You can use the same argument for banks.

The number of transactions in voting and at banks are pretty wildly different.

It doesn't have to be. If online voting ever becomes widespread, there is a reasonable chance there will be more votes and referendums.

There would be a absurd cognitive overhead for people to have to study and prepare an opinion for as many votes as they have financial transactions. Every other productive activity in the US would grind to a halt.

Why would polls like :

Do you want a bridge built here? Yes, No, Don't Care.

be a cognitive overhead?

The California voting guide for June 2018 is 96 pages long:

http://voterguide.sos.ca.gov/pdf/complete-vig.pdf

The info associated with each ballot measure is considerably more extensive than you seem to realize.

That's because there are so few votes. Imagine having one single exam at the end of 4 years of education.

No, this is the 2018 guide for the primary election. That is less than 2 years after the 2016 general presidential election. In 4 more months, there will be another 2018 guide for the general election. That means there will be 6 elections in 4 years, not counting any special elections:

http://vigarchive.sos.ca.gov/

You don't even know how many votes we have even though they are so purportedly few. Imagine what little the average voter knows.

No you can’t, your analogy is absurd. Canada uses a fully paper system for federal elections and the costs of manually counting are negligible, and still completed the evening of. The same can be said of American elections before the allure of electronic systems took hold.