Rewards Eligibility Oracle: A Two-Month Retrospective

A Brief Reminder of What REO Does

On August 25th, 2026, The Graph's Rewards Eligibility Oracle (REO) changed how rewards flow to participants on The Graph Subgraph Service. REO requires Indexers to serve queries and continuously improve their operations to meet higher quality of service standards for The Graph ecosystem's consumers. This post documents how those promises have held up so far.

For most of The Graph Network's history, indexing rewards followed curation signal and allocated stake. Neither one checks whether an Indexer served a query or if their service meets a specific quality of service threshold. In 2025, 15.2% of indexing rewards went to Indexers that were not serving queries or indexing data that consumers used. While REO was proposed in GIP-0079 in October 2023, it did not become a reality until August 2026. The Foundation's new mandate made it the first of three changes to network economics, alongside GRT Liquid Staking and Direct Indexer Payments.

REO's rules are straight-forward for Indexers to understand. An Indexer is eligible for rewards if it has at least 10 online days in a rolling 28-day window. These Indexers must serve data to a minimum threshold of Subgraphs with a qualifying query where a qualifying query returns HTTP 200, responds in under 5,000 ms, and comes from an Indexer within 50,000 blocks of chainhead. The oracle checks once a day at 10:00 UTC and renews each eligible Indexer on-chain for 14 days. An Indexer that fails keeps its stake and can keep serving queries and earning query fees. An Indexer cannot claim indexing rewards until it qualifies again. Total issuance on The Graph protocol does not change. REO only changes who receives it.

The original post also made four promises: publish the criteria before enforcing them, announce changes about 14 days ahead, make every Indexer's status public, and raise the bar over time. Over the past two months, this threshold has updated twice, redirecting 667,000 GRT that would have otherwise gone to under-performing Indexers to a smart contract that can be used to the protocol’s benefit later. This post reports what each step changed with the number of Indexers, GRT, and Delegators it affected.

The First Two Months

REO has had three eligibility steps in nine weeks, and each change after launch took effect 14 days after it was announced. Before enforcement, a testnet oracle on Arbitrum Sepolia scored the same mainnet query data against the same rules for seven months. On the 46 days when the testnet and mainnet oracles both posted, they renewed exactly the same Indexers. Here's what the timeline has looked like so far:

The other rules have not changed: 5 online days in 28, HTTP 200, under 5,000 ms, and within 50,000 blocks of chainhead. Every step was presented with its data at weekly Graph Office Hours calls over the course of the past two months.

Step 1: The REO Baseline (August 25-October 6)

When enforcement began on August 25th, 70 Indexers were active on The Graph's Subgraph Service. 51 of them passed REO. By October 5th, the last day under the 1-Subgraph rule, 55 Indexers were eligible¹, and no eligible Indexer had lost its eligibility. Metrics in this section cover August 25th to October 5th.

Indexers

• Eligible: 51 on August 25th and 55 on October 5th. The count never fell below 51, and 55 different Indexers were renewed during the step.
• Newly eligible: 4 Indexers were renewed for the first time.
• Failing: On October 6th, 14 of 69 active Indexers did not pass the REO threshold.

Individual cases show the change more clearly than the totals. Two Indexers went from serving on 11% of days to 100% and were renewed on August 24th (one day before enforcement). Three more went from serving on 0% of days to 79% and were first renewed on September 5th.

One event in this step was not REO. From the afternoon of September 24th, about 13 Indexers still received queries but rejected every one, and the daily count of serving Indexers fell from 51 to 40 by September 28th. The timing matches a change to the gateway's TAP sender address, but that is not confirmed.

The decline in eligible Indexers earlier in the year came before enforcement, and it came from Indexers leaving. Sixteen Indexers that passed in the first quarter were gone by July, and 10 of them withdrew all of their stake, about 267M GRT.

GRT

During Step 1, 52 Indexers collected 26.11M GRT in indexing rewards. The protocol minted another 2.83M GRT to the reclaim contract, 9.8% of indexing issuance for the period. Each reclaim records a reason on-chain.

Before this upgrade, rewards that could not be paid were never minted. Examples are rewards on an allocation closed with a zero Proof of Indexing (POI), or on a Subgraph whose rewards are denied. These rewards now go to the reclaim contract. This GRT may be reintroduced to the ecosystem through protocol governance. No decision has been made on how it will be used.

