The analytics page tracks how healthy, representative and decentralized Cardano governance is, one epoch at a time. Everything on it comes from on-chain data, refreshed several times a day. The page is organized in five chapters: governance today, representation and participation, accountability and behavior, decentralization, and governance throughput. The homepage carries a “Governance at a glance” strip with four of its headline figures, DReps holding delegated power, delegated voting power, DReps that voted in the recent window, and the share held by the top 10 DReps, each with a sparkline of the last epochs and a link into the chapter that explains it.
The two layers
DRep statistics on DRepTalk always separate two things. The representative layer covers real registered DReps: their count, their combined voting power, and how concentrated that power is. The default delegation layer covers the two predefined options, always abstain and always no confidence, which hold real voting weight but are not representatives. The analytics page shows them side by side without ever mixing them, so a rise in default delegation never reads as a change in DRep concentration. If the two options are new to you, start with what the default options do.
Reading the trends
Almost every trend chart starts in the same epoch, and the analytics page states that epoch once, in the note at the bottom. A chart that starts later says so on the chart itself. That start is not cosmetic: it marks the first epoch the metric can be measured reliably, and a chart refuses to show anything earlier rather than guessing.
Changes between epochs are shown as net change. If a chart shows 2,000 more delegators than the epoch before, that is the balance of everyone who arrived and everyone who left. On-chain epoch aggregates cannot tell those two groups apart, so the page never claims inflow or outflow.
The “voted in the last 12 epochs” figure counts DReps with at least one on-chain vote in that window, including votes that were later changed. Twelve epochs is roughly two months, long enough that a quiet stretch between governance actions does not make the whole network look inactive.
The silent counterpart uses the same window but a different base: it counts the DReps that held delegated power in an epoch and cast no vote in the twelve epochs up to it. The two charts are deliberately not each other’s mirror. A DRep that voted and has since retired still counts as having voted in the window, but it holds no power and so is never silent. The share chart next to them reads the silent count against the power holders of the same epoch, so it is the cleaner participation rate of the two.
The default options also get a share chart: the two predefined options together, as a percentage of everything delegated to them and to representative DReps. It is the per-epoch version of the comparison sentence under the two option cards.
The “DRep activity” section reads the current registrations rather than a trend. Two bars split all registered DReps by registration state, active against inactive or retired, and by voting history, ever voted on chain against never voted, with a note on how many active registrations hold delegated power. Below them the section states how much stake is still delegated to inactive DReps, how much of it the ten largest hold, and lists the largest inactive DReps by voting power.
Voting concentration in practice
The effective-representation panel and the Positions tab of a governance action also read how concentrated the cast votes were. These numbers describe exercised power, the voting power that actually voted, not the distribution of all delegated power. The half-count says how few of the largest voters together cast at least half of the voted power. The action page adds the largest voter’s share, the combined top-5 share, and, where the action has an approval threshold, how many of the largest voters together held as much power as the threshold required in yes votes. That reading is arithmetic on cast votes, it does not claim those voters coordinated or voted the same way.
These stats only appear when the voting power behind every single vote on the action is recorded. An action with incomplete per-vote power shows no concentration numbers at all rather than a misleading partial sum.
Changed votes
A DRep can re-vote on an action at any time while voting is open, and the analytics page tracks what those re-votes actually did. A re-vote only counts as a changed vote when the voter’s final position differs from their first recorded one. Many re-votes keep the position and only update the attached rationale, and those are shown separately rather than inflated into change numbers. The panel also shows where changed votes moved, to yes, to no or to abstain, and which decided actions were reconsidered the most.
Only actions whose complete vote history has been swept from the chain are counted, and the panel says how many are still queued. The same honest split appears on each action’s Votes tab.
Vote rationales
Every on-chain vote can carry a metadata document explaining the decision, the rationale. The rationales panel measures how many DRep votes on decided actions carry that link, both as a share of votes and weighted by the voting power behind them. A linked document is not always retrievable, so these figures can read higher than the readable rationales on an action’s own page. It also shows which decided actions were best and least covered, and how many rationales arrived only later, through a re-vote on the same action.
