The short version
- Rejected pages are not deleted. They stay in the workspace with their score, their tier, and a specific rejection reason.
- Rejection reasons are structured codes, not free text — irrelevant, no placement fit, thin, spam, dead page, already linked, below threshold, and others.
- Any page can be manually promoted for consideration when you see an angle the model missed.
- Every opportunity carries a complete evidence trail: the queries that found it, what reading the page produced, and how each score component was reached.
- This is what makes disagreement productive — you can see exactly which judgement to override.
Automated qualification has an uncomfortable property: the more aggressively it filters, the more valuable it is, and the harder it is to trust. A pipeline that rejects nine pages out of ten has done a great deal of work on your behalf — and you have no way of knowing whether it threw away the best opportunity in the set.
There is only one honest answer to that, and it is to show the work.
Nothing is silently deleted
Every candidate that enters the pipeline and does not become a qualified opportunity remains in the workspace, with the reason it did not make it. Rejection reasons are structured codes rather than prose, so they can be filtered, counted, and reasoned about:
| Reason | What it means |
|---|---|
| Already linked to you | The page prominently links your company already. You have the placement. |
| Irrelevant | Content relevance came in below the floor. The page is not about your market. |
| No placement fit | Placement suitability came in below the floor. Relevant, but nowhere to add you. |
| Dead page | The page no longer resolves. |
| Thin content | Too little substance to hold editorial value. |
| Spam | A link here would be a liability. |
| Unsupported language | The analysis declined to judge placement in a language it cannot read. |
| Analysis failed | The record is incomplete. An unknown, not a low score. |
| Below threshold | Survived the hard checks, scored below the qualification floor. |
| Product limit | Qualified, but beyond the number of opportunities retained for the run. |
The distinction between “irrelevant” and “no placement fit” is the one people use most. The first means discovery reached into the wrong market and the query set may need adjusting. The second means discovery worked and the page simply has nowhere to put you — which is not a problem to fix, just a page to skip.
“We found nothing” and “we found sixty pages and rejected all of them, for these reasons” are completely different answers. Only one of them tells you what to do next.
The evidence trail
Every opportunity — selected or not — accumulates a record as it moves through the pipeline. Three layers, each from a different phase:
- 01
Discovery evidence
Which queries found this page, which strategy each belonged to, where it ranked, and which competitor seed surfaced it. This is why a page found four times across three strategies is treated differently from one found once at position 40.
- 02
Enrichment evidence
What reading the page produced: its type, quality status, language, publication date, the author, whether you are already mentioned or linked, and the written reasons behind the relevance and placement scores.
- 03
Qualification evidence
How each scoring component was reached, which hard checks were applied, and the resulting tier — or the reason for rejection.
47 opportunities for acme.com
Ranked by priority
Best AI visibility platforms for SaaS
growthlab.example · Comparison roundup
0
- Directly compares products in your category
- Three competitors are listed while your company is absent
- The roundup structure supports a new entry without an edit war
Who to email
Sarah Jones · Author · Senior editor
sarah@growthlab.exampleVerified
Example data. Every field shown comes with its source in the product.
Promoting a page yourself
Sometimes you will disagree, and sometimes you will be right. You know things the pipeline cannot: that you already know the editor, that the publisher is about to launch a section you fit perfectly, that a page scored as marginal is read by exactly forty people and all of them buy software like yours.
Any page — including a rejected one — can be manually promoted for consideration. It joins the working list, contacts can be discovered for it, and a draft can be written against it, exactly as if it had qualified on its own.
Why this is a product decision, not a debug feature
It would be simpler to hide rejections. The list would look cleaner and nobody would ask why a particular page is missing.
But the value of automated qualification is not that it makes decisions — it is that it makes the same decisions consistently, at a volume no person can match, while leaving the judgement calls to someone who can make them. That only works if the judgement calls are visible. A filter you cannot inspect is not a colleague doing the first pass; it is a black box you either accept wholesale or stop using.
Dice therefore keeps the evidence for every decision, publishes the scoring model in full, and leaves the final say with you. The pipeline is built for inspection, not blind automation — and the rejections are the part where that commitment is actually tested.