Issue 037 - Media - Human attention
How much human attention is going into microdramas?
Reuters reported that short, vertically filmed microdramas are rapidly growing in the U.S., with industry revenue projected around $1.5 billion in 2026 and about 66 million U.S. monthly active viewers last year.
The problem
Estimate the total number of hours Americans will spend watching microdramas during 2026, and the approximate number of individual episodes that represents.
Then estimate the industry's approximate revenue per viewer-hour and revenue per episode viewed.
Is a $1.5 billion microdrama industry primarily the result of an enormous amount of audience attention, or unusually effective monetization of a smaller amount of viewing time?
Because Fermi problems target an order of magnitude, I normally use no more than two significant digits and write most calculations in scientific notation; the Fermi reference explains both conventions.
Before checking sources
Matt's first pass
My assumptions must be off, because this does not look right. Some starting assumptions:
I started with 3.3 x 10^8 Americans, assumed 80% have smartphones, and got 2.6 x 10^8 smartphone users. I then assumed 30 minutes daily on average watching these microdramas. This is the assumption I thought was high: maybe some Americans are watching this many, but on average across all of them? Probably high.
smartphone users ~= 3.3 x 10^8 x 0.8
~= 2.6 x 10^8 users
daily viewing ~= 2.6 x 10^8 users x 30 min/day
~= 7.8 x 10^9 user-minutes/dayLooking at the projected $1.5 billion revenue in 2026, that is about $1.3 x 10^8 per month.
monthly revenue ~= $1.5 x 10^9 / 12
~= $1.3 x 10^8/monthIf I take the estimated 2.6 x 10^8 users and assume 900 minutes watched monthly, 30 minutes for 30 days, that is about 2.3 x 10^11 user-minutes per month, or 4 x 10^9 user-hours per month.
monthly viewing ~= 2.6 x 10^8 users x 900 min/month
~= 2.3 x 10^11 user-minutes/month
monthly viewing ~= 2.3 x 10^11 / 60
~= 4 x 10^9 user-hours/monthBrief aside: the typical user will not be able to view all new content, but let's say they can view 5% of the new content produced monthly, and that is estimated at 900 minutes viewed per the above assumption. Then the total amount of new content produced might be closer to 18,000 minutes monthly, probably a huge underestimate. At 2 minutes per episode, that is 9,000 new episodes monthly.
new content/month ~= 900 min / 0.05
~= 1.8 x 10^4 min/month
new episodes/month ~= 1.8 x 10^4 / 2
~= 9 x 10^3 episodes/monthUsing the viewing estimate:
revenue/viewer-hour ~= $1.3 x 10^8 / 4 x 10^9
~= $3 x 10^-2 per viewer-hourThis is why I thought my assumptions about time spent watching must be too high, because that is a tiny revenue per viewer-hour.
Continuing on the assumption the viewer-hours are correct, revenue per episode should be about 1/30th of that, or about 0.1 cents per episode viewed. Assuming 9,000 new episodes are produced monthly, if their ad revenue is divided equally, which it will not be, that would be:
revenue/new episode ~= $1.3 x 10^8 / 9 x 10^3
~= $1.4 x 10^4 per episodeAgain, my estimate for the number of new episodes per month is low.
Calibration Score
Matt's Calibration Score: 30 / 100
Higher is better: earn points for accurate pegs, sound models, correct math, and a result close to the sourced answer. The image shows percent full of it: 100 minus the Calibration Score.
Pegs: 0/30. The reported monthly-active-user anchor was overlooked, and daily-use assumptions were too high.
Model: 15/30. The viewer-hours model was right, but starting from all smartphone users distorted the calculation.
Math: 5/10. Some revenue-per-unit handling blurred cents, CPM, and episode economics.
Result: 10/30. The revenue-per-viewer-hour result was about an order of magnitude low.
Grounding facts
A few billion viewer-hours is large, but not unimaginable compared with major social video platforms. The striking part is that a mobile-first niche format can already command mass-market attention.
Revenue per episode viewed around a penny also changes the story. If episodes are tiny and cliffhanger-driven, a large business can be built from very small payments, subscriptions, ads, or unlocks multiplied across enormous repeat viewing.
After checking sources
Check and recalibrate
The big correction is to use the reported audience: 66 million monthly active viewers, not all smartphone users. That is already a broad audience, about one-fifth of the U.S. population.
The other correction is that monthly active viewers are not necessarily daily viewers. A reasonable central Fermi scenario might be: each monthly active viewer watches on 10 to 20 days per month, for 15 to 30 minutes on a viewing day. That gives about 2.5 to 10 hours per month. Use 5 hours per month as a clean middle estimate:
monthly active viewers ~= 6.6 x 10^7
viewing per viewer ~= 5 hours/month
months/year ~= 12
annual viewer-hours ~= 6.6 x 10^7 x 5 x 12
~= 4.0 x 10^9 viewer-hours/yearA high-engagement scenario using 30 minutes every day would be:
high scenario ~= 6.6 x 10^7 x 0.5 h/day x 365
~= 1.2 x 10^10 viewer-hours/yearSo a reasonable answer is probably on the order of a few billion to about ten billion U.S. viewer-hours per year, with 4 billion viewer-hours as a useful central estimate.
If the average episode is about 2 minutes, then each viewer-hour contains about 30 episodes:
episode-views ~= 4.0 x 10^9 viewer-hours x 30 episodes/hour
~= 1.2 x 10^11 episode-views/yearAt 1-minute average length, the same attention would represent about 2.4 x 10^11 episode-views. So the answer is roughly 100 billion to 250 billion individual episode views.
Now divide the projected revenue by the attention:
revenue/viewer-hour ~= $1.5 x 10^9 / 4.0 x 10^9 viewer-hours
~= $0.38 per viewer-hourAnd divide by episode-views:
revenue/episode-view ~= $1.5 x 10^9 / 1.2 x 10^11
~= $1.3 x 10^-2
~= 1 cent per 2-minute episode viewedThe annual revenue per monthly active viewer is also helpful:
revenue/monthly active viewer ~= $1.5 x 10^9 / 6.6 x 10^7
~= $23 per yearThat does not look like extreme monetization per viewer if averaged across the whole monthly active audience. The business can reach $1.5 billion because the audience is already large, attention is recurring, and a smaller highly engaged paying group can subsidize many casual viewers.
Post-check reflection
Matt's reflection
I botched it. I overlooked the explicit number given of 66 million active users in the U.S. and worked backward from assumptions about smartphone users, then made a poor assumption about viewership among that group. I further botched it by assuming daily watching from that monthly active user pool, and these two assumptions should have pushed my answers down by at least an order of magnitude when looking at revenue per user and viewer-hour.
Ultimately the answers I got were not terrible. I was about an order of magnitude low for revenue per viewer-hour. It looks like there is no hard estimate for the number of new episodes monthly, but my given answer is not ridiculously absurd.
This is not an insignificant percentage of Americans, about 20% of us, and not a trivial amount of money. It is probably a good idea to dig in on this matter and become more familiar with it. A spend of about $25 per active monthly user sounds high, but reasonable.
Recommended memory peg
For attention-economy problems, remember viewer-hours = active viewers x hours per viewer per period, episode-views = viewer-hours x 60 / minutes per episode, and revenue per viewer-hour = revenue / viewer-hours.
Reader results
Bars show how submitted estimates sort into the answer choices from the gut-check prompt.