Coverage is a presence check, nothing more. A vote with a rationale is not automatically better reasoned than one without, and the panel makes no attempt to judge content. The power-weighted figure only counts actions where the voting power behind every single vote is recorded, and says how many actions that excludes. The best and least covered lists only rank actions with at least 20 votes, and the panel says how many smaller ones were left out.
Voting timing
The timing panel reads when votes arrive. It counts votes, not weight: every vote counts once, whatever voting power stands behind it, so the figures describe the behavior of the voters rather than the movement of the stake. The headline numbers are the median day after submission for DRep votes and for SPO votes, and the day by which half of an action’s DRep votes had arrived, taken as a median across decided actions. A vote can only be timed when both the action’s submission and the vote’s block time are on record. A DRep who changed a vote is timed at the final vote. The medians only cover decided actions, an action still open can only hold its early votes so far.
The window split reads every DRep vote against its own action’s voting window instead of the calendar. The span from submission to the close of voting is cut into three equal parts, and each vote falls in the early, middle or late third of that action’s window, so an action open for twenty days and one open for five are read on the same scale. Votes recorded after the window closed are counted separately and never as late, because they never had a third to fall into. The per-type table only lists types with at least 20 timed DRep votes, and a type’s median SPO day needs 20 timed SPO votes of its own, otherwise that cell shows n/a.
None of this is a quality measure. Timing describes when votes arrive, not how carefully they were cast. An early vote is not more diligent than a late one, and a late vote can be the result of waiting for a discussion to run its course.
The Constitutional Committee
The committee panel shows how the Constitutional Committee participates: the median share of members voting on decided actions, how often the committee split rather than voting one way, how many actions finished below its approval threshold, and each member’s participation by name. Eligibility follows the committee’s actual membership at each action’s tally epoch, so resignations, term expirations and hot-key rotations are accounted for, and only a member’s final vote on an action counts.
A committee vote against an action expresses that the member did not find it consistent with the constitution in force at the time. The panel reports these outcomes as numbers and takes no position on any individual judgment.
SPO participation
Stake pool operators vote alongside DReps and the Constitutional Committee, but only on a subset of governance actions: hard forks, motions of no confidence, committee changes and security-relevant parameter changes. The panel counts how many decided actions fall in that eligible set, then reports two figures over them. Median SPO turnout weighs each pool’s stake against the stake of every pool that could vote, the pools permanently set to abstain included and the No side counted the way the ledger counts it, the same denominator the on-chain ratification check uses. The second figure counts how often the SPO verdict and the DRep verdict on the same action fell on opposite sides of their respective thresholds. The actions behind that count are listed under the tiles, each one naming the body whose tally stayed below its own threshold.
A pool is on a default stance when its reward account delegates to always abstain or always no confidence: its stake counts in every SPO tally without a vote of its own, and by now a large part of all pool stake sits there. The panel therefore adds a second turnout reading that leaves that stake out of the denominator, the turnout among the pools that could still take a position, and a tile with the share of pool stake on a default stance as of the most recently decided eligible action. The eligible set is small and stays small, so every action in it is listed with both readings, newest decision first, which is where a surprising median can be traced to the actions behind it.
Both figures only cover actions with a complete reading. An action whose stake tally is missing or unparseable is left out of the turnout median rather than folded in as a guess, and the panel states how many actions that affected.
An enacted action counts as meeting both bodies’ thresholds because the chain could not have enacted it otherwise. The stored SPO percentage for every action type other than hard forks still reflects the reading from before the Plomin hard fork, where a pool that did not vote counted as No rather than Abstain, so it can understate a real pass, which is why the chain outcome takes precedence over that stored figure.
