Why standard models don’t work for lotteries
Determination of the optimal media spend is often complicated by a great number of factors of influence. For draw-based games, business dynamics heavily influence the result of any investments in media, to the extent that models such as regular multi-touch attribution (MTA) do not produce accurate output. The most important business dynamics are the size of the jackpot and the number of days left before the draw.
The display channel, for example, can be extremely effective when the jackpot is high, but not effective at all for lower amounts. Similarly, more people tend to buy a lottery ticket on impulse directly after seeing a banner a few days before the draw compared to a draw that is weeks away. Using this input to model optimal media spend for the coming period would not make sense.
It is, however, possible for us to quantify these two business dynamics and incorporate them in our MTA model to determine optimal media spend.