The dilatometer test is the best tool that geotechnical engineers can use to predict settlement of structures or embankments.  It is a calibrated static deformation test that is performed at close depth intervals (10 or 20 cm/4 to 8 inches).  The predicted modulus value is at a strain level of approximately 0.5%, which is a strain value that corresponds to what buildings and embankments impose on the soil.  And thus the DMT modulus is referred to as an "operative" modulus.  Schmertmann (1986) presented a settlement design method using DMT data and showed with case histories its superb accuracy.

I am attaching two technical papers, “DMT-predicted vs. observed settlements a review of the available experience” and “DMT testing for redesign using shallow foundations” that show the value of the DMT for settlement prediction.   

Schmertmann's (1970 and 1978) methods for predicting settlement using CPT data was only intended for unaged sands.  It has often been misapplied for other soil types and stress histories.  Too often (once may be too often), engineers have converted SPT N-values to CPT qc-values using approximate correlations and used Schmertmann's CPT unaged sand method to predict settlement.  Schmertmann abandoned CPT settlement predictions in 1981 in favor of the superior DMT method and so should all engineers. 

The SPT and CPT are penetration tests--not deformation tests--and strain the soil to failure.  These strain levels are many times larger than what a building or embankment will impose on the soil.  Thus settlement predictions based on these methods are highly inaccurate and should not be used unless there are extensive site specific correlations.

Duncan (2000) showed that the Burland and Burbridge (1985) SPT method for predicting settlement in sands using N₆₀-values had a coefficient of variation (standard deviation/average) of 0.67.  The value is so high that in order to safely (95% success) predict that allowable settlement will not exceed 1.0 inch the average settlement value is 0.3 inches.  His analysis only considered the inaccuracy of the SPT method for predicting settlement.  In my discussion of Duncan 's paper, I showed that when one considers error from SPT repeatability and subsurface variability the average settlement value is even lower.

Why don't more buildings settle if so many engineers in the U.S. and elsewhere use the SPT to predict settlements? --Overly conservative design.  Although Skempton developed the N₆₀ correction method in 1986, it has been a rare gem to find N₆₀ values shown on soil boring logs.  The SPT method has significantly changed over the last 50 to 60 years.  Well-maintained automatic SPT hammers deliver about 55% more energy than safety hammers.  Today's split spoons have an inner diameter that is larger than the tip diameter to accommodate liners, but liners are rarely used.  Not making these corrections can reduce the N-value to 50% of an N₆₀-value.  And thus we can be conservative by 200%. 

When engineers recommend deep foundations without proving that they are necessary, they have not provided good service to the owner.  DMT data can be used to prove whether deep foundations are necessary.