Settlement computation, in my opinion, is fraught with many uncertainties.
First there is the question of what loads to use. You want the "real" loads. Ideally, for clays, you would want the dead loads and the permanent "live" loads - i.e., furniture, storage of materials and the like - live loads that will be more or less permanent. Then you will also want the transient live loads - wind loads, unbalanced water pressures (on mass concrete dams) due to reservoir level flucutations - you will also want to have an idea of any seismic loading. For sands, you want the similar loadings but realize that for transient dead loads, they can induce additional settlements - the rebound after removal may not always be the same as the loading (although, in practice, it would probably be negligible). This is not an issue as much is clays as the transient load will be taken immediately by porewater pressure build-up and then subsequent porewater pressure drop on the transient load removal - of course, there are times that the transient loadings might exceed the equilibrium porewater pressure and the transient loading can add a permanent porewater pressure load that will add to the consolidation.
Secondly, you need to determine what methods of computation you will use. For sands, some, like OldestGuy, likes Hough's method (which I believe you can also find in the WSDOT geotechnical manual). Many others, though like Burland and Burbidge (see M.J. Tomlinson's Fdn Dsng and Construction book). Others like Schmertman's Method. Many years ago, Ground Engineering had a great article that compared about 15 different methods of settlement in sands to actual tests. There conclusion was to pick three methods and then take the "average" of the three for your estimate. The problem with sands, too, is that many methods use "elastic" theory and hence the E value is a necessity. Being 20% off on the E value chosen will put you 20% off on the settlement computation. Of course, one could always look into using the theory of particulate half space (see Milton Harr's book on Mechanics of Particulate Media (1977) - he also wrote a book on Mechanics of Soils as Particulate Media Under Moving Loads (1989)) - have really never seen this referenced any any other text book that are typically used in soil mechanics. Okay - so the bottom line is you have uncertainty in E - hence uncertainty in the estimated settlements.
For clays - you will have the problem of what the overconsolidation ratio is. Using traditional oedometer results, you might think you have a normally consolidated clay - but it is much more likely that there will be some quasi-preconsolidation that you may or may not know from your oedometer tests - or estimate - like "assume" (as bad as some think) that the Pc' (preconsoldiation pressure) is 10% higher than the existing Overburden pressure. You will also have issues with Cc/1+eo (in my view, always plot strain vs log p' rather than e - as the slopes will already, then, have the 1+eo built in). Some posit that you plot log strain vs log p' to get the Pc' value - it is subjective. Personal note is that years ago Casagrande came up to the Geocon office in Toronto (Fred Matich studied under Terzaghi and Casagrandes) - and some engineer proudly showed him about 20 plots using his graphical method of determining Pc'. He looked at the plots and said - "No - this should be about here and that about there . . ." Dejectedly the engineer said "but I did it according to your method." - and he replied, well, I had to have some way to get my students to come close to the correct value . . . To add to this - do you take into account Bjerrum's pore pressure settlement coefficients? . . . and Fox's depth correction factor. There is also estimating settlement using stress paths. Som and Das (Theory and Practice of Foundation Design) compared the "traditional" method and the stress path method for the same "footing model" - they got 48.2 mm (using Skempton/Bjerrum method but only 30.8 mm using the stress path method. So, there is the rub - they differ by approximately half of the computed settlements. (by the way I do not believe in decimals of a mm !!)
Now, you also have the effects of creep - in both clays and sands - see Schmertman.
Sorry to ramble on - it is a Monday night - at the office as we have aircon and internet . . . but, the basic gist is that settlement computations provide what? - estimations!
I remember some learned types in the past say that if you are within 30% of the actual value - you've had a good day - owe yourself a beer! Steel and concrete have Coefficients of Variation of 5% say - soils are in the range of 25 to 50% depending on the parameter. So many times, you might just see soil reports simply state that settlements are expected to be less than 25 mm, say, or 40 mm . . . there is no hard number to put on it.
Safety Factor on settlements? NO. But certainly computations and judgment to come to a "close" range? Yes - but always with a needed amount of wiggle room. The estimations depend on quality of soils data and reliability of the methods of computation.
