A clean memo begins with a dirty market.
That is not a complaint about the market. It is the starting condition.
Retail pages change. Sellers shift. Availability disappears. Product titles blur model differences. Promo tags appear without full context. Marketplace offers look comparable until condition, seller quality, fulfillment, or channel posture says otherwise.
The job of a serious category decision system is not to pretend that the market is clean.
The job is to build a layer between visible market noise and commercial action.
I think of that layer as the refusal layer.
The refusal layer is the part of the system that says:
- this row should not become action language yet,
- this seller should not be treated as comparable,
- this product match is not strong enough,
- this availability state is unknown, not unavailable,
- this price pressure needs recheck,
- this signal is visible, but not decision-ready.
That may sound conservative, but it is not passive.
It is commercial hygiene.
A pricing, category, ecommerce, or marketplace team does not need every visible market signal pushed into a meeting. It needs the right signals separated from the weak ones before the meeting starts.
More data is easy to show.
Cleaner decisions are harder.
The difference is refusal.
Without a refusal layer, a dashboard can become a pressure machine. Every price gap looks urgent. Every low seller looks like the market. Every scraped row looks like evidence. Every blocked page is quietly ignored. Every unknown state is collapsed into a fake answer.
That is how bad market data becomes bad commercial behavior.
A weak reference can create false margin pressure.
A wrong product match can start an unnecessary pricing conversation.
A seller that should have been excluded can make a clean channel look distorted.
An unknown availability state can be misread as a real market condition.
A blocked PDP can still end up influencing a decision if the system does not know how to say no.
This is why blocked, unknown, recheck, and watch are not failure states.
They are control states.
Blocked means the system does not have enough usable evidence from that source.
Unknown means the system should not pretend certainty.
Recheck means the signal may matter, but not before proof improves.
Watch means the signal is commercially interesting, but not yet strong enough for action.
Those states protect the decision memo.
They keep weak evidence from entering the action path.
They also make the output more useful for leadership. A clean memo should not only show what deserves attention. It should also show what was deliberately held back and why.
That matters in TV category work because a television is rarely just a simple SKU row. Model identity, screen size, panel family, year, seller, condition, availability, promo posture, MAP pressure, marketplace behavior, and channel context can all change the meaning of a price.
If the system cannot refuse a weak row, it cannot protect the decision.
This is one of the reasons I do not think of CategoryVantage as a price tracker.
A price tracker can collect.
A decision desk has to judge.
And judgment starts before recommendation.
It starts with the right to refuse.
Before a row becomes part of a weekly commercial decision memo, the system should ask:
- Is the product identity strong enough?
- Is the seller good enough to compare?
- Is the condition comparable?
- Is availability known?
- Is the source usable?
- Is the signal explained by promo, MAP pressure, channel divergence, seller noise, or a real same-product threat?
- Is the proof strong enough for action language?
If the answer is no, the row should not be polished into confidence.
It should be held, explained, and routed correctly.
That is the discipline behind a clean memo.
The best decision surface is not the one with the most rows. It is the one that knows which rows should never reach the decision table.
In market evidence, refusal is not a weakness.
It is the control that makes the rest of the system worth trusting.