Attracting Delegation
Delegators compare validators on historical uptime, on-chain trust and consensus scores, take rate, and how much the operator communicates. A new validator has none of the first, so it competes on the last two while the record accumulates — then earns the right to compete on performance once it has three to six months of clean data.
Visibility is most of the rest. Operators who post performance updates, answer questions and publish their scoring methodology attract attention that silent addresses do not, and the delegators they attract are the ones who stay through a bad month because they understand what they bought.
Uptime and Reliability
Uptime is the metric delegators weigh most, and rightly: a missed epoch earns nothing for your stake or theirs. A validator at 98% is meaningfully more attractive than one at 90%, even if the second scores better in the epochs it does show up for.
Publish the history. Weight-submission frequency is verifiable by anyone in the extrinsic explorer, but almost nobody will do the analysis, so do it for them: monthly uptime, incident post-mortems, and your redundancy architecture. Operators who explain an outage and what they changed afterwards routinely gain delegation from an incident that would have cost a silent operator stake.
Transparency
Publish a validator information page: the subnets you validate, your scoring approach on each, your hardware and failover setup, your take and why, your uptime history, and a way to reach you. Trust of that kind is slow to build and is what keeps delegators through a dip they do not have to interpret alone.
Take Rate Strategy
Your take is bounded by the chain, not chosen freely: MaxDelegateTake is 11796/65535 — 18% — and has been capped there since runtime v320. Below that ceiling it is a positioning decision. A rate under the field’s makes you the cheap option while you have no record to sell; a rate at the ceiling needs evidence behind it, because a delegator paying more is buying something and will want to see what.
Move it slowly and announce it early. Surprise increases are the fastest way to trigger an exodus, and the delegators most likely to leave are the ones who chose you on price. Give at least two weeks’ notice, say why, and expect questions.
Scaling Past One Subnet
A second subnet diversifies revenue and widens the delegation you can attract — delegators staking across several subnets prefer one operator across all of them. Each needs its own hotkey and process; the binding constraint is scoring compute and attention, not server capacity, so pair a GPU-bound subnet with a CPU-bound one rather than two of the same.
Where a subnet lets validators design their own evaluation, that pipeline is the moat. Everyone running the default produces near-identical vectors and near-identical consensus alignment; a genuinely better judge model, curated reference data, or multi-signal scoring puts you closer to the ground truth, which is what consensus rewards. The improvement compounds — better alignment means more influence, which makes the next improvement worth more.
Scale the operations alongside it: active-passive failover so no single failure costs an epoch, infrastructure-as-code so any box can be rebuilt in an hour, centralized alerting on dropped scores, and a rollback path for updates that degrade scoring. And stay in the governance conversation — protocol changes to consensus, registration or emission allocation land on validator economics first, and the operators who track them upgrade ahead of the deadline instead of after it.