v0.1

PPC Dashboard

CAD→USD: 0.7365
KW/ASIN optimizationThe decision flow every keyword / ASIN target is run through — live, drives the attention panel
Campaign budget optimizationWhen a daily budget gets raised, lowered, or left alone — live, all campaign types
Negative keywordsHow the match type is decided, and exactly what each negative blocks — word order included
Audiences & retargetingSponsored Display audiences. Reserved — the rules are not decided yet
More to optimizeThe waiting list: placements, search terms, strategies — nothing built yet
ACOS coloursThe six colour steps — this page and Sales, in one window. Display only, changes no action

▶ Breakdown by campaign

▶ Charts

to
All US CA

How a negative keyword is decided

×
a question what gets added you never pick the match type — the destination ad group does
1What each one actually blocks
You negateAsThese searches STOPThese KEEP RUNNING
dog socks for wounds
4 words, into a broad ad group		 negative phrase dog socks for wounds
best dog socks for wounds
dog socks for wounds xl		 socks for wounds dog — same words, wrong order
dog socks
socks for injured dog
bonsai starter kit
3 words, into a phrase or exact ad group		 negative exact bonsai starter kit — and nothing else bonsai starter kit for beginners
large bonsai starter kit
starter kit bonsai — wrong order
japanese
ONE word — into any ad group		 negative exact japanese — the bare one-word search only japanese bonsai tree
japanese maple kit
gift japanese style
Word order is the part people miss. A negative phrase only blocks its words next to each other, in that order. “socks for wounds dog” contains every word of “dog socks for wounds” and still runs — