Don’t get caught off guard with inaccurate or ambiguous data
As the 2016 election results swung in Donald Trump’s favor, many people were stunned by the inaccuracies in polling. And although the miss was not unprecedented, it did question the use of data techniques to accurately measure voter sentiment. So, what happened and why should the casino marketer care?
During the run-up to the election, polling was used to help campaign staffers understand how their investments were paying off and where to allocate additional resources. This makes polling data a vital tool in understanding the return on campaign investments. But every poll has some statistical error that could affect the data’s reliability. Many pollsters will try to reduce this error by aggregating multiple polls together; weaker polls will be combined with stronger polls that use a better sampling methodology.
Sometimes there are more systematic problems that affect the overall quality of the polling. For example, some polls could underrepresent certain demographics of voters – there might be more women represented in the poll, or less white men, or not enough minorities. This could introduce sampling biases that inaccurately represent the voting population. In most cases, these biases can be “smoothed” out when looking at poll data together. But sometimes, state and national polls all miss in the same direction. And regardless of the methodology used or the size of the sample, it produces the same inaccurate result. In this case, national polls overstated Clinton’s lead over Trump by almost four points – and state polls, especially in Wisconsin, Michigan, and Pennsylvania, grossly underestimated Trump’s level of support with certain demographics.
Of course, there were other factors at play here as well. Voter turnout this election cycle was at its lowest point in 20 years. Fewer Democratic voters came out than in the last two elections. Whereas Republicans, riding a wave of populism, had an easier time bringing voters to the polls with targeted messages that resonated with the demographic. And ultimately, what a voter says in polling matters far less than a voter showing up to vote.
The challenge with inaccurate polling is that not only do you get voter sentiment wrong, but you allocate resources inefficiently. Clinton’s campaign team had her traveling to Tempe, Arizona, for a campaign rally the week before the election. Polls seemed to indicate that Clinton had a reasonable opportunity in Arizona. But the data was inaccurate, and Arizona was an easy win for Trump. For Clinton’s team, this trip was a bad investment at the expense of shoring up support in Midwestern states.
How this applies to casino marketers
In gaming, casino marketers have similar challenges in understanding if they are “winning.” The data might give you a false sense of comfort that the investments you are making are paying off. The reality might be very different, and the ramifications of bad resource allocations could prove costly.
For example, let’s say that you run a free play promotion for one month and you see that gaming revenue is up by 5% from the prior month. You may conclude that the promotion investment paid off. However, what if you compared this result to the prior year’s month and saw no material change? Could you still say with conviction that the investment was successful? What if you reviewed the audience who received the promotion and evaluated the spend relative to overall monthly gaming revenue? This analysis might show that the conversion rate for the promotion was only 2%. And of those players that converted and used the promotion, the impact to monthly revenue was marginal because this demographic of player did not respond to the promotion message. This deeper analysis could lead you to a more accurate conclusion.
A distorted view of the success in your marketing investments might lead you to more bad choices, compounding the problem. And if you get it wrong, not only is it a wasted investment, but the opportunity cost had you invested in something else will be very high (remember Clinton in Tempe, Arizona).
So what should you do to prevent getting caught off guard by inaccurate or ambiguous data?
- Be diligent with your analysis – always question the outcome. If it is too good to be true, it probably is.
- Know what your goal is and the metrics you will use to track progress towards your goal.
- Invest in the skills and technology your team needs to be successful with data.
Remember, the poll that counts is the one where you win.
