Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
116 lines (67 sloc) 13.5 KB
title date featured_image caption featured_image_caption
Four (Hundred and Thirty-Two) Weddings
2018-03-11
four_weddings.jpg
Using data to study what makes a wedding work.
Four brides featured in an episode of four weddings. Image credit: TLC.

When my wife and I first got engaged, we spent a fair amount of time doing what any newly engaged American couple does: watching a lot of wedding-themed reality television. I'm not ashamed to admit that I've seen more than my fair share of Say Yes to the Dress. There were other shows that made it into the rotation too. However, the one I remember most fondly was a little gem called Four Weddings, which aired on the TLC Network.

This show had everything. It featured couples from all over the country, with a wide variety of wedding themes, budgets, and eccentricities. There was just the right blend of community and competition. And for a math nerd like myself, there were enough numbers flying across the screen to keep my interest.

If you've never seen the show, here's a quick overview. In each episode, four strangers go to each other's weddings, and then rate the weddings on a few different scales. The bride with the highest-scoring wedding receives an all-expenses paid honeymoon (at least, in most cases). As the show progresses, you get to see not only how everyone rated the weddings, but you also get some basic demographic information about the brides. This includes things like their age, location, wedding size, and wedding budget.

Here's a quick preview clip from one of the episodes, if you'd like a little more context:

When I first discovered Four Weddings, I thought it might be fun to compile the statistics and see what sorts of trends emerged. But alas, this was in the stone age of 2011, before streaming television had taken off to the extent that it has now. In order to collect data, I had to wait for an episode to air, record it on our DVR, and then speed through it as best I could to record any relevant data. This was a slow process, and after a while I threw in the towel.

Recently, however, I discovered that Four Weddings is making a comeback. And whether out of pure coincidence, or to help market this reboot, it's now possible to find almost all of these episodes available for streaming on the TLC website. When I made this discovery, I knew what I had to do: make a pot of coffee, and watch a lot of women get married. Four hundred and thirty-two, to be precise, spread out over 108 episodes.

In what follows, I'd like to share with you what I learned.

"I Do" Demographics

Before we get into the competition aspect of the show, I'd like to start with some basic demographic information regarding the couples and their weddings. First, here's a heat map highlighting where each of those 432 weddings took place. You can also adjust the map to display the average budget per wedding by state.

As you can see, brides in most states were not represented in the show. Of the states that made an appearance, New York and Florida are by far the most frequent destinations; these two states alone account for nearly 42% of the weddings in the data set. On the other end of the spectrum, Ohio and New Hampshire are only represented by a single wedding each.

In terms of average budget, New York and New Jersey take the cake; weddings in these states on the show cost nearly $40,000, on average.

For more detailed budget information, here's a histogram with budget information for every wedding on the show. You can also look at breakdowns of the number of guests at the wedding, budget per guest, bride's age, spouse's age, and the age gap between the two people getting married.

The chart should speak for itself, but here are some additional highlights:

  • Overall, weddings on the show had an average budget of $31,058. They had a median budget of $25,000.
  • The average guest count was nearly 154. The median guest count was 145.
  • The average bride age was 29.2. The average age of the spouse was 31.6.
  • The average gap between the age of the spouse and the age of the bride was 2.4 years.
  • 52.5% of the time, the gap between the age of the bride and the age of the spouse was 3 years or less. 6.7% of the time, the bride was more than 3 years older than the spouse. 40.7% of the time, the spouse was at least 3 years older than the bride.
  • Not all spouses are husbands! There were a handful of same-sex weddings featured on the show.

This demographic information is nice to know, but doesn't really address the question that first piqued my curiosity: given the data about each wedding, can we predict which wedding will win?

Ratings and Rankings

Before digging deeper into the data, we'll need to know a bit more about how the winner is determined. First, each bride gives the other three weddings an overall experience score. This is a rating on a ten-point scale that reflects one's overall feelings about the wedding.

This isn't the entire story, of course. If it were, it's easy to imagine that the scoring system would quickly devolve into a race to the bottom, where each woman scores the other weddings poorly in the hopes that somehow her own wedding will be able to eke out a win.

To keep this from happening, brides also assess three other categories: the food, the venue, and the wedding dress. Unlike the overall experience score, however, brides simply rank the weddings in these categories as first, second, or third place. Each ranking has a corresponding point value:

Rank Point Value
1st 10
2nd 6
3rd 3

In this way, it's harder to game the system, since each bride is forced to dole out a perfect 10 once in each category.

This means that the maximum possible score a wedding can receive is 120 points: 30 points in each category. On the other hand, the minimum possible score is 27 (9 points in each of the ranked categories, and 0 for overall experience).

In the event that there's a tie in total score, the winner is the person with the higher overall experience score. There are tie-breaking rules in the event tha the overall experience scores are the same too, but we won't go into them here, because this was a rare occurrence and the rules changed over the course of the show.

Winning Weddings

Now that we know a bit more about how the game is played, let's explore some statistics that might predict the winning wedding.

One natural hypothesis is that wedding budget should have some predictive power. After all, more money likely means a fancier dress, better food, a more elaborate venue, or all of the above!

How can we look at the effect of wedding budget on the final outcome? One way to do it is to group the winners by whether they had the highest, second highest, third highest, or lowest budget among their foursome. If budget has little impact, you'd expect brides with the lowest budget to win about as often as brides with the highest budget. In other words, you'd expect each group to win roughly 25% of the time.

However, that's not what happens. As the chart below shows, weddings with the highest budget in their group won 45% of the time. On the other end of the spectrum, weddings with the lowest budget in their group won only 11% of the time. Perhaps unsurprisingly, people who spend more on weddings tend to have the weddings that people enjoy the most.

There are a few other stats you can play around with in the pie chart. Initially I'd thought that budget per guest might be a strong predictor of success as well, but it turns out to be weaker than the budget on its own (32.4% vs. 45%).

The other predictors you can explore don't have to do with budget, and instead involve the experience points ratings that the brides give and receive. First, note that the number of overall experience points a bride gives out doesn't seem to have an effect on the likelihood of her winning. This isn't so surprising; there's no real incentive for altruism on the show.

On the other hand, getting the highest overall experience score is an even stronger predictor for winning than budget: 60.6% of winners had the highest experience score in the group. Again, this shouldn't be so surprising: if people rate your wedding high in overall experience, it's probable that they also ranked your food, dress, or venue highly as well. In addition, as mentioned before, overall experience score is used as a tie-breaker in the event that multiple people tie for highest total.

Here's where things take an interesting turn. The last category you can examine is what I call the overall experience gap, that is, the difference between how many overall experience points a bride received, and how many overall experience points a bride gave. As you can see, this gap is the second best predictor for success among winners: over half of all winning brides also had the largest overall experience gap!

Of course, this doesn't necessarily mean that brides with popular weddings are sabotaging their competitors by giving them poor experience scores. In fact, since winning is strongly associated with high experience scores, it could be that the gap is larger for women who win simply because the overall experience scores they received were also high in their group.

Plotting All The Things

To close things out, let's take a kitchen sink approach. Here's a visualizaiton that lets you plot data from every wedding on the show. You can adjust what each axis is displaying, but the colors of the circles always correspond to final rankings. As with the pie chart above, winners are in blue, second place finishers are in green, then orange, and last place finishers are in red.

Here are a few things I noticed:

  • Wedding budget may be a predictor of success, it's worth considering whether there's a limit to this rule of thumb. Among the eight weddings that had a budget of $100,000 or more, only two of them were ranked highest in among their foursome.
  • Setting the y-axis to total points received tells a clear story about what it takes to secure a win. Nobody has received a "perfect" score of 120, but if you score 90 or above, you're pretty likely to win. On the flip side, if you score less than 80 points, you're probably not getting that honeymoon. In other words, there's a relatively narrow range where there's a lot of uncertainty about whether or not you'll win.
  • This is part of the reason why even a small experience point gap can help secure a win. In fact, nearly 38% of all episodes featured a winner and runner-up whose total score differed by no more than three points.
  • Of the three other categories that are scored, it appears that a strong showing in venue is the best predictor of success. Of the weddings that scored 26 or 30 points in venue (the highest marks possible), 66% won the episode. For food, 50% of weddings that scored 26 or 30 points won, and for dress only 39% of weddings that scored 26 or 30 points won.

Conclusion

Full disclosure: several episodes had incorrect tallies, or tallies that were simply impossible given the rules of the game. Where it was clear what the tallies should have been, I've adjusted the data. But there were a couple of times where it wasn't clear why things were off. In those cases, I've left the data as-is. Check out the notes in the CSV for more details.

So what did I learn from all of this reality television? As expected, on Four Weddings, money helps people win honeymoons. Sabotaging the overall experience scores of your peers can also help, though this doesn't seem to be happening too much, as it's pull off without looking like a jerk. Also, if you're going to splurge on one aspect of the wedding, go for the venue.

There are probably other conclusions to be drawn from the data, but I'll stop here for now. If you're interested in exploring the data more on your own, here's a link to the raw CSV file. If you notice anything interesting, please let me know!

There's plenty more data to mine, too. I just grabbed the lowest-hanging fruit. In an ideal world, I think it would be fun to perform some kind of sentiment analysis on the transcripts of the episodes, and see if there are any trends that can be pegged to how the women on the show speak.

But alas, there are only so many hours I can devote to this show in good conscience. For now, my own honeymoon with this show is over. But talk to me again in a couple of years, and maybe I'll be ready to review the data one more time.