Curators are critical to the Graph decentralized economy. They use their knowledge of the web3 ecosystem to assess and signal on the subgraphs that should be indexed by The Graph Network. Through the Explorer, curators are able to view network data to make signaling decisions. The Graph Network rewards curators who signal on good quality subgraphs with a share of the query fees that subgraphs generate. Curators are economically incentivized to signal early. These cues from curators are important for Indexers, who can then process or index the data from these signaled subgraphs.
When signaling, curators can decide to signal on a specific version of the subgraph or to signal using auto-migrate. When signaling using auto-migrate, a curator’s shares will always be upgraded to the latest version published by the developer. If you decide to signal on a specific version instead, shares will always stay on this specific version.
Remember that curation is risky. Please do your diligence to make sure you curate on subgraphs you trust. Creating a subgraph is permissionless, so people can create subgraphs and call them any name they'd like. For more guidance on curation risks, check out The Graph Academy's Curation Guide.
First, we take a step back. Each subgraph has a bonding curve on which curation shares are minted when a user adds signal into the curve. Each subgraph’s bonding curve is unique. The bonding curves are architected so that the price to mint a curation share on a subgraph increases linearly, over the number of shares minted.
As a result, price increases linearly, meaning that it will get more expensive to purchase a share over time. Here’s an example of what we mean, see the bonding curve below:
Consider we have two curators that mint shares for a subgraph:
- Curator A is the first to signal on the subgraph. By adding 120,000 GRT into the curve, they are able to mint 2000 shares.
- Curator B’s signal is on the subgraph at some point in time later. To receive the same amount of shares as Curator A, they would have to add 360,000 GRT into the curve.
- Since both curators hold half the total of curation shares, they would receive an equal amount of curator royalties.
- If any of the curators were now to burn their 2000 curation shares, they would receive 360,000 GRT.
- The remaining curator would now receive all the curator royalties for that subgraph. If they were to burn their shares to withdraw GRT, they would receive 120,000 GRT.
- TLDR: The GRT valuation of curation shares is determined by the bonding curve and can be volatile. There is potential to incur big losses. Signaling early means you put in less GRT for each share. By extension, this means you earn more curator royalties per GRT than later curators for the same subgraph.
In general, a bonding curve is a mathematical curve that defines the relationship between token supply and asset price. In the specific case of subgraph curation, the price of each subgraph share increases with each token invested and the price of each share decreases with each token sold.
In the case of The Graph, Bancor’s implementation of a bonding curve formula is leveraged.
Now that we’ve covered the basics about how the bonding curve works, this is how you will proceed to signal on a subgraph. Within the Curator tab on the Graph Explorer, curators will be able to signal and unsignal on certain subgraphs based on network stats. For a step-by-step overview of how to do this in the Explorer, click here.
A curator can choose to signal on a specific subgraph version, or they can choose to have their signal automatically migrate to the newest production build of that subgraph. Both are valid strategies and come with their own pros and cons.
Signaling on a specific version is especially useful when one subgraph is used by multiple dApps. One dApp might need to regularly update the subgraph with new features. Another dApp might prefer to use an older, well-tested subgraph version. Upon initial curation, a 1% standard tax is incurred.
Having your signal automatically migrate to the newest production build can be valuable to ensure you keep accruing query fees. Every time you curate, a 1% curation tax is incurred. You will also pay a 0.5% curation tax on every migration. Subgraph developers are discouraged from frequently publishing new versions - they have to pay a 0.5% curation tax on all auto-migrated curation shares.
Note: The first address to signal a particular subgraph is considered the first curator and will have to do much more gas-intensive work than the rest of the following curators because the first curator initializes the curation share tokens, initializes the bonding curve, and also transfers tokens into the Graph proxy.
For end consumers to be able to query a subgraph, the subgraph must first be indexed. Indexing is a process where files, data, and metadata are looked at, cataloged, and then indexed so that results can be found faster. In order for a subgraph’s data to be searchable, it needs to be organized.
