t a quirk of one direction of travel. And it showed up dry as well as wet, which matters, because water gets replenished between swings and grease does not. A contaminant that the slider carries away and that nothing puts back is exactly the kind that produces a one-way staircase like this.
None of this is only our experience. There is published test data showing that pendulum readings get unstable specifically when a contaminant is present, and we come back to it further down. It is also worth noting that the relationship between a pendulum value and an actual coefficient of friction is device-specific and nonlinear, which is a further reason not to treat one pendulum average as a direct measure of available friction (Cui et al., 2022).
What the BOT-3000E added
If the pendulum path is being cleaned by the test, we have a problem: the pendulum can tell us a contaminant is present, but it can no longer tell us how bad the floor actually is. The evidence of the contamination is also the thing destroying the measurement.
That is where the second instrument earns its place. We ran full ANSI A326.3 DCOF tests with the BOT-3000E, three samples of four directional measurements at each station, with the lowest value governing.
The dry results are the ones worth sitting with:
- Dining area, in the walkway near the kitchen: 0.33 DCOF, dry
- Entry foyer, the same flooring, away from kitchen traffic: 0.71 DCOF, dry
Same floor, same building, same instrument, same morning. One area produced better than double the friction of the other with no water involved at all.
A sound floor of this type is supposed to be comfortably slip-resistant when it is dry. A dry reading of 0.33 means the grease was not sitting in a puddle on top of the surface, it was down in the texture, filling the micro-profile that gives the floor its grip. The floor was compromised before a single drop of liquid arrived. The wet values in that same dining walkway, 0.39 and 0.42, then sat at or below the 0.42 that A326.3 recommends as a minimum for interior areas expected to get wet, and far below the 0.55 recommended where oils and greases are present, which is the honest category for a dining walkway served straight out of a kitchen.
The BOT is not magic. Drag it over the same eight inches enough times and it would clean the path too. What makes it the right partner here is the protocol and the comparison it enables. A326.3 moves to fresh surface and takes the lowest of three samples rather than averaging a sequence, so a rising series can’t quietly become the reported result. And because the number is an absolute coefficient of friction anchored to a reference surface, you can hold the suspect area up against a control area on the same floor and read the difference straight off. The 0.33 against 0.71 is that difference. The pendulum, on a path it had already swept, could not have produced it.
There was a third sign, and it cost nothing to observe. After we toweled the test area dry during wet testing, it stayed visibly damp for a quarter of an hour. A clean surface of this type sheds water and dries quickly. One that holds moisture that long is fouled, and it also means any spill on that floor stays a wet floor for much longer than anyone would expect.
The published data says the same thing
We are not the first people to notice that instruments behave differently once a floor is dirty. It has been measured directly.
A study in Safety and Health at Work ran three in-situ slip meters, the pendulum, the English XL and the BOT-3000, across the same floors and the same contaminants, and looked at how steady each instrument’s repeated readings were (Kim, 2012). The test set was built around restaurants on purpose: ceramic, vinyl, and asphalt tile with and without wax, measured dry and then under water, detergent solution, soybean oil and engine oil.
The number that matters is the coefficient of variation, which is just the spread of repeated measurements divided by their average. Smaller means steadier.
The pendulum did well on dry floors. Its coefficient of variation there was 0.055, which is tight and repeatable. Then a contaminant went down and it rose above 0.2. The paper’s own wording is that pendulum repeatability “reduced sharply” once the specimens were covered with liquid contaminants. Good dry, unsteady dirty, which is the same behavior we watched in the swing series.
The BOT-3000 did not come apart the same way. Its static measurements held a coefficient of variation under 0.1, the steadiest of the three instruments in the study, and its dynamic measurements ran between 0.1 and 0.2.
Two things about that study are worth stating plainly, because it gets cited loosely in both directions and anyone can go read it.
The figure below 0.1 belongs to the BOT’s static measurement, not its dynamic one. The dynamic figure was 0.1 to 0.2, and the English XL stayed under 0.2 across every condition.
More importantly, none of those dynamic numbers describe DCOF as it is measured today under ANSI A326.3. That study used a Neolite slider, while A326.3 requires SBR. Its contaminants were plain water, a hypochlorite detergent, and two oils, where A326.3 specifies a 0.05% sodium lauryl sulfate solution. The device was the older BOT-3000, and the work was published in 2012, before A326.3 existed as a standard in its own right. Different rubber, different wetting agent, different machine, written before the method. So we do not offer it as a repeatability figure for A326.3, and it should not be used as one against A326.3 either.
What the study does establish is the thing that matters here, and it establishes it well: put a contaminant on the floor, and the pendulum is the instrument that loses the most stability while the drag sled holds together.
The same study, with the same limits on how far its numbers travel, found one more thing worth repeating, because it asks a different question. Repeatability only tells you whether an instrument gives the same answer twice. It says nothing about whether the answer is tracking the contamination or just holding steady. So the study also compared each instrument’s readings against the measured viscosity of the contaminant on the floor. All of them correlated. The drag sled’s dynamic readings correlated most closely of anything tested, at r = 0.987. The thicker the contaminant, the further the reading moved, in step with it.
That is a statement about direction of response rather than about precision, so the protocol differences above don’t undo it. And it is the property you actually want from an instrument you intend to point at a greasy floor: not merely consistent, but sensitive to the thing you are there to find.
The mechanics behind the difference are not mysterious. The pendulum’s slider crosses its contact path at roughly 3 m/s. The BOT’s sensor travels at about 0.2 m/s. A fast sweep across the same fixed strip of floor, over and over, is an efficient way to displace a viscous contaminant, which is precisely what a restaurant floor has on it.
So the published work measured the symptom, which is scatter, and our swing series shows the cause, which is the test cleaning its own path. They are two views of one problem.
Why we run both, every time
The two instruments were not answering the same question, and that is the point.
The pendulum told us what kind of problem we had. No single averaged value would have said “grease.” The shape of the swing series said it, and only because we kept every swing instead of writing down one number per orientation. The pendulum remains our reference method for slip resistance, and this job is a good argument for it rather than against it. What failed was not the instrument. What failed would have been reducing it to one averaged figure on a floor where the test changes the thing being tested.
The BOT-3000E told us how bad it was. It gave a defensible magnitude, a control comparison on the same flooring, and a dry-condition finding that the pendulum could not have supplied once its own path was clean.
Neither instrument on its own gets you there. Pendulum alone on this floor, reported conventionally, produces averages that put a grease-fouled restaurant walkway in the low slip potential band. DCOF alone gives you 0.33 versus 0.71 and a strong argument, but not the mechanism, and no explanation of why the dry number is so low. Run both and the picture closes: a rising swing series that identifies contaminant removal, and a dry DCOF gap that measures what the contaminant cost.
We also treat one convention as settled because of jobs like this one. When the swing series climbs and never plateaus, the first-contact swing is the as-found value, because it is the only swing taken on the floor as we found it. Every swing after it was taken on a floor our own instrument had started cleaning. We still report the average, clearly labeled, so nothing is hidden. We just don’t pretend it describes the floor anyone walked on.
If you are reading someone else’s report
Whether you manage the building or you are reviewing a test report someone else produced, a few questions separate a report you can lean on from one you can’t.
Ask whether the individual swing values were kept, or only the averages. If only the averages survive, a rising series is invisible and unrecoverable, and you cannot tell a stable floor from a floor that was being cleaned by the test.
Ask whether the floor was tested dry as well as wet. Dry testing is where impregnated contamination shows itself. A floor that underperforms dry is telling you something a wet test alone will never isolate.
Ask whether there was a control location. A number by itself invites an argument about thresholds. The same number next to a reading from the same flooring in a cleaner part of the building is much harder to talk past.
And ask whether more than one method was used. Two instruments working on different principles either agree, which is worth something, or they disagree in a way that tells you why.
One last caution, because it cuts both ways. A DCOF value is a comparative measurement, and A326.3 says plainly that a single value does not by itself predict whether any particular person will slip. Pendulum values and DCOF values are separate measurements on separate scales and cannot be converted into one another. No one number, from either instrument, makes a floor safe or unsafe on its own. What builds a conclusion you can defend is convergence: several independent measurements, taken different ways, pointing at the same answer.
On this floor they did.
This article is informational and describes slip-resistance measurement practice. It is not legal advice. The test data described here is from a de-identified field evaluation and is presented to illustrate measurement behavior, not to characterize any specific property.
