Morgan Stanley’s Mary Meeker and team rechecked their YouTube advertising calculations and produced a new set of estimates this morning. As expected, if Mary had stuck with the assumptions used yesterday, the correct maths would have produced a ludicrously small estimate. So she changed the assumptions to boost the estimates.
Hats off to Mary and her team for acknowledging (and repairing) the error, but the episode does reveal a hazard of much financial forecasting. It seems clear that the “1% of streams have ads” used in yesterday’s calculation was chosen not because it was the most reasonable assumption (why would YouTube only put ads on 1% of streams? Why not, say, 4%? Or 10% Or 0.5%?) but because higher assumptions produced absurdly high revenue estimates. That the problem was a failure to divide by 1,000 escaped notice, so the estimates were made reasonable-seeming by adjusting the assumptions.
And now that the maths has been fixed–and the calculation has produced an immaterial revenue estimate–the assumptions have been changed. This is what’s known in the trade as “backing into the numbers.” It’s common, and, used appropriately, it can be helpful: The new estimates are probably still too high, but they seem far more reasonable than those with yesterday’s assumptions would have. But the episode clearly reveals the risk of blindly accepting what appear to be carefully developed assumptions, not to mention the estimates derived from them.
Morgan Stanley’s YouTube Estimates:
Yesterday: $720 million. ($720 thousand using same assumptions and correct maths)
Today: $75 – $189 million (using new assumptions).
The backstory: Mary Meeker’s YouTube maths
Henry’s email: [email protected]m