Analytical Procedures as Substantive Test

The desired level of assurance is generally 95%. Since you question is abstract in nature, I will best try to answer.

Based on knowledge of the client obtained from historical understanding as well as the understanding the client and various industry and market data, one can use analytical procedures to determine the expected amount earned or expensed by a client for a particular activity (e.g. salaries and wages)

Lets assume that the client has had no change in the number of employees and the salaries increased on average 7% over the previous year, then if salaries was $100 million last year, the expectation is for $107 million this year. In this case the expectations would be sufficiently precise to enable identification of a material misstatement, since the level of salaries is controlled.

Lets say you tried to use this method to assess the cost of sales, which consists of utilisation of 10 different raw materials from various sources, both local and international. During the year the following impacted the cost of raw materials:
1. exchange rate movement
2. government revenue measures
3. cost of fuel
4. cost of electricity
5. transportation costs
6. price increases by suppliers
It would be difficult to ascertain the impact on the COS overall, as there are too many variables. You might assume a 10% overall increase, however when you look at the actual figures, it shows an increase of 22%. Worse yet, it could show an increase of 11% which is in line with your expectation, however included in the actual figures is a material misstatement which your analytical procedures may not detect as the figures fall within your expectation, leading you to conclude that there is no material missstatement, when there actually is (a detection error).

The more variables affecting a balance, the harder it is to derive and expectation which is sufficiently precise to identify material misstatements.