Concentration of delegated power
The lower part of the page looks at how concentrated the delegated voting power itself is, independent of who actually votes. The Gini coefficient summarizes how unevenly power is spread across DReps, from 0 for a perfectly even spread to values near 1 when a few DReps hold most of it. The top-10 share and the minimum-coalition counts make the same idea concrete: how much the ten largest DReps hold, and how few of the largest DReps together reach half or two thirds of all delegated power.
The minimum coalition table applies the live approval thresholds to the current distribution: for each threshold it shows how few of the largest DReps together hold that share of the delegated power, and which action types the threshold gates. This is arithmetic on delegated power, not a claim that those DReps coordinate or vote at all. How the power that actually voted concentrates is a separate reading, described under voting concentration above.
Governance throughput
The last chapter steps back from any single vote and looks at the pipeline of governance actions as a whole: how many are submitted, how they conclude, and how long a conclusion takes. An action’s outcome is one of enacted, expired, closed, or dropped. Closed is the outcome reserved for info actions, which by design never enact anything, they conclude once the vote has run its course. Dropped covers actions removed from the proposal set before a decision was reached, which happens when another action building on the same predecessor is enacted first and leaves this one no path to enactment. Because a bare count of those would raise more questions than it answers, the page lists the dropped actions by name, each with the enacted action that took its place where the chain records one. Actions still open have not reached any of these outcomes yet.
Timing is measured in epochs, not calendar days, because both submission and decision are recorded per epoch on chain. The overall median and the per-type breakdown only include actions with a known submission epoch, and a type needs at least three such actions before its own median is shown, otherwise the table reports its counts without one.
Where the numbers come from
DRepTalk records one row of governance aggregates per epoch, built from the same chain data that powers the rest of the site. Voting power snapshots come from the per-epoch stake distribution, vote counts from the on-chain votes themselves, and the concentration figures from the full distribution of delegated power across DReps, with the two default options excluded. The effective-representation panel measures each action against its decision epoch, while the tally bar on an action page uses the epoch of the latest tally, so the two can sit one epoch apart for the same action. The threshold marker on the full-stake bar sits on the ratification denominator, the yes votes plus the No side as the ledger counts it, which already includes the always-no-confidence weight, so the marker is the same bar the chain measures against.
Frequently asked questions
Why do some charts start later than others?
Each metric shows data from the first epoch it can be measured reliably. Voting power can be reconstructed from chain history, but delegator counts only exist from the point DRepTalk started observing them live, so that chart starts later instead of pretending older data exists.
Why does the current epoch look provisional?
An epoch’s vote activity is only final once the epoch has ended. The page treats the running epoch as incomplete by design and finalizes its numbers right after the epoch rolls over.
What is the difference between active and with voting power?
Active follows the on-chain registration state. A DRep can be active with zero delegated stake, and stake can still sit with a DRep whose registration has lapsed. The activity section shows both layers separately.
Why do some actions show no concentration numbers?
The per-vote voting power for at least one vote on that action is not recorded. Concentration stats are computed only over complete data, a partial reading would understate how concentrated the vote really was.
What counts as a changed vote?
Only a re-vote whose final position differs from the voter’s first recorded one. Re-votes that keep the position, for example to attach or update a rationale, are counted separately and never shown as changed votes.
Does a missing rationale mean a DRep voted carelessly?
No. The rationale figures measure whether a vote carries a published explanation, nothing about its reasoning. Some DReps explain their votes in other places, and a vote without an attached rationale can be as considered as any other.
Could the largest DReps really decide an action on their own?
The minimum coalition numbers are arithmetic on the current delegated voting power: those DReps together hold enough weight to reach the threshold’s share. Nothing suggests they coordinate, and in practice votes come from a much broader set, which the voting concentration and effective representation panels measure.
Why do committee counts differ from the seat count?
Membership changes over time. A member who resigned, whose term expired, or who had not yet registered a voting key does not count as eligible for an action decided in that period, so each action is measured against the committee as it stood at the time.