Changelog
What changed, and when
Every rule on this site has a date. When we change how sample is screened, how a reward is held, or what goes into a delivery, the change is recorded here with the reason behind it, including the changes that made us look worse.
Publishing a method is easy. Publishing its history is the part that can be checked — it is what separates a set of rules from a page of marketing that happens to be phrased as rules. If a claim on this site is older than you expected, or newer, this is where you find out.
The record
Supply integration contract published in full
The entry link format, the four terminal states, how callbacks are signed and how settlement stays idempotent are now a public page rather than a document released after an NDA.
The old order wasted the time of every partner who read the document and then discovered their platform could not sign a callback. Publishing it moves that discovery to before the first email.
Screen-out compensation stated as a rule, not a courtesy
Participants are paid a fixed amount for a screen-out or a full quota. This was already how we ran; it is now written into the participant pages as something we can be held to.
Someone who reads four screening questions carefully and turns out to own the wrong car spent real minutes doing it. Pay nothing for those minutes and you teach that person to guess at whatever answer sounds like it qualifies — which is how a panel degrades its own data.
Rewards and payouts · Join the OpinionReady panel · Participant support
Inattention separated from fraud in quality scoring
A response that was rushed or straight-lined is scored differently from one produced by a script across many accounts. Different deductions, different appeal routes, and only the second affects withdrawal review.
The distinction is the one the Insights Association draws in its own data integrity work, and collapsing the two into a single word is how honest members end up treated as fraudsters over a bad afternoon.
Unverified figures render as description, not numbers
Any statistic the platform cannot yet export is shown as a qualitative phrase instead of a digit. The only figures printed as numbers are the ones we could produce a report for today.
A number nobody can reproduce is worth less than an honest sentence saying we do not have it yet, and a buyer who asks for the backing report and gets nothing will discount everything else on the page.
Research-grade sample · Who is in the panel · Data quality and the rejection log · Method and insights
Rejection log named as part of every delivery
Deliveries include the respondents removed, the rule that removed each one, and when it fired — at the row level, reconcilable against the buyer’s own counts.
Sample is bought on trust and inspected almost never. A removal you cannot see is indistinguishable from a removal that did not happen.
What counts as a change
Anything that alters what we owe a buyer or a participant. A new screening check, a change to what holds a reward, a different rule for what goes into a delivery file, a figure moving from description to a real number.
Not in here: wording, layout, and the ordinary work of making a page clearer. A log that records every comma becomes a log nobody reads, which defeats the point of keeping one.
If a change affected you
Rules are not applied backwards. A removal or a held reward is judged by the rules in force on the day it happened, and the date on each entry is what tells you which set that was. If you think a decision used the wrong version, say so — that is a specific enough claim for someone to check, which is the whole reason these dates exist.
Buyers with a question about a delivery can reach the sample team. Participants disputing a reward should use the route on the rewards page.