7 ms·

decrease in monthly uniques (-2.69%) against continued strong growth (10.70%)? The overall market is growing because there are two (2M visitors/month combined,

by dmv 17y ago

decrease in monthly uniques (-2.69%) against continued strong growth (10.70%)? The overall market is growing because there are two (2M visitors/month combined, vs 900k a year ago). But the mix has steadily changed. If that's not market share loss, what is it?

andrewljohnson 17y ago
That's correlation at best, not causation. And the significance of the difference is questionable at that. I presume there are other programmer websites in the world than stackoverflow and expertsexchange. The miniscule decrease you note can hardly be attributed to stackoverflow, just because stack overflow is the other company on the graph. It needn't even be competitive forces at play. It could be that expertsexchange got slower and Google fed it less traffic for a while, which is something that might be permanent or transient. Attributing this to anything based on this graph will be a fruitless labor.
- j_baker 17y ago
  I agree, but there's always wishful thinking!
- dmv 17y ago
  I argue correlation/causation all the time -- but this is really a semantics argument on both our parts. The statement was "There's nothing in this graph that suggests market share is being taken away." I suppose the emphasis could be placed either on taken or away. No graph can show action ('taken') so as to demonstrate causation. That's why so many arguments are susceptible to "correlation/causation" trouble. As you suggest, there are any number of causes for the data given. To seek proof of causation from a single graph is a flawed venture. But if we look at whether market share was reduced ('away'), an observation may be made from a single graph. Not to say 'why' or 'how much', but the observation is still valid. If you sell 90 cars last year, and 110 cars this year -- and I sell 5 comparable cars last year and 50 comparable cars this year, I have "taken market share" from you in a generally accepted sense. Within the market space of two, I have more share of our market this year than the previous, even if your sales did not decrease. It doesn't matter how I did it (cause). It may not be true next year (correlation). It may not be universally true -- if there were 500 of this kind of car sold last year, and only 400 cars this year, your share of the total market increased. My conclusion before, and again: "the mix has steadily changed. If that's not market share loss, what is it?"

by dmv 17y ago

andrewljohnson 17y ago

That's correlation at best, not causation. And the significance of the difference is questionable at that. I presume there are other programmer websites in the world than stackoverflow and expertsexchange. The miniscule decrease you note can hardly be attributed to stackoverflow, just because stack overflow is the other company on the graph. It needn't even be competitive forces at play. It could be that expertsexchange got slower and Google fed it less traffic for a while, which is something that might be permanent or transient. Attributing this to anything based on this graph will be a fruitless labor.

j_baker 17y ago
I agree, but there's always wishful thinking!
dmv 17y ago
I argue correlation/causation all the time -- but this is really a semantics argument on both our parts. The statement was "There's nothing in this graph that suggests market share is being taken away." I suppose the emphasis could be placed either on taken or away. No graph can show action ('taken') so as to demonstrate causation. That's why so many arguments are susceptible to "correlation/causation" trouble. As you suggest, there are any number of causes for the data given. To seek proof of causation from a single graph is a flawed venture. But if we look at whether market share was reduced ('away'), an observation may be made from a single graph. Not to say 'why' or 'how much', but the observation is still valid. If you sell 90 cars last year, and 110 cars this year -- and I sell 5 comparable cars last year and 50 comparable cars this year, I have "taken market share" from you in a generally accepted sense. Within the market space of two, I have more share of our market this year than the previous, even if your sales did not decrease. It doesn't matter how I did it (cause). It may not be true next year (correlation). It may not be universally true -- if there were 500 of this kind of car sold last year, and only 400 cars this year, your share of the total market increased. My conclusion before, and again: "the mix has steadily changed. If that's not market share loss, what is it?"