Know what you’re looking at.

Inclusion & publication

A candidate needs an identifiable creator and a public sponsorship proposal or campaign. Automatic discovery adds candidates to a private queue. Emily reviews the source, campaign identity, and status before publication. Proposals are included when clearly labeled.

The initial research contains 23 candidates, recorded on September 16, 2026. Where those records appear, they are labeled historical research snapshots. Their original amounts and caveats are preserved. They are not evidence that sales remain open today.

Status is separate from success

Open

A source supports that placements are available, as of the stated date.

Sold out

The listed inventory was fully allocated. This alone does not prove payment.

Proposed

A sponsorship idea has been posted, without confirmed active sales.

Closed

The campaign has ended or the creator has closed sales.

Needs review

There is insufficient evidence of the current sales status.

A reachable website never establishes open sales by itself. Status dates appear on campaign pages.

Amounts mean different things

Raised totals come from campaign websites. X supplies original-post impressions, followers, engagement, and related updates. Creator claims remain in the source history and do not replace website totals. Sold placements can still carry a committed amount: allocation and payment are separate facts.

Paid

An amount explicitly identified as paid. Its evidence label still matters.

Winning bids

The current or final winning bids, which may not have settled.

Committed / reserved

A promise or reservation, distinct from a payment.

Creator reported

The creator’s stated result where settlement details are not established.

Site displayed

A website counter whose precise financial meaning is not established.

Votes

A noncash measure, excluded from money comparisons.

These measures are never added together. The directory and Trends show USD estimates beside original amounts, with labels that keep payments, bids, pledges, and reported totals separate. Filter by result type for a closer comparison. Refunds, decreases, changed goals, or conflicting values require review. Claimed placements are distinguished from confirmed sales.

Currency estimates

EUR and CAD amounts use European Central Bank reference rates supplied through Frankfurter. The rate date appears beside the tables; converted amounts carry an ≈ and retain the original figure. These are estimates at the stated reference rate, not exchange rates on the date of payment. Rates are checked daily, and the last valid rate is retained if a check fails. Noncash votes are never converted. Historical case studies retain their original currencies. Exports include the original amount, USD estimate, rate, date, and source.

Post dates, impressions & updates

Launch dates link to the original campaign tweet where it is available. An update post does not establish a launch date; “first found” identifies an earliest known post without claiming it was the launch. Missing dates remain unknown when no original can be located or retrieved. Impressions are the original tweet’s public view counter, not unique viewers or total account reach.

Campaign updates come from the creator’s own posts. We look for links to the campaign website, connections to the launch post, or campaign-specific topics alongside sponsorship language. Unrelated tweets are filtered out. This rule-based matching can miss vague updates, and a new campaign post does not by itself change the amount raised.

Evidence & missing data

Observations are labeled public payment record, creator report, website display, or unknown. A website counter is a statement from that website, not independently verified revenue. A creator’s claim of payment does not become a public payment record.

“Not reported” means unknown. A historical zero means a source showed zero at its recorded time. It is not today’s zero. Failed requests, loading placeholders, missing posts, and unavailable values never become zero.

Freshness

Social data uses public counters observed via twitterapi.io, an independent third-party provider. These are not X’s official metrics. An official X API adapter can be selected instead; each observation records its actual provider.

For connected, supported sources, active campaigns are checked every 15 minutes. Inexpensive structured sources may be checked every five minutes. Campaign-related original posts are checked hourly. Profiles and closed campaigns are checked daily. Discovery runs hourly; an open dashboard reads the cache every 30 seconds.

Each source carries its own observation time. If collection fails, we retain the last valid observation and show that the update is delayed. A recent post check does not make an older amount current. Provider rate limits, monthly quotas, or unsupported websites can delay updates.

The project targets a $95 monthly operating budget for up to 50 active campaigns. The seven-day pilot must verify the cadence and cost before live availability is promised. Public visits do not trigger paid upstream requests.

What the comparisons can tell us

Diminishing returns is a hypothesis. A small curated sample, different audiences and prices, and uneven campaign ages do not isolate novelty as a cause. These data also do not measure advertiser ROI.

24-hour and 48-hour comparisons require a precise launch timestamp and a successful source check within 15 minutes of the chosen age. An earlier unchanged value is used only when that check confirms coverage. Historical cumulative totals cannot fill those gaps. Progress uses the funding target recorded at that time, and current follower counts retain their own timestamp.

Earlier advertising cases establish precedent. They are not combined with current campaigns as a financial cohort. Historical outcomes retain their original source’s definition.

Sources & corrections

Campaign pages link to original posts, campaign websites, observations, and research. CSV exports preserve values, definitions, evidence, sources, and timestamps. Chart SVG exports contain the plotted observations in their metadata.

Corrections create new observations and an editorial audit record. Existing historical observations remain available. Source failures and conflicting results enter the review queue; automated collection cannot publish an unsupported success claim.

Collection references: X public metrics · Cloudflare Workers