Skip to main content
Finished work has two kinds of results: observations, such as a page going live or being cited, and a measured change on the questions it targets. A citation alone never means the work is paying off. Home’s week includes these observations and measured outcomes, with a link to the work behind each result.

Steps by kind of work

Each kind of work tracks its own signals. A correction request tracks citations on the source site and its measured answer outcome. A site fix tracks its live check and AI crawler visits. It has no citation step. A product page’s markup fix is a site fix. Work to get your products onto a seller tracks whether AI then shows them sold there. A title and description change tracks the page going live, Google rankings and Search Console, followed by its measured answer outcome. The API serves these in order as results.steps, including signals that are pending, unchanged or unavailable. There is no requirement to reach one step before another. An unavailable signal keeps its reason, such as no linked questions, no tracked keywords or a missing connection.

Site fixes

A site fix is settled by the first daily update or weekly report that checks the fix on your live site. If the check passes, the action is marked Fix is live. If it fails, it is marked Not live. Heralded waits up to seven updates or reports for a check it can use. If none comes, the action settles as Done · not measured. A product page’s markup fix is settled by the first weekly report whose catalogue read checks that page after the fix. It is Fix is live when the page states every field the fix named as missing, and Not live when it still lacks one. A field a later report finds missing opens new work and never changes what the delivered fix is checked against. That holds when you did the work yourself: Heralded keeps the fields you were shown when you marked it done. Two site fixes are judged like work aimed at questions instead, over 14 days: letting AI crawlers read a page, and redirecting or restoring a dead page.

Seller listings

Work to get your products onto a seller is settled by what the seller’s cards show. It ends Your products are shown sold there when an answer observed after you marked it done shows one of your products with that seller. It ends Not shown sold there when seven updates or reports pass without one, and at least one answer since had its product cards read, even if it showed none. If every capture failed or an engine reads none, it ends Done · not measured. It credits no points, because it carries no target questions.

Work aimed at questions

Content, rewrites and outreach are aimed at questions. Heralded measures the metric and engines recorded in their targets. Work without explicit targets uses Mentioned for its linked questions.
  1. Baseline. Heralded takes the 14 days that end a week before the action was filed. The week in between is left out, because that is usually the dip that prompted the work. If this baseline has too few answers for the test, Heralded uses answers from your first read up to delivery, on the same questions and engines. The same minimum still applies. The verification record identifies the baseline as standard or since_first_read.
  2. After. It pools the answers after the work was delivered: 28 days for new content, competitor gaps and page rewrites, and 14 days for everything else, including title and description rewrites, outreach and correction requests. Work on a page also waits, up to 28 days, for Heralded to see the page change. A page changes when what it says changes: its text, headings, description, image descriptions or where a link leads. A link or image address that differs on every visit, such as a tracking code, is not a change. If two checks in a row find the page removed, Heralded stops checking it and the work ends Inconclusive because the page was removed. A page that is missing once and comes back changed is measured from that change.
  3. Control. It measures the same change on questions no other work touched, and subtracts it. A market that rose for everyone doesn’t get credited to your work.
  4. Test. What is left must pass a paired test, so its 95% range has to clear zero.
Heralded needs at least three targeted questions and three untouched questions to judge. With fewer, the engines’ normal variation would swamp any move one piece of work could make. Both groups also need enough paired units for an interval. If too few answers pair in either group, the outcome is Inconclusive.

Measured outcomes

Citation evidence stays visible through no-effect, falling and inconclusive measurements. A passing site check means the fix is live. It does not establish an answer gain and never makes the work paying off.

What happened since

Delivered work also gets a trail: dated observations from data Heralded already collects. The trail starts when the work is delivered, and nothing seen before delivery counts. Each daily update and weekly report looks again at every signal that hasn’t shown anything yet. What it watches depends on the kind of work. Each signal is in one of four states:
  • Good. It happened, dated when it was observed.
  • No change yet. Heralded looked, and it hasn’t happened.
  • Couldn’t check. Heralded couldn’t look, and says why, such as that Search Console isn’t connected, Google’s results couldn’t be read, every answer since delivery failed or the page was removed. This is not the same as no change.
  • Pending. Nothing to say yet, such as Search Console before its first window after delivery.
Each finding is recorded once, on the action’s activity. A signal is recorded again only when what it shows changes, and is no longer checked once it is good. A published piece’s live and cited lines are its publication and its first citation, which its activity already shows. Work reopened and delivered again starts a new trail. The measured result, from the tests above, is the strongest entry. The trail credits no gain: it shows what happened, and only the measured result says the work moved the answers.

Citation results

A piece’s first citation is the first qualifying answer observed after that publication. An answer collected before publication does not count, even if its report finishes later. Heralded resolves the published URL through the page’s canonical identity. A superseded publication or a page already cited before publication earns no new citation result. One observation cannot earn credit for two pieces at the same address. The action and Studio reads carry the same first citation, with the engine’s name, the question, the observation date and an answer link when the answer is available. Further citations come from the publication’s measurement window. They count distinct engine and stable question pairs, so repeated answers to the same question do not inflate the total. The list shows up to five pairs and carries the full count. The grouped results field puts confirmed signals in success tone, pending and unchanged signals under watching, and failed reads in warning tone. Missing Search Console or crawler connections produce a neutral not_connected offer naming only the missing sources. The ordered results.steps keeps every declared signal, including one with no questions, keywords or page to watch. The flat trail keeps its four observation states. The result status is Delivered until a qualifying citation makes it Cited, for work that tracks citations. Paying off requires verification of the targeted answer metric against untouched questions. An inconclusive or no-gain measurement leaves a citation as the headline and shows the measurement separately. A citation records what an engine did. It does not show that the work caused a change.