What do you consider a sample to mean during an EICR

Well, you can test 100% of the accessories on 20% of the circuits, or 20% of the accessories on 100% of the circuits, but you mustn't sample the accessories on a sample of circuits.

Ideally, the sample should be selected at random.

With 6 - 10 circuits in a domestic, I think that all should be I&T'd. The contract may have been for 20%, but IMHO, that doesn't meet the legislative requirement.

By contrast, if you have an office block with dozens of circuits, by all means sample them. Even better is to test say 20% each year on a 5 year cycle, but a record must be kept of which have been done when.

If you do sample, and find a duff circuit, what then? If it was 1 out of 5, the conclusion must be that the whole installation is unsound. 

Yes Chris I can see the merits of the 20% per year ethos - providing you start at a straight edge of a full EICR/EIC prior to that (either one you have done yourself preferably or one that you have faith in (rare as a unicorn!!!)) then a rolling 20% as you say does have some meaning and could have merits. 

In domestic installations I test everything. I do not "sample." M.O.T. inspectors test everything as well.

Z.

We had a third 2391 test rig installed last year. Sparky was keen to please and left a very tidy job indeed. Four subsequent class assessments have now brought it to a state where Sparky would need a re-visit to address what he might see as a bit of a shambles! 
I would be more inclined to minimise dismantling and carry out fundamental tests without disconnecting where possible. A review of my own EICRs stretching back years will reinforce the point that the emphasis should be squarely on inspection with testing only a supplementary consideration. The vast majority of serious issues were uncovered by inspection. Those serious issues revealed by testing were largely by means where disconnecting was not required although, admittedly, there were quite a few open RFCs over the years. 
Trying to impose a globally accepted sample set is perhaps a handy tool for the bean counters but every job should be subject to a detailed risk assessment before arriving at a determination on sample size. That is easy for me to say but I appreciate that it would be horrendously difficult to implement in reality. 
My approach has always been to attempt to arrive at a reasonable assessment of the condition of the electrical installation by minimal disruption and as such I set my own sample sizes. The key thing is that I believe I can justify them. 

We can again discuss this forever, but as Lyle says the important bit for assessing general condition is inspection. I see no reason to do sampling at all in a domestic, but this is where a proviso comes, if any significant problems are found I do 100% inspection of every accessible connection. Usually, the CU has numerous connections I consider to be loose when checking all the tightness, and this says one thing, the installer was not careful to tighten properly so every screw is suspect. Almost always they are all looser than the current torques suggested by manufacturers, and the reason is easy to see. The screws on modern accessories are larger than older types, and many of the screws were tightened with a 3mm screwdriver, which has a small handle and cannot be turned very hard, often the ubiquitous "neon". Modern tools have bigger handles and so the screws, which may have suffered some creep or always been loose, can be tightened more. I remove accessory fixing screws with an electric screwdriver which makes the job quite quick so looking in 20 or 30 sockets takes an hour or so unless they are very old patress types that are always loaded with very short wire ends! The light switches and ceiling roses are also done and then while dismantled can have wander lead tests and inspection for lack of CPCs.

At this point, the general condition is fairly obvious, and a realistic set of tests may be applied easily, but there is little point in doing everything as if a new installation. I like to loop test every socket and light (My old 2391 tutor, Tom, loved this), and usually, an insulation value for the whole installation is adequate (only L&N to Earth). Any problems can then be noted, and remember this is not a fault-finding job, it is I&T to decide if continued use is satisfactory.

I will now have a lot of comments of the "what if" kind but think a bit, you have tested the Earthing, fixed any loose connections, seen any install problems that can be accessed, noted any broken accessories, and seen any likely high-temperature bits. You have seen the quality of the installer and got the condition. You may want full ring tests if you have seen modifications or new work, and you will have done RCD tests if any present when you did the loops (perhaps by accident!).

I consider excessive dismantling a serious problem as it is likely to induce faults, and these will probably not show unless you do all the testing again. I find very few installs that are fully satisfactory, I do use a 3 neon test plug to check polarity and neutral at every socket (surprising how many are wired backward) and an MFT with wander lead. My worst hate is undoing painted ceiling roses! I rarely even think about the design in a domestic, they are too small to have Zs troubles, and the loop tester gives a warning anyway.

Thanks to all those who sent poor EICRs to me, the matter is now under serious consideration by the IET, all data gained is anonymous.

I am a bit disappointed that the test rigs are a bit worn Lyle, careful reassembly is very necessary!

David Stone said:

I remove accessory fixing screws with an electric screwdriver which makes the job quite quick so looking in 20 or 30 sockets takes an hour or so

That's fine until you come across the screws with chambled heads; rusted in; or stripped threads, either of the screws themselves or the back boxes.

For sampling to be valid, it requires an adequate "population" to sample.

The minimum population size for a sample is 100. That's "things in general" - so it depends on what aspects of "inspection" or "testing" you are talking about - and what the risks of missing something are.

So, in a caravan for example, there's little room for missing out any inspections or tests, as a sample will be meaningless.

I agree with all of that. Are you me?

Z.

Not Exactly you zoom, probably old school though!Chris the stuck ones get the full impact driver treatment (either loose or broken) then a tap or a new box. Adequate violence is good. If it won't come apart it fails!