The REO row (with 658,987 GRT from 8 Indexers) marks where REO may have contributed. It does not prove that REO caused the loss, because an Indexer that goes offline also fails REO. REO does not reclaim on the day an Indexer fails. It blocks the claim and leaves the rewards on the allocation. If the Indexer passes again and presents a POI within 28 days, it still collects them. Only after that are those rewards reclaimed as a stale POI, so REO's effect on reclaims lags the eligibility data by about a month.

Separately, 5.92M GRT went to the Innovation Allocation approved in GIP-0089 during this step. That is a fixed share of issuance and does not come from reclaimed rewards.

Delegators

Delegators who moved stake away from failing Indexers were mostly leaving Indexers that had never qualified.

• Undelegated from failing Indexers: 28.37M GRT, by 54 Delegators. Of that, 28.19M came from Indexers the oracle has never renewed.
• One large exit: In the week after enforcement, one Delegator undelegated 25.69M GRT from Indexers that had no tokens allocated and had never been renewed. Without that Delegator, the step's outflow from failing Indexers was 2.68M GRT.
• New delegation to failing Indexers: 10 Delegators added 0.40M GRT.
• GRT delegated at the end of the step: 951.5M to passing Indexers and 71.5M to failing Indexers on October 6th, against 821.1M and 220.9M on August 25th. These figures leave out GRT that was already thawing, and part of the change comes from Indexers that started passing. Note that undelegating starts a 28-day thaw, so it is not the same as withdrawing.

Step 2: Moving to Five Eligible Subgraphs (October 6-October 27)

On September 22nd, The Graph Foundation announced that Indexers would need to serve qualifying queries on at least five Subgraphs a day for that day to count. The justification is to get Indexers to sync and reliably serve a range of Subgraphs rather than just a single one. Over the 21 days of Step 2, the number of eligible Indexers fell from 55 to 48. Figures in this section cover October 6th to October 26th.

Indexers

• Eligible: 55 on October 6th and 48 on October 26th.
• Failed a daily check: 12 Indexers dropped out of the oracle's daily renewals, nine of them on October 7th, the first run after the change. Seven of them lost eligibility once their 14-day renewal ran out. No Indexer became eligible for the first time.
• Failing: On October 27th, 21 of 69 active Indexers were not passing REO, up from 14 at the start of the step.
• Breadth of service: On each day of the step, between 40 and 46 Indexers served qualifying queries on at least five Subgraphs.

The Indexers that lost eligibility did not serve at least five Subgraphs on enough days, which is precisely what the change was designed to affect. Before the change, the model built on gateway data from September 4th to October 1st projected about 45 eligible Indexers at five Subgraphs.

GRT

During Step 2, 47 Indexers collected 13.36M GRT in indexing rewards, down from 52 Indexers in Step 1. Rewards paid per day held steady: 636K GRT a day, against 622K in Step 1. The protocol minted 1.47M GRT to the reclaim contract, 9.9% of indexing issuance for the period.

The REO row is small in this step for the same reason as before: reclaims lag eligibility. Rewards on the allocations of the 7 Indexers that lost eligibility in October show up as reclaims only after those allocations go 28 days without a POI. Separately, 3.49M GRT went to the Innovation Allocation.

Delegators

• Undelegated from failing Indexers: 538K GRT, by 32 Delegators. Of that, 380K came from Indexers the oracle has never renewed. Without the largest single Delegator, the outflow was 407K GRT.
• New delegation to failing Indexers: None.
• Pace: 10.7 Delegators a week left failing Indexers, against 9.0 a week in Step 1.
• GRT delegated at the end of the step: 956.8M to passing Indexers and 75.6M to failing Indexers on October 27th. The failing total rose from 71.5M because 7 Indexers, with their Delegators, moved into the failing group.

Step 3: Scaling to 10 Eligible Subgraphs (Started October 27)

From October 27th, an online day needs at least one qualifying query on each of at least 10 Subgraphs. Each daily run checks every day still in the 28-day window against the new minimum, including days before October 27th. Two days in, 47 Indexers are eligible. Figures in this section cover October 27th-28th with the eligible count as of October 29th.

Indexers

• Eligible: As of October 29th, 47 Indexers were eligible, 4 of them in grace, and the oracle renewed 43 that day.
• Failed a daily check: 3 Indexers dropped out of the daily renewals in the first two days.
• Breadth of service: 39 Indexers served qualifying queries on at least 10 Subgraphs on October 27th, and 38 on October 28th.

