In general, I think spreadsheets get "out of control" because they offer a model by which non-programmers can program. This lets spreadsheets "get out of contr

In general, I think spreadsheets get "out of control" because they offer a model by which non-programmers can program. This lets spreadsheets "get out of control" in the sense that individual files become quite huge (even if they never get extended to inherently unreasonable tasks). Digging deeper, why are huge files a problem? I believe it's because spreadsheets hide the manipulations they encode. This makes them hard to verify, hard to test, and hard to re-use. The larger the file, the more difficult it is to inspect and verify. In short, spreadsheets present an opaque programming model, whose problems compound as the file size increases. My personal solution has been to use command-oriented data analysis tools, mainly "R". Programs like these combat huge file sizes, and opacity, by storing vector-oriented commands, rather than vectors of cell-oriented commands. To give a practical example, suppose you have a spreadsheet listing 100 employees, their hours worked, and their hourly rate. You want to multiply out the cost per employee, and add all those values up. With a spreadsheet, you have 100 cells, each containing a product of two numbers, then another cell containing the sum. To verify the calculations, you would have to click through 101 cells, inspecting their contents. With a tool like R, the identical operation would be something like: employeeCosts <- employeeHours * employeeHourlyRate totalCost <- sum(employeeCosts) Furthermore, with a spreadsheet you may not really care about that column of {employeeCosts}. Still, you have to calculate them, and you have to look at them as you attempt to make sense of your calculations. So you can either: (1) leave the column along with all the other columns, making it harder to focus on the results you care about; (2) put the column on a separate sheet, making it harder to verify; or (3) do some visual trick like minimizing column widths, making it harder for somebody else to explore and understand your code. A command-line tool, on the other hand, gives you lots of ways to avoid having to look at unwanted intermediate values: (1) rm(exployeeCosts) (2) totalCost <- sum( employeeHours * employeeHourlyRate ) (3) findTotalCost <- function(timeSpent, rate) { return(sum(timeSpent*rate)) } The last solution leads to re-usability, which is where spreadsheets really start to fall down. Every spreadsheet I've ever been sent that links to another spreadsheet has been a usability disaster. Once you make the jump to a real data analysis tool, the benefits begin to multiply. For exammple: (1) It won't take 5 minutes to upload your files to an e-mail to send somebody. (2) You can version-control your work. (3) Your graphics will almost certainly improve in quality. (4) You will find it easier to maintain "Single Point Of Truth"-- for example, the table of employee rates can "live" in a single file, which individual analysis scripts load as needed, rather than getting duplicated in every spreadsheet that needs the numbers.