You should probably fix that so you're not storing numeric values in string columns. We have numeric types precisely so we can avoid this kind of problem.

Anyway, float is a little more forgiving, and unlike decimal , can accept strings in scientific notation. But note that it could introduce accuracy issues:

SELECT TRY_CONVERT(decimal(18,9), TRY_CONVERT(float, MyCol))...

Example . Also see this answer .

The 6E-06 and -6E-06 in the question example are already binary literals. You'll get same results in your test if you run it like this:

SELECT TRY_CAST('6E-06' AS DECIMAL(18, 9)) AS OutCol;
SELECT TRY_CAST('-6E-06' AS DECIMAL(18, 9)) AS OutCol;

Unfortunately, SQL Server doesn't have built-in support for directly converting to numeric or decimal types from strings formatted in scientific notation. Therefore your best option is probably pre-processing this data. Failing that, you're left with some rather complex code. , but you might be able to simplify things by converting to a float or real first, depending on the possible range of the data.