We are looking for newsagents willing to work in a development project to verify this section of our system.
Now our estimation program can calculate publications sales values in subagents to within 22% accuracy for an individual item. But I do believe as there are many items we can overall get a required return rate of 7% with better sales figures then now. The arguement is if I said one head, and you threw a coin in the air once, I would be 50% wrong. Say I said 500 heads and you threw 1,000 times a coin in the air, I would be pretty close to the result. The more times you are going to throw the coin, the more accurate I can be.
The reason for this belief is that years ago, I discovered that most newsagents would be better off inflating the subagents orders by far more then they do now. My computer results suggested that often newsagents were losing many sales by dramatically and unintentionally undersupplying. Interestingly HWT did a detailed survey in the industry with the public and came to a similar result. Many subagents were not getting enough supply.
Until now the problem with these tests we did because if a newsagent thinks that 15 will sell in a subagent. No matter what the computer said 20, 12,17 10 etc, the newsagent will deliver 15. So the computer is not given a chance.
So recently I took it to HWT. Please let us do a proper test. We know how the newsagent is doing now. Let the computer do a few weeks and let see how it does. Recently HWT have expressed some interest in this project. So we are looking to get a project going that hopefully will prove it way or another.
If anyone is interested in working on such a project DOS or posbrowser, please let me know.