The eligible count trails the new rule by up to 14 days. An Indexer renewed before October 27th keeps its eligibility for the rest of its 14-day period, even if it fails the new rule. The full effect of this step shows by about November 10th. Before the change was announced, the model projected 42 eligible Indexers at 10 Subgraphs.

GRT

Two days is too short to measure the likely impact. In that time, 18 Indexers collected 1.25M GRT in indexing rewards, and 13,243 GRT was reclaimed, none of it tied to REO. In the model, the 42 Indexers projected to stay eligible at 10 Subgraphs would claim 92.7% of indexing rewards, against 94.3% at 5. The Indexers that serve the most Subgraphs also earn the most rewards, so the bar can rise without disrupting them.

Delegators

In the first two days, 2 Delegators undelegated 47K GRT from failing Indexers. On October 29th, 22 of 69 active Indexers were failing, with 79.0M GRT delegated to them.

What comes next

The team will measure how the network responds before the next change. One idea debated during Graph Office Hours on October 20th is to scale the minimum with each Indexer's total stake. For example, at 1 Subgraph per 10 million GRT with a floor -- meaning an Indexer with 100M GRT would need 10 Subgraphs. Nothing is decided yet on next steps, but the Foundation team continues to monitor and model what will. Any change goes into the criteria document with a justification and takes effect 14 days later.

How REO Improves Economic Efficiency for The Graph

REO has done what it set out to do: make the Subgraph Service more economically efficient. In two months, it has brought the supply side of the Subgraph market into closer balance with the demand side. Indexing rewards now follow the Indexers that serve the queries data consumers send.

Indexers that were underperforming have improved. Four Indexers qualified for the first time during Step 1, and three of them went from serving on 0% of days to 79%. Indexers that were not serving have been excluded from rewards, which is the intended outcome. They were collecting rewards without providing value to data consumers.

REO has also cut the GRT paid to Indexers that were not doing the work. About 667,000 GRT built up on the allocations of Indexers that failed REO. That GRT went to the reclaim contract instead of to those Indexers.

The threshold has risen twice in the nine weeks since its onset. Each step came with 14 days' notice and with its data presented at Graph Office Hours. To date, the community has not pushed back on that pace, but the Foundation checks in weekly about this pace at Graph Office Hours.

Delegators have also started to act. Almost all of the GRT undelegated from failing Indexers came from Indexers that never qualified. Even so, 79.0M GRT was still delegated to Indexers failing REO on October 29th. Delegators can check their Indexer's status on the REO dashboard and move stake away from Indexers that do not qualify.

REO is the first of three changes to network economics in the Foundation's mandate. GRT Liquid Staking lets delegation that the Foundation manages go toward protocol-aligned work. One example is directing delegation from the liquid staking vault to Indexers that support specific chains or community Subgraphs. More detail will follow in the coming months as the community aligns on these efforts.

Direct Indexer Payments (DIPs) take the next step, and the Foundation will outline them in the coming months. Today, indexing rewards are the main tool for steering Indexer behavior. With DIPs, recurring payments from data consumers will drive what Indexers index. That makes the protocol more sustainable over time.

Taken together, REO makes every GRT of issuance work harder. The protocol issues the same amount as before, and more of it now reaches Indexers that serve data consumers. That ties GRT more closely to real use of The Graph, and that kind of economic efficiency serves everyone who holds GRT.

¹ Eligible counts in this post include Indexers in grace, which are still eligible but due for renewal. The REO dashboard lists them separately. On October 20th, for example, the dashboard showed 46 eligible and 2 in grace, which this post counts as 48 eligible.

About The Graph

The Graph is a suite of blockchain data infrastructure products that extract, process, and deliver scalable blockchain data solutions across 60+ networks. The Graph enables application developers, data analysts, AI agents, and enterprise teams that need structured, real-time access to blockchain data. Products include Subgraphs, Firehose, Substreams, and Amp. As of early 2026, The Graph has served over 1.27 trillion queries to more than 75,000 projects, powered by a network of independent Indexers around the world.

Follow The Graph on X, LinkedIn, Instagram, and Reddit. Join the community on The Graph’s Telegram, join technical discussions on The Graph’s Discord.