The Blur blitzkrieg
Or, how to turn the tables on the largest NFT marketplace
Poker's the only game fit for a grown man. Then your hand is against every man and every man's hand is against yours. Team-work? Who ever made a fortune by team-work? There's only one way to make a fortune and that's to down the fellow who's up against you. -W. Somerset Maugham, 'Straight Flush'
In the span of four months, Blur managed to unseat incumbent OpenSea as the highest-volume NFT marketplace in crypto. It’s one of the biggest David-and-Goliath battles in recent crypto history, won largely due to a very clever reward program designed to drive loyalty and liquidity.
Unlike other drops Spindl has analyzed (and which we’ll be publishing soon), Blur managed to aggressively steal marketshare while attracting users whose retention rates are healthy by quick-churning Web 3 standards. Like any two-sided market, Blur faces a chicken-and-egg problem: without a supply of NFTs by creators, there’s no demand by buyers, and vice versa.
Their first drop, like other drops, merely rewarded NFT trading activity in general. The two subsequent drops were aimed specifically at incentivizing the buy and sell sides, while also rewarding ‘loyalty’ to Blur over other marketplaces. In effect, it was a very studied and precise vampire attack that aimed to draw users from other marketplaces while carefully building Blur’s own.
From October 19th to December 6th, anyone who listed NFTs on Blur would qualify for ‘Care Packages’ (boxes of token rewards for an as-yet unlisted token). The size of the kickback would be a function of a ‘loyalty score’ which reflected how much you favored Blur in listing your NFTs, and/or listed them for less on Blur. It was a classic ‘vampire attack’ which worked spectacularly well: Blur NFT listings absolutely exploded.1
From December 6th until February 13th (the day before the $BLUR token launched), the incentive scheme flipped to being all about bidders.
If you placed many and/or very risky bids (i.e. close to the floor price) for prominent collections, you received a number of points that translated into yet more ‘Care Packages’ (i.e. air-dropped bonuses). The actual amount of the kickback varied as a convoluted function of bid price over floor price, but directionally it was clear what buyers needed to do: bid more and higher. As with the prior drop, it worked very well indeed: trading volume popped and stayed high.
How much of this newfound volume was actually driven by air-dropped users though?
Well, without something like Spindl attribution to measure effectiveness, you’re only speculating based on correlations, but that’s about the industry standard at the moment so let’s go with it. As you can see below, most of the elevated volume in listing was by airdropped users.
Ditto on the bidding side, the airdrop incentive drove most of the increased trading volume on Blur, following by surge in non-drop bidding after the token launch.
Lots of marketplaces and chains use drops to incentive user engagement. Along with referrals programs and maybe quests, it’s about the only user-acquisition lever available to Web 3 growth teams (very few Web 3 projects we’ve talked to run conventional Web 2 ads).
The mechanism is unique because it’s not rewarding easily-gameable actions like transactions, but instead actual trading liquidity. Whether you’re an exchange dealing in IBM shares, ad impressions or NFTs, you want market depth around the clearing price so that big buyers can’t come in and pump or tank the market (too much). That’s what Blur was really incentivizing in addition to loyalty, and on that front it absolutely worked.
But is it real?
One of the doubts around airdrops (and quests and referrals) is this: are these real users or Sybils/farmers? Do they stick around after the cash has been doled out? Does this even work out on a return-on-investment basis?
Below is the retention chart for Blur non-airdropped users. Cohort retention is a key type of growth analysis: bucket all users who started in a certain time window, and measure what percentage of them stayed active through time. The steeper the plunge, the worse the retention and the higher the churn (and likelihood these are fake or at least farming users).
The picture is rough for non-airdropped users:
But…it’s much, much better for airdropped users. Users are not merely trying to rack up some points and leave, they’re actually sticking around (as the incentive scheme requires).
Of course, a cynic might say Blur is buying itself retention metrics with a cleverly-designed reward system…but so what? Re-engagement campaigns are a tried and true marketing technique in Web 2: gaming companies routinely spend mountains of ad dollars just to induce users to come back. The loyalty programs at airlines and grocery stores are all about bribing users with kickbacks to stick around. The question is if it’s net profitable and, in the case of Web 3, are the users real?
Most drops we’ve looked at are heavily farmed and Sybiled2: humans programmatically create lots of wallet IDs and jump through the hoops to qualify for a drop. Once dropped, they transfer their gains to a single wallet and dump the token (or list any NFTs). The resulting circle pattern of post-drop transfers has become so common at Spindl we call it the 'bacteria plot,' because it looks like a Petri dish exposed to microbes.
There’s plenty of fraud in Web 2 too: an entire ads anti-fraud industry is there to stanch the flow of fake ad impressions and clicks. Marketers just treat the loss like retailers do ‘shrinkage’ (accounting-speak for shoplifting): what really matters is the resulting return on the marketing budget (even with fraud baked in).
What’s the ROI of drops?
In normie Web 2 marketing, return on advertising spend (ROAS) is the single most important metric in any growth-driven company:
In the most basic of marketing campaigns, that’s just value extracted from a user (before they churn) over how much you paid the Facebook ads system (say) to bring that user in. The Blur math though, as with most things in crypto, is a lot weirder than in Web 2.
Let’s calculate Blur’s CAC: users were ‘given’ a reward in a token that hadn’t launched yet, and whose value was unknown. Anticipation for the reward made the value of said token absolutely pop when it did finally launch. Do we now calculate a user’s acquisition cost by the retroactive value of that token, or something else?
If we do use the later pumped price as the CAC, then as the token rises in value—i.e. the more successful the token launch—the more it seems Blur is paying for users. Physicists dismiss theories that don’t jibe with self-evident reality as ‘unphysical’: any user-acquisition calculus where return on spend goes down when token value goes up is clearly ‘unphysical’. Driving the token value up is as much a marketing goal as a flat retention graph.
Strictly speaking, the ROAS of the Blur airdrop was definitionally zero, as they don’t charge a platform fee. Even assuming a (hypothetical) OpenSea-matching royalty rate of 2.5%, the ROAS to date was a piddly 8% (meaning they ‘lost’ 92 cents of every acquisition dollar from the drop).
This is premature as rewarded users haven’t had much of a lifetime to earn out yet. Also, given Blur’s sudden dominance of the NFT market, a low ROAS seems directionally wrong as clearly something worked here. Conventional ROAS, at least during an airdrop blitz like this one, is probably a mis-framing of the problem.
Blur will have to figure out the correct calculus to incentivize long-term retention and monetization, not just excitement around the next token pump. In Web 3, the ‘acquisition cost’ isn’t a one-time cost as in Web 2, but a running rent Blur is paying to keep its users around. Any Web 3 ROAS needs to reflect that new calculus.
For one, there isn’t a single, one-time and discreet acquisition cost for every user. Protocols don’t pay an ads system for a new user, and then it’s just revenue gravy from then on. From the excitement around airdrop 2 (and the above retention curve), it’s pretty clear Blur will have to continue to pay rewards to keep many users around in airdrop 3 (and 4 and 5…).
That’s not necessarily a bad thing if the math works out, but Blur will have to figure out the correct calculus to incentivize long-term retention and monetization, not just excitement around the next token pump. Like an airline, Blur will need to reward loyalty while also evading drop farmers who scam the drop system. The ‘acquisition cost’ isn’t a one-time cost as in Web 2, but a running rent Blur is paying to keep its users around. Any Web 3 ROAS needs to reflect that new calculus.
Second, the nature of native tokens make this math even more complex.
Crypto projects are like the Federal Reserve, printing the very currency that their entire economy is denominated in, and which is ultimately what’s typically deployed for user acquisition. One of the oddities of Web 3 is that marketing budgets are often locked up in a treasury run by a DAO or foundation, and that a new rewards (AKA marketing) campaign must be subject to a DAO proposal process. Perhaps more importantly, that budget is denominated in the protocol’s native token, not USD or other fiat.
The numéraire (to use a fancy economist term) of the Blur NFT economy is the $BLUR token itself. It’s almost as if Ford Motor Company could buy Facebook Ads with shares in Ford to attract buyers: what exactly is Ford’s CAC if the resulting sales bolster Ford’s very share price? Assuming Ford can pay users to buy cars in their own shares, when exactly does the music stop?
Well, when does the music stop for the Federal Reserve?
When people stop accepting dollars in exchange for global petroleum supplies and beer from your corner store. Unlike dollars, there are neither aircraft carriers nor nuclear weapons backing up $BLUR. But money is as money does: if someone will trade you goods or services (or other currencies) for it, that’s money. In the end, how is it any different than Spindl selling equity for USD and then using that fiat and yet more equity to hire employees? The only difference is in the liquidity of this new, magical unit of account called a token, and the nature of its buyers and sellers (and how it’s regulated).
Marketplaces without tokens, such as Blur competitor OpenSea, are kind of stuck. Like smaller economies that must borrow in a foreign currency (and hence can’t inflate or deflate their currencies at will), OpenSea must either pay for user incentives in fiat or stablecoin, and/or take marketplace royalties in order to keep the lights on. That of course subjects them to being undercut on price on other marketplaces which don’t need royalty revenue, in addition to vampire attacks via token rewards as we saw here.
OpenSea still has a strong conventional equity balance sheet, and it can raise money via that funny money instead of a token, but that’s far more involved and harder to conjure. In a sense, they’re like a developing country that must borrow in dollars rather than their own sovereign currency, on the hook for a debt whose denominating currency they don’t control.
Or is Blur the developing-world economy, much like the high-flying Asian economies in the 1997 currency crisis, and will suddenly see a flight of capital and a plunging exchange rate once the excitement is over? In that case, you’d rather be OpenSea and their old-school equity balance sheet.
Whether Blur or OpenSea, at some point NFT marketplaces will have to develop business models that provide ongoing revenue beyond the speculations of hard-money venture capitalists or token pumpers. Then, the ‘right’ way to calculate the economics of Web 3 user acquisition will matter much more than it does now.
One of the amusing features of Web 3 is that every pre-existing Web 2 marketing concept is re-named as if it had just been invented. ‘Vampire attacks’ are just ‘conquesting’, a practice that dates back to newspapers and which was ported online via hostile keyword targeting on Google Adwords (among other tricks). The openness of the blockchain has made it even more common in Web 3.
The etymology of ‘Sybil attack’ has to be one of the most fascinating in computer security: it stems from a book Sybil, named after the pseudonym of a case study in multiple personality disorder. The patient was later revealed to be Shirley Ardell Mason, an art teacher who lived in obscurity, and who claimed 16 different personalities. The paintings found in her house after her death were sold at auction, and possess a grim, haunting quality.