Saving Twitter: 6 Steps for Brand Ad Networks
Brand ad driven social networks are hard, but also well understood
The most important dichotomy in advertising
There are two types of advertisers on the internet:
Performance: I spend $100 on ads and expect to make $200 in sales (within the time period I care about)
Small advertisers turn their ads on and off to make sure they are still driving sales.
Large advertisers build complex ML systems to measure the return on ad spend.
Brand: I spend $100,000 on ads to make sure people know my brand the next time they go to the store.
Small advertisers are mostly wasting their money (hard to get people to recall ads they only saw a few times).
Large advertisers are building durable moats (think Coca Cola). They use impressions as a metric of success.
According to Elon, 70% of Twitter’s ad spend is brand. Facebook’s is only around 50%. Google’s is probably even less (search ads are the exact opposite of brand development).
Step 1: Create a new ad unit that delivers “verified” impressions
Brand advertisers want their ads to be seen by real people (ideally real people within their targeting criteria, but at least real). They only have one metric they can really measure: impressions (and maybe reach).
Twitter should use the new blue checkmark subscription and create a type of ad that will only be shown to subscribers. This will result in brand advertisers paying much more per ad impressions, and quickly make up for the revenue loss due to 1/2 ad load (as promised in the copy for the subscription). In general, brand advertisers are competing against each other driving up CPMs. They don’t really have a good way to measure ROI in the short run, so they’ll probably overshoot the 2x CPM for a “verified” viewer.
Performance advertisers may also bid for these ads. But they’ll be just as happy to bid on unverified impressions if they result in sales.
The most sophisticated advertisers will run both and build multi-touch models to buy the cheapest inventory (no matter what you do to your ad system).
Step 2: Create a new ad unit that delivers impressions only adjacent to a subscriber tweet
Brand advertisers hate having their ads show up next to “bad content.” In general, this is why they care about moderation. They simply don’t want their ads associated, even subliminally, with racism, violence, porn, etc.
Whether this is rational or not is irrelevant. It’s been this way since the early days of cable ads where advertiser boycotts created a limit on the type of content a channel could broadcast.
Creating an ad unit that guarantees only to show next to verified tweets reduces the risk of this spillover. It does not eliminate it, so this won’t be as effective as only showing ads to verified tweeters. But it’ll create some amount of extra willingness to pay, which is critical for Twitter right now.
Step 3: Hire adtech experts from Snap, FB, Google (in that order)
Adtech is the software layer that decides which ad to show which user, and reports the results to the advertiser. This is a surprisingly difficult problem (more complicated than Tiktok’s feed ranking). Twitter generally has the worst performance of the big tech companies here. The reporting tools are not as good. The conversion optimized bidding is not as good. The auction dynamics are not as good. The creative generation/optimization is not as good. The advertiser tools to manage and test thousands of variants are not as good.
Each step that’s not as good in this funnel reduces ROI for an advertiser, resulting in less spend.
Snap is a good example of a company that made the investments in 2017 and 2018 and reaped the rewards in 2019 and 2020. This isn’t an infinite well of profit, but it’s definitely a large amount of profit that’s being left on the table right now.
Hiring from Snap first is due to the lack of sophistication at Twitter today. Many things that Google/FB engineers/PMs take as water probably don’t even exist at Twitter.
Tiktok could also be a good source, but probably hard to recruit from (given their own strengths and opportunities).
Start the investment now and within 3mo you should start returns (and more importantly, restore some confidence in advertisers).
Step 4: Relaunch Vine
Just take trending videos on Twitter and use the Vine brand to create a video only feed. Limit the duration to 30 seconds for posts on Vine. Initially, it won’t be as good as Tiktok, but it won’t be useless either. Limiting to 30 seconds is both on-brand for Twitter, and a vulnerable spot for Tiktok (who has really been pushing length to push time spent) and Youtube (who has no idea what to do with short videos).
Videos generally earn much higher CPMs which is why it’s worthwhile. Monetize immediately by giving creators tools to see who saw their video and which parts resonated. They’ll happily pay to be able to see this, and it’ll improve their iteration velocity on Vine to be faster than Tiktok. Short videos also have lower production cost increasing iteration velocity.
Step 5: Stop investing more in search
There is a world where Twitter search displaces Google (really. real conversations are just better than seo content farms). But today is not that day.
Build highly relevant search is very hard (particularly in domains where Pagerank and TFIDF heuristics don’t work… like Twitter).
Build search intent (getting people to search you first) is even harder and requires months/years of incredibly relevant search results.
Both of these are very hard technical problems demanding a large budget of smart engineers and large amounts of hardware for data processing. Both of those are scarce resources that would be better spent on ad optimization today. Once the ads are really good, you can move them into search.
Step 6: Start decreasing organic reach for non-subscribers
“At some point the executives at Google realized it was possible to get to the top of the search results without paying Google anything. They resolved to quickly fix this problem.”
That’s probably a fake quote, but it does explain how Google has (1) grown revenue and (2) slowly destroyed their product.
Twitter should not do the same. However today, there is too much organic reach on Twitter. This means if I post something, it’s too easy for people who are not connected to me to see it, and it’s too easy for me to see something that’s not paying for me to see it. On balance, if you worry about Twitter not existing in 2 years, you should be willing to slightly increase the cost of reach in exchange for a 90% as good Twitter definitely existing for the next 10 years.
Subscription is a much better model than ads for organic reach, and by definition limits one person’s ability to outbid everyone else (the root cause of all low quality ads).
It should be possible for Twitter to decrease organic reach, require a subscription to reach your followers, and converge on a stable equilibrium where most of the established people pay, and most of the new people get free boosts until they get 1000 followers.
Once that happens, stop. Don’t push organic reach toward 0 like Facebook Pages did… this is a path to unhappiness as only the most extreme content becomes capable of organic reach.
All unhappy families are unhappy in the same way
This playbook has been used by every ad driven consumer startup in SV, and for good reason. It creates very profitable businesses. It also completely destroys the product if taken too far.
But today, Elon’s challenge is not completely avoiding ads, but rather, dialing the ads to get a sustainable business that produces $1-2B in profit per year. Without that, Twitter will always be at the mercy of mob, rather than the container.
Hopefully Elon survives the next 2 years there, and then sets up a governance structure that is not incentivized to dilute the organic experience any further than necessary to sustain the $1-2B/yr profit level.
If he pulls that off, he might actually create a perpetual town square for all of humanity.