A Product Manager’s reading of Twitter’s S-1

Unearthing business drivers: the US vs. ROW growth divide, ARPUs, timeline RPMs, and more

Despite a government shutdown, Twitter was able to file their S-1 on sec.gov yesterday. There has already been extensive coverage, the best of which has been on Quartz and WSJ.

This is different than a reporter’s coverage of the S-1. It’s a tour through the lens of a product manager who has worked at the intersection of consumer tech and monetization the past five years, half at Google and half at Foursquare. There are two sections: users and revenue.

Disclaimer: This is not a recommendation for or against investing in Twitter. I stick to a random walk down diversified ETFs for my personal investments.

Revised S-1 Q3 Update

Twitter filed a revised S-1 on Oct. 15th to include Q3. The data and graphs in the rest of this post reflect this new data. Q3 highlights:

• Renewed US growth: Twitter added as many US MAUs in Q3 (4m) as they did in the first two quarters of 2013 combined. QoQ growth was 8% in Q3 compared to just 2% in Q2.
• Declining engagement per user: Timeline views per MAU fell -2% in the US and -1% ROW, despite growing in the first two quarters of the year.
• Revenue still strong: Global timeline RPM is up 27% YoY in Q3 vs. 21% in Q2. Global ARPU is up 34% YoY, though that growth rate has declined for 4 straight quarters from 73% in Q4 2012.
• CPC decline continues: Q3, with a -20% drop, is the sixth straight quarter CPC has fallen.

Users

Below is a flash card worth of the key facts on Twitter’s usage:

• 53m US MAUs + 179m ROW MAUs = 232m Global MAUs
• ~100m DAUs implies 43% DAU/MAU (vs. FB’s insane 58%)
• 76% of MAUs are on mobile (vs. FB’s 71%)
• 5% of MAUs are estimated to be spam accounts
• 500m tweets per day implies 5 tweets / DAU / day, which strikes me as very high. Twitter never defines whether retweets or favorites count in this metric, but I have to assume they do.There’s certainly a power-law distribution of tweets, which likely means the median is still 0 or 1.
• 350bn all-time tweets

The graphs below show the deceleration of US MAU growth starting at the end of 2011, with a slight bounce back in Q3.

ROW MAU growth starts declining later in mid 2012. Despite the much larger user base, ROW YoY growth as of Q3 2013 is still 26% higher than US (41% vs. 32.5%).

Twitter uses timeline views as their core metric to measure engagement. They provided quarterly timeline views in their S-1, from which you can derive the monthly views per MAU below. Engagement per active user has been flat in the US for the past year, while still growing and playing catch up internationally. Both regions saw small declines in Q3, however.

Revenue

Revenue growth looks like a stronger story than user growth. This is likely in large part due to Twitter starting to monetize 3 years after launching, leaving lots of potential views to turn into ad impressions. Adam Bain and Kevin Weil, Twitter’s sales and product leads for revenue, have clearly been busy the past few years introducing ad products to increase sell-through rates.

As a result, global revenue has grown healthily: $28m in 2010, $106m in 2011 (+276%), $317m in 2012 (+198%). The first 9 months of 2012 was $204m vs. $422m in 2013 (+107%).

A big monetization opportunity is ROW, which accounts for 25% of revenue even though it is 77% of MAUs (as of H1 2013). ROW has gone from 2% of revenue in 2010 to 4% in 2011 to 17% in 2012 (pg F-46).

The graphs below show the beginnings of a deceleration in annual revenue growth rates. The difference between the US and ROW is stark here, where YoY growth in the first 9 months of 2013 was 296% in ROW vs. 78% in the US.

We can only calculate global ARPU on the quarterly level because Twitter does not provide a US/ROW split for monthly revenue in their S-1. As of Q2 2013, Twitter’s global ARPU of $0.64 is 39% of Facebook’s $1.60 ARPU.

It’s possible to calculate RPM for timeline impressions given the quarterly revenue and timeline view counts Twitter provides. While RPM was flat for the first 3 quarters of 2012, it has risen since.

Without knowing the mix of US vs. ROW, it’s hard to isolate US RPM movement. Undoubtedly, the growing contribution of lower RPM, higher volume ROW impressions is weighing down the global figure.
While RPM has increased in 2013 the rate of growth of timeline views has steadily decreased (except for Q3 US), which reflects Twitter’s slowing MAU growth combined with plateaued engagement per user (timeline views per MAU).

Bonus topic: decreasing CPCs from lack of auction pressure

“Average cost per ad engagement decreased 18%, 9%, 19%, 12%, 46% and 20% sequentially in the three months ended June 30, 2012, September 30, 2012, December 31, 2012, March 31, 2013, June 30, 2013 and September 30, 2013, respectively.” (p65)

As Twitter has expanded its available ad impressions, advertiser demand has not kept pace and average CPCs have fallen for the last 6 quarters.

The bear case: existing advertisers aren’t growing their budgets or raising bids, and not enough new advertisers are coming on to the platform.

The bull case: there exists an opportunity to raise ad prices by bringing more advertisers into the auction, which the acquisition of MoPub sets Twitter up well for. Expect programmatic buying and Twitter inventory availability on MoPub’s exchange to bring more demand into the Twitter advertising ecosystem.

Bonus revenue facts

• 65% of revenue is mobile (vs. 41% for Facebook)
• $2.58 Timeline RPM in US vs. $0.36 internationally (pg 66)
• Data licensing as a percent of revenue has decreased from 74% to 27% to 15% to 11% in 2010, 2011, 2012, 2013 9mo respectively. Data revenue has continued to grow ($47m vs. $35m in 2013 9mo vs. 2012, 35% YoY), but the advertising business has grown much faster.

What keeps Dick up at Night

Based only on what I can glean from the S-1, the three areas below are what I assume Twitter will focus on for 2014:

1. Revive US user growth: In the first half of 2013, US MAUs had only grown 8% and timeline views per MAU had grown just 3%. Twitter’s timeline RPM is 7x higher in the US than abroad. Reigniting US user growth is the fastest way it can grow its top-line revenue. Fortunately Q3 saw renewed signs of US MAU growth (8% QoQ vs. 2% in Q2).
2. Monetize the international usage boom: Twitter is growing 26% faster abroad than in the US, despite already having 3.5x as many MAUs outside the US. Twitter has a huge opportunity to generate more of its revenue abroad. It still has not opened its self-service ad platform outside the US, which is likely a good place to start since scaling direct on-the-ground sales will be costly and time consuming.
3. Stem the CPC decline: After 6 straight quarters of CPC compression, Twitter needs a turnaround. Bringing more advertisers into its auction by making its inventory available on exchanges and using MoPub to open up programmatic buying could be the panacea.

Appendix

All of the calculated stats and graphs can be found in this spreadsheet. The S-1 was the only data source used.