Alex Constantinescu

EA launches: what drives buzz and sales

A game's sales drop about 16% every week after launch, so the first five weeks decide it. Teasers created more than twice the buzz of trailers, and paid tweets stopped helping at around 20 a week.

Bar chart of regression coefficients showing teasers drive the most launch buzz
Bar chart of regression coefficients showing teasers drive the most launch buzz
  • 16EA game launches, 2017–2018
  • −0.177log-sales per week after launch (p < 0.001)
  • 2.6×teaser effect vs trailer (β 0.259 vs 0.098)
  • ~20paid tweets a week before returns diminish

Context

The brief came from a fictional small publisher, ABA Games, that wanted to learn from Electronic Arts before planning its own launch. There were two datasets: daily media and buzz data, and weekly sales, across 16 EA launches in 2017 and 2018.

The question

  1. What drives game sales?
  2. What drives online buzz?
  3. What should a smaller publisher do with that?

Approach

  • Data work. Aggregated the daily data to weeks so it could be merged with sales. Cut the panel at the first launch. Built launch windows per game. Dummy-coded overlapping launches instead of dropping them.
  • Buzz. A composite of Reddit volume, Reddit sentiment and news volume.
  • Models. Game fixed effects throughout, to absorb franchise strength, with HC3 robust standard errors. Log-sales on buzz, weeks-to-launch and seasonality. Buzz on owned media, paid volume and its square (to catch diminishing returns), news, trailers, teasers, press releases, conventions and lagged buzz.

What I found

  • Sales decay fast. About −0.177 log-sales per week after launch (p < 0.001). The first five weeks decide the result.
  • Buzz did not predict sales (p = 0.87). I report this rather than hide it: buzz and sales are likely determined together (Duan et al., 2008), and weekly aggregation blurs the link.
  • Franchise strength dominates. FIFA and Battlefield carry the largest fixed effects; NHL is negative.
  • What creates buzz: news coverage (β = 0.021, p < 0.001), teasers (β = 0.259) well ahead of trailers (β = 0.098), and press releases (β = 0.056, p = 0.002). Paid tweet volume alone is not significant, but its squared term shows diminishing returns.
  • Timing: buzz peaks 16 to 20 days after a trailer, at about six times baseline, with a half-life of about 17 days.

Recommendation

  1. Front-load the budget around launch. Spend after launch has little left to win.
  2. Cap paid tweets near 20 a week (about $10,200 at EA scale). A smaller publisher with a smaller network will hit that ceiling sooner.
  3. Release the trailer about 2.5 weeks before launch, paired with a paid push, so the buzz peak lands on launch week.

Limits

These results are observational, from one publisher and two years. The null result on buzz and sales is a measurement and identification problem, not proof that buzz doesn’t matter. Convention effects are biased by several games competing in the same week.