Two half-working stepsare not half a path.
Every scorecard you have ever used adds up weighted marks. This one multiplies, because the reader has to get through all five steps in order. Two steps at half score 25, not 50 — and 25 is the number the reader actually experiences.
Averaging a funnel is arithmetic that flatters you.
Of the 243 possible combinations of five steps, a product and a weighted mean return different verdicts on 110. The verifier counts them; the page does not guess.
0.50 × 0.50 = 0.25. A weighted mean would have called the same path 80, which is the number that keeps content libraries feeling fine while nothing converts.
0, 3, 6, 13, 25, 50, 100. That is a property of a five-step product over three marks, not a design choice — so the page states the scale instead of implying a continuum.
You have written the pieces. Some of them are good. The question nobody asks is whether a reader who finished one could get from there to something you sell — and the honest answer, for most libraries, is that half the pieces name nothing, a quarter point somewhere adjacent, and the rest have a link somebody moved in a redesign eleven months ago.
Fix one step and the score doubles. That is what multiplying does.
Seven published pieces. Select C-03 and set the link resolves back to 2 — the score doubles from 25 to 50, because the steps multiply.
| Piece | Points at | Score | Verdict |
|---|---|---|---|
| C-01 How we price a fixed-scope build | Proposal System Toolkit | 100 | ROUTED |
| C-02 The five questions before you automate | AI Cost-Per-Task Calculator | 50 | ROUTED |
| C-03 Why your forms lose leads | Website Lead-Leak Gate | 25 | DEAD-END |
| C-04 A field guide to AI slop | Instagram Profile Conversion Audit | 0 | MISROUTED |
| C-05 What we learned shipping 300 products | Channel Continuation Assay | 0 | DEAD-END |
| C-06 Notes on hiring a first contractor | nothing named | 0 | DEAD-END |
| C-07 The anatomy of a cold email that works | Cold Email Reply-Rate Optimizer | 13 | DEAD-END |
Editing C-03
Why your forms lose leads
1. The ask exists×1.00 → running 100
Is a specific thing you sell named in the piece at all?
2. The ask is visible×0.50 → running 50
Would a reader who stopped at 60% still have seen it?
3. The link resolves×0.50 → running 25
Tested today, not assumed?
4. The destination matches×1.00 → running 25
Does the offer solve the problem this piece raised?
5. The destination is ready×1.00 → running 25
Price named, no further hop?
This model — the product
25
A weighted mean would say
80
Fix first: The ask is visible — the earliest break in path order. Everything after it is untestable.
Routed
2/7
29% against a 60% target
Misrouted
1
overrides the rate
Fix first
C-04 - The destination matches
Library verdict
TRAFFIC IS BEING SENT WRONG
Two half-working steps are not half a path. 0.50 × 0.50 is 0.25.
Your coverage target, your marks — no benchmark. It grades paths, never people, and connects to nothing. Not legal or financial advice.
C-03 is loaded by default. Set “the link resolves” back to 2 and the score goes 25 → 50, because the running product recovers. The panel beside the score shows what a weighted mean would have said, which for that piece is 80. Then look at C-04 and C-05: both score exactly 0, and they get different verdicts.
This is the real output of the shipped engine on the shipped sample.
Nothing below is written by hand. It is the verbatim stdout of python3 engine.py pieces-sample.csv, captured programmatically and diffed against this page before publishing. Note the running column: it is the product so far, step by step.
==============================================================================
CONTENT-TO-OFFER JUNCTION - COJ-089 - RedHub AI LLC
==============================================================================
The score is a PRODUCT of five steps, not a weighted sum. A path is not
a portfolio - a strong step cannot offset a broken one. Grades paths,
never people. Nothing is fetched and nothing is sent.
------------------------------------------------------------------------------
PIECE C-01 [How we price a fixed-scope build]
points at: Proposal System Toolkit
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 2/2 x1.00 running 100.00
The link resolves 2/2 x1.00 running 100.00
The destination matches 2/2 x1.00 running 100.00
The destination is ready 2/2 x1.00 running 100.00
Score (product x 100) 100/100
VERDICT ROUTED
Fix first nothing - every step is at 2
------------------------------------------------------------------------------
PIECE C-02 [The five questions before you automate]
points at: AI Cost-Per-Task Calculator
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 1/2 x0.50 running 50.00
The link resolves 2/2 x1.00 running 50.00
The destination matches 2/2 x1.00 running 50.00
The destination is ready 2/2 x1.00 running 50.00
Score (product x 100) 50/100
VERDICT ROUTED
Fix first The ask is visible
(earliest break in path order - every
step after it is untestable)
------------------------------------------------------------------------------
PIECE C-03 [Why your forms lose leads]
points at: Website Lead-Leak Gate
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 1/2 x0.50 running 50.00
The link resolves 1/2 x0.50 running 25.00
The destination matches 2/2 x1.00 running 25.00
The destination is ready 2/2 x1.00 running 25.00
Score (product x 100) 25/100
VERDICT DEAD-END
Fix first The ask is visible
(earliest break in path order - every
step after it is untestable)
------------------------------------------------------------------------------
PIECE C-04 [A field guide to AI slop]
points at: Instagram Profile Conversion Audit
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 2/2 x1.00 running 100.00
The link resolves 2/2 x1.00 running 100.00
The destination matches 0/2 x0.00 running 0.00
The destination is ready 2/2 x1.00 running 0.00
Score (product x 100) 0/100
DISPOSITIVE FIRED the destination does not match the piece.
Ranked below DEAD-END: a dead end wastes a
click, a wrong destination spends trust.
VERDICT MISROUTED
Fix first The destination matches
(earliest break in path order - every
step after it is untestable)
------------------------------------------------------------------------------
PIECE C-05 [What we learned shipping 300 products]
points at: Channel Continuation Assay
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 2/2 x1.00 running 100.00
The link resolves 0/2 x0.00 running 0.00
The destination matches 2/2 x1.00 running 0.00
The destination is ready 2/2 x1.00 running 0.00
Score (product x 100) 0/100
VERDICT DEAD-END
Fix first The link resolves
(earliest break in path order - every
step after it is untestable)
------------------------------------------------------------------------------
PIECE C-06 [Notes on hiring a first contractor]
points at: nothing named
------------------------------------------------------------------------------
The ask exists 0/2 x0.00 running 0.00
The ask is visible 0/2 x0.00 running 0.00
The link resolves 0/2 x0.00 running 0.00
The destination matches 1/2 x0.50 running 0.00
The destination is ready 1/2 x0.50 running 0.00
Score (product x 100) 0/100
VERDICT DEAD-END
Fix first The ask exists
(earliest break in path order - every
step after it is untestable)
------------------------------------------------------------------------------
PIECE C-07 [The anatomy of a cold email that works]
points at: Cold Email Reply-Rate Optimizer
------------------------------------------------------------------------------
The ask exists 2/2 x1.00 running 100.00
The ask is visible 1/2 x0.50 running 50.00
The link resolves 1/2 x0.50 running 25.00
The destination matches 2/2 x1.00 running 25.00
The destination is ready 1/2 x0.50 running 12.50
Score (product x 100) 13/100
VERDICT DEAD-END
Fix first The ask is visible
(earliest break in path order - every
step after it is untestable)
==============================================================================
LIBRARY
==============================================================================
Pieces graded 7
Routed 2 (29%)
Dead ends 4
Misrouted 1
Your coverage target 60%
LIBRARY VERDICT TRAFFIC IS BEING SENT WRONG
Fix first C-04 - The destination matches
Two half-working steps are not half a path. 0.50 x 0.50 is 0.25, and
that is the number the reader experiences. Averaging them to 0.50
would be arithmetic that flatters you.
==============================================================================Five steps in path order, multiplied — and one verdict that is not a score.
The ask exists · the ask is visible · the link resolves · the destination matches · the destination is ready. Each mark becomes a pass fraction: 0 → 0.00, 1 → 0.50, 2 → 1.00. The score is the five multiplied, times 100. ROUTED at 50 or more, DEAD-END below. There are no weights in the model.
A weighted sum grades a set of independent things. This grades a sequence the reader walks. A mark of 1 is not half credit — it is half the readers, and half of half is a quarter. Any single zero zeroes the whole path, and the verifier proves no combination of strong steps can lift it.
If the destination does not match the piece, the verdict is MISROUTED and it ranks below DEAD-END. A dead end wastes a click; a wrong destination spends trust, at the moment the reader was most interested. It does not come from the score — two paths scoring exactly 0 can receive different verdicts.
Not the worst step — the first one below 2, walking the path in order. Everything after a break is untestable: you cannot learn whether a destination converts when nobody reaches it. The verifier reports honestly that for this model the rule coincides with lowest-mark-then-file-order, so no distinct behaviour is being claimed for it.
The library reads TRAFFIC IS BEING SENT WRONG if any piece is misrouted, whatever your coverage rate is. Otherwise it compares the routed rate against a coverage target you set yourself. There is no benchmark in the product — your target, your marks.
It grades the path you describe. It does not click anything for you.
- A structural read of the route from a published piece to an offer.
- Deterministic and offline — same inputs, same verdict, every time.
- A named model you can argue with: the five steps, the pass fractions, the product and the threshold are all on this page.
- Dateless. Nothing in it drifts.
- Not a crawler. It does not fetch your pages or test your links — you click, then you mark.
- Not a conversion-rate model. A ROUTED path is a path that exists, not a path that sells.
- Not a judgement on the writing. Whether the piece is any good is a different question and a different product.
- Not legal or financial advice, and it scores no people.
Scope note: a ROUTED verdict means a reader could get from the piece to the offer. It does not mean they will, and no verdict here guarantees any conversion, revenue or traffic outcome. The five marks are your own observations, so the output is only as honest as the marking — which is why the anchor for “the link resolves” is that you clicked it today rather than that you believe it works.
Anyone who has published more than they have audited.
- Founders with a blog that gets read and a product that does not get bought.
- Teams after a site redesign, when every internal link moved at once.
- Creators whose back catalogue still gets traffic and points nowhere.
- Anyone who has ever found their own broken CTA by accident.
- Automated link checking — this is a judgement pass, not a crawl.
- Grading whether the content itself is good.
- Forecasting conversion rate from a path score.
- Evaluating the person who wrote the piece. It grades the path.
The Junction grades the route. These three grade the two ends of it.
Step five is 'the destination is ready'. This is what you do when it is not — the offer and the page the path lands on.
ViewThe Junction says the path exists. This says whether the piece at the start of it was worth finishing.
ViewFor libraries built at volume, where the routing breaks fastest — one drafting system with the ask built in beats auditing 200 pieces later.
ViewThe mechanics, stated plainly.
They finished the piece.
Then what?
One purchase, lifetime access, 12 months of updates. $89, once.
Deterministic and offline. It grades paths never people, crawls nothing, and guarantees no conversion or traffic outcome. Not legal or financial advice.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai