I run a small video hosting service called Stillhaven. We are a small team of four based out of Michigan. This week my team and I decided to close the free plan to new signups.
This was not a hard decision. The free plan was an experiment in marketing. Our hope was that in running ads and marketing efforts to our free plan, users would see the value and upgrade when they hit the free limits. Seemed reasonable enough. We expected some abuse, of course, but we limited our marketing efforts to the US and Canada, in lieu of worldwide, in hopes of keeping the cost reasonable while we tested this marketing drive after launching to the public.
Here is what it actually did in just 30 days.
The numbers
- 50 free accounts in a month.
- 0 upgrades to a paid plan.
- 5 suspended for breaking the terms.
- Most never signed in again after their first two days, and one big video upload.
- Total cost of over $1,000 in transcoding and transcription alone.
The cost alone wasn't a deal breaker, nor did it break us. We already had enough paying customers inside of Stillhaven before we launched publicly, so it was a small hit on our bottom line, not a crippling expense for a new startup. A few months' worth of testing this free option was a lesson, but when you add in the advertising, time spent on moderation, storage, etc., it was a $2,000 lesson that was only going to get worse.
What the free accounts did
The pattern was the same each time: sign up, upload within minutes, never come back.
- A full episode of a current TV drama, uploaded four minutes after signup. Nobody ever played it.
- A 1990 feature film, uploaded three times by one account from release rips, then passed around by link.
- Dubbed episodes of a streaming series, embedded within hours on a site selling "Unlock 720p".
- 5 accounts were suspended within 24 hours for uploading pornography.
- A subscription adult site asked for an account. I said no. Our payment processor does not allow it, and one processor account carries every customer's billing. Ironically, this was the one legitimate contact from these specific marketing efforts, and we had to tell them no.
Our automated scanning caught all of this, and I am grateful that my team had the foresight to include this scanning before we opened it to the public.
Where they came from
Two sources surprised us.
The first was ChatGPT. One day, five free signups arrived within twelve hours, and every one was referred by ChatGPT. Two of them uploaded the dubbed episodes. I had written the site to be easy for AI assistants to read, and it worked. I assume someone asked for free private video hosting, and the AI assistant answered accurately.
The second was our own advertising. Over 30 days Google Ads spent $401.48 for 120 clicks and 3 signups. All three were free accounts. The ad group that took the most money was built around "private video hosting". Its search terms included "anonymous video host" and "upload video anonymously". To me, private meant a client's unreleased cut. To many searchers it means hiding what you upload. We were paying about $3.30 a click to recruit the accounts that we ultimately ended up suspending.
What we tried first
Before closing the plan we added:
- A block on disposable email domains at signup.
- One account per inbox, counting Gmail's dot and plus variants as the same address.
- A hold on free uploads whose file names look like release rips.
- A review queue that shows me the flagged frames.
- A hard bandwidth cap and a monthly upload cap on free accounts.
Each one did its job. None changed the result, because the problem was not a hole in the plan. It was who the plan was for.
What replaced it
- Free is closed to new signups. People who were using it are being moved to the new entry plan at no charge.
- The entry plan is $9 a month: 100 GB of storage and 300 GB of bandwidth.
- Monthly plans start with 14 days free, with a card. The trial has its own limits, including two hours of new video on the entry plan. A feature film does not fit.
- Every plan now has a published monthly limit on new video. Transcoding is our largest variable cost, about 8 cents per source minute. Until this week the limits were daily and unpublished.
What we got wrong
We kept the free plan for the email addresses. We thought a list of free users was a list of future customers. It was a list of people who wanted free hosting.
For a while we also read abandoned checkouts as demand. Over 50% of the abandoned checkouts coming from Google and Bing ads were bots, and every bot chose the most expensive plan.
The free plan gave us a signup count to look at, and nothing more.
What we don't know
Whether $9 with a card is the right door. We will know more after the first few hundred people see it. And we're still allowing a 14-day trial, but with a card.
If you run a small service with real storage or compute costs, I would like to hear what your free plan did. Write to me.