And so, if Indexers had to guess which subgraphs they should index, there would be a low chance that they would earn robust query fees because they’d have no way of validating which subgraphs are good quality. Enter curation.
Curators make The Graph network efficient and signaling is the process that curators use to let Indexers know that a subgraph is good to index, where GRT is added to a bonding curve for a subgraph. Indexers can inherently trust the signal from a curator because upon signaling, curators mint a curation share for the subgraph, entitling them to a portion of future query fees that the subgraph drives. Curator signal is represented as ERC20 tokens called Graph Curation Shares (GCS). Curators that want to earn more query fees should signal their GRT to subgraphs that they predict will generate a strong flow of fees to the network. Curators cannot be slashed for bad behavior, but there is a deposit tax on Curators to disincentivize poor decision-making that could harm the integrity of the network. Curators also earn fewer query fees if they choose to curate on a low-quality Subgraph since there will be fewer queries to process or fewer Indexers to process those queries. See the diagram below!
Indexers can find subgraphs to index based on curation signals they see in The Graph Explorer (screenshot below).
- The query market is inherently young at The Graph and there is risk that your %APY may be lower than you expect due to nascent market dynamics.
- Curation Fee - when a curator signals GRT on a subgraph, they incur a 1% curation tax. This fee is burned and the rest is deposited into the reserve supply of the bonding curve.
- When curators burn their shares to withdraw GRT, the GRT valuation of the remaining shares will be reduced. Be aware that in some cases, curators may decide to burn their shares all at once. This situation may be common if a dApp developer stops versioning/improving and querying their subgraph or if a subgraph fails. As a result, remaining curators might only be able to withdraw a fraction of their initial GRT. For a network role with a lower risk profile, see Delegators.
- A subgraph can fail due to a bug. A failed subgraph does not accrue query fees. As a result, you’ll have to wait until the developer fixes the bug and deploys a new version.
- If you are subscribed to the newest version of a subgraph, your shares will auto-migrate to that new version. This will incur a 0.5% curation tax.
- If you have signaled on a specific subgraph version and it fails, you will have to manually burn your curation shares. Note that you may receive more or less GRT than you initially deposited into the curation curve, which is a risk associated with being a curator. You can then signal on the new subgraph version, thus incurring a 1% curation tax.
By signalling on a subgraph, you will earn a share of all the query fees that this subgraph generates. 10% of all query fees goes to the Curators pro-rata to their curation shares. This 10% is subject to governance.
Finding high-quality subgraphs is a complex task, but it can be approached in many different ways. As a Curator, you want to look for trustworthy subgraphs that are driving query volume. A trustworthy subgraph may be valuable if it is complete, accurate, and supports a dApp’s data needs. A poorly architected subgraph might need to be revised or re-published, and can also end up failing. It is critical for Curators to review a subgraph’s architecture or code in order to assess if a subgraph is valuable. As a result:
- Curators can use their understanding of a network to try and predict how an individual subgraph may generate a higher or lower query volume in the future
- Curators should also understand the metrics that are available through The Graph Explorer. Metrics like past query volume and who the subgraph developer is can help determine whether or not a subgraph is worth signalling on.
Migrating your curation shares to a new subgraph version incurs a curation tax of 1%. Curators can choose to subscribe to the newest version of a subgraph. When curator shares get auto-migrated to a new version, Curators will also pay half curation tax, ie. 0.5%, because upgrading subgraphs is an on-chain action that costs gas.
It’s suggested that you don’t upgrade your subgraphs too frequently. See the question above for more details.
Curation shares cannot be "bought" or "sold" like other ERC20 tokens that you may be familiar with. They can only be minted (created) or burned (destroyed) along the bonding curve for a particular subgraph. The amount of GRT needed to mint a new signal, and the amount of GRT you receive when you burn your existing signal are determined by that bonding curve. As a Curator, you need to know that when you burn your curation shares to withdraw GRT, you can end up with more or less GRT than you initially deposited.
Still confused? Check out our Curation video guide below: