We're about to start buying guest post placements from our own curated marketplace shelf, pointed at this website, to measure how the product performs before any customer can buy it. This post is the measurement protocol, published before the first order, so that when the results appear you can check them against rules that existed in advance. If we have to change anything, it goes in the changelog at the bottom, dated, with the reason.
Why publish first? Because results-first studies get to quietly choose what to measure after seeing the data. Committing now means a failed placement is as much a finding as a successful one. In an industry where every seller reports 100% success, our failures may end up the most useful numbers we publish.
What we're buying
Ten to twelve guest post placements (a new article containing one dofollow link, written by us, approved by our reviewer) on sites from our vetted shelf, concentrated in the DR 40 to 60 range, across at least two niches, all pointing at pages on rank.fast. Suppliers are not named, per our standing policy; every placement's measured outcome is published, per the same policy.
The two evidence streams, kept separate
- Stream A, the placement monitor (ours). Per placement: was it delivered as ordered; is the link live, dofollow, anchor intact; is the hosting page indexed in Google. Checked at delivery, then re-verified at 30 and 90 days ("stickiness"). This is the stream our future guarantee would cover, so it's the one the pilot exists to test.
- Stream B, Search Console observation. For each target page on our site: trailing 28-day impressions, clicks, and average position, snapshotted as a baseline at order time and compared at 28 days after the placement is indexed.
What we record per purchase, at order time
- Order date, price paid, DR band and quality grade of the host site, niche, target page on our site, anchor text
- Baseline: the target page's trailing 28-day Search Console numbers, and the host page's index state before placement
- The comparison windows above, fixed in advance
What we'll conclude, and what we won't
The pilot's conclusions will be of the class "this placement was delivered, indexed in N days, and survived to its checkpoints": delivery rate, time-to-live, time-to-indexed, and stickiness, with the full per-placement table published.
What we will not claim: that any placement caused a ranking change. Ten links pointing at a site that is simultaneously publishing content, rewriting pages, and growing cannot support causal attribution, and we won't pretend otherwise. Stream B movements will be reported as labeled observations, nothing more. Anyone who tells you their twelve-link study proved a ranking effect is telling you about their marketing, not their measurement.
What would count as failure
We're stating the bad outcomes in advance: placements that never go live, links delivered nofollow or with altered anchors, host pages that never enter Google's index or drop out by a checkpoint, and sites that take the money and vanish. Every one of these, if it happens, appears in the published table with the same prominence as the successes. The failure rate is the number our future guarantee gets priced on, so we have no incentive to hide it and every incentive to know it precisely.
When
Orders begin when the pilot budget is committed; results publish as they accrue on the live log and in a full writeup when the 30-day checkpoints complete. Pricing and any guarantee for the marketplace publish only after that writeup, built on these numbers.
Where the results will live
Every placement gets a row in a public, machine-readable log at
/data/pilot-log.json:
anonymized id, order timestamp, DR band and quality grade, niche, target
page, the order-time baseline, the exact date it was delivered, the exact
date it entered Google's index, and its 30/90-day checkpoint results.
Auditable end to end without identifying any host site or supplier.
Changelog
2026-08-24: added the public per-placement log
format at /data/pilot-log.json, following external review
feedback that order-time baselines and exact indexed dates should be
prominent and auditable. No purchases made yet.
2026-08-23: protocol published. No purchases made yet.