Arbitrum’s $85m Growth Campaign

What happens when you incentivise growth without permission-less identity

Imagine being a business and running a promotion where you offer people $3 of value for every $1 they spend. Oh, and there’s absolutely no conditions on who can claim this offer. Your grandma, the homeless dude down the road, a well paid executive, or a normal middle class person are all eligible for this offer.

What do you think is going to happen? Well, the people who need the money the most and are least likely to be repeat customers will be coming in droves to swipe you clean until you run out of money or inventory to sustain this offer.

The good news is that the real world doesn’t work this way as free markets ensure business like that die quickly.

The bad news is that crypto businesses do work this way and the free markets continue to promote their bankroll.

Introduction

The above scenario is exactly what Arbitrum more or less did, except with $85m and ended up with a $60m loss in the process. Let’s dig into what exactly the scheme was, how it was structured, and what we can learn from this all.

Arbitrum DAO structured this in a way where certain verticals and their respective apps would receive ARB tokens to incentivise on their platform. Ultimately the idea is that by incentivising usage of these platforms, Arbitrum as a network would receive more fees and the end protocols would benefit as well. Turns out one party won here and the other one less so (I’m sure you already know who the loser is here).

The analysis is pretty high quality with the sophistication around measurement and give props to the Blockwork team for clearly laying out why, what, and how around their approach.

You can read/examine the results yourself here: https://forum.arbitrum.foundation/t/ardc-research-deliverables/23438/9

Approach

At a high level, you can break this campaign into two high level components:

1. Create a benchmark to understand what % of the incentives can be attributed to the spend versus baseline. They call this a “Synthetic Control” methodology with some fancy math. This doesn’t matter too much except for whatever our final numbers are we need to adjust them down because not everything can be attributed to this single effort. You can read more about it in the original forum post.

2. Incentivise the end users of apps in different verticals on Arbitrum by giving them ARB tokens to juice their metrics. Three verticals were chosen (perps, DEXs, liquidity aggregators). Each app was given how to best spend the incentive.

I did find some interesting excerpts that I thought I’d include in here for your own judgement:

• “Many protocols missed several bi-weekly reports or didn’t post them at all. Around 35% of all STIP recipients didn’t post a final report.”

• “It was infrequent that protocols rigorously justified why they should be allocated a certain amount of incentives when applying for the STIP. Rather, the final allocations were generally a result of back-and-forth between protocols and the community, often resulting in an allocation based on something akin to “we feel like this ask is too big/small”.

Anywayyyyysss, moving on. I’ve included screenshots for the different categories, how much was spent, and the mechanism (no methodology screenshot for the DEXs but basically they just incentivised liquidity). The key thing to remember here is that 1 ARB is more or less $1. So yes, these are millions of dollars being distributed.

Results

I want to break up the results into two parts here because there are two things that this experiment aimed to understand.

1. The impact of these incentives on the apps

2. The impact of these incentives for sequencer revenue

We’re going to start our analysis with the first because that paints a slightly happier story. Well if we think from first principles, if someone gives you free money to run promotions for your business, what do you think will happen? Well, generally business will improve — for a while. That is what we saw across the board with this experiment.

Starting off with the Spot DEXs, their results seem pretty decent on the surface:

So basically what we’re seeing here is that anywhere from $2 to $24 was earned in TVL for every dollar spent which sounds good. However, we need to ask the real question here — how much of that was retained? This is where it becomes a bit tricky. Balancer’s TVL basically dropped after the rewards ran out, as evident through this chart:

However, Camelot, on the other hand, actually managed to retain this TVL! I’m not sure why these two protocols differed in their retention but if I was to take a guess it’d be the way they ran their incentive program and the types of users they attracted for the campaign itself. This is something I’ve bookmarked and will analyse myself in a future article.

Now that you understand some of the micro nuances, lets zoom out and understand how effective this was for the apps and the three top-line categories that matter (spot volume, perp volume, and loans). I present to you our key chart. I had to annotate on top of it to help it make sense of it all so stay with me as I explain it.

• I drew two red vertical lines to mark the start of the program and the end of the program. This will help us understand the timeframe we’re dealing with here.

• I then drew multiple horizontal lines to understand the different metrics and visualise how the program impacted those metrics over the course of its lifetime.

  • The first blue line basically shows that TVL spiked massively (no surprise) but then basically dropped to below where the program first began indicating virtually none of it was sticky!

  • The second line is spot volumes. I want to pause here and mention that unlike TVL which is supply side and costs nothing, spot volume represents demand. As we can see demand was constant at best but was actually lower by the time the program ended!

  • The third line is loans outstanding which is also demand driver and saw no change. While no lending protocols were incentivised I do find it as another strong metric of demand. This actually dropped throughout the program!

So what can we conclude from all of the above? Well basically Arbitrum spent $85m on all these other businesses to juice their supply side metrics (which clearly worked) but was rendered useless as there was no corresponding demand to soak up that TVL and tighter liquidity. In essence, you could say all of that money was lit on fire and given to mercenary farmers. At least certain protocols have higher TVL and a higher token price making some people richer in the process 😇

Speaking of demand side metrics, surely all this activity was good for the chain and led to higher revenues from all those transactions — right?!

Well, not exactly.

Actually, no, not at all.

So here’s the chart of sequencer revenues from Jan 2022 to July 2024. The big spike near April is when crypto started to go up massively and the Synthetic Control helps us account for this.

On the surface we can see revenue went up, hitting as high as $400k per day for certain months. Here’s a clearer chart that shows the impact just for Arbitrum and taking into account the Synthetic Control:

So what’s the area under the curve? $15.2m. If you remove the Synthetic Control you get a total of $35.1m in sequencer revenue in total. We’re still far from god here given $85m was spent!

Learnings

To summarise all of the above:

• Arbitrum decided to spend $85m to incentivise activity on its network to boost marketshare and revenue

• They did this by giving free tokens to apps/protocols that would distribute them to their end users

• Upon analysis, all of these free tokens were given to supply side drivers and virtually no change was shown on the demand side

• Looking even deeper, the sequencer revenue from all this activity was $60m less than the amount spent

What’s my takeaway from this? The first is supply side incentives are as good as burning money and should not be done unless you have a supply-side problem (usually not though, demand is the struggle).

The second, which is the premise around what I touched upon at the start of the article was: if you give money out to randos without discernment of who they are and where they come from, you will get what you pay for — which to clarify, is 💩. Protocols that continue to dump money to users without understand who they are, what their intents are will end as the business described at the start of this article.

Lets imagine that this incentive scheme discerned who these tokens were given to via a wallet’s permission-less identity and had criteria such as:

• Does this user actually use DEXs or is it a brand new wallet?

• What is the net-worth of this wallet and are they a potentially valuable wallet to acquire?

• How much has this wallet spent on fees? Are they stick on the platforms they use?

• Is this address currently using all the things that have upcoming tokens? They probably smell like a farmer.

What do you think the end result would be?

---

I believe that the work that my team and I are doing at 0xArc will solve these issues. We still have some key components to build but if you’re interested in learning more feel free to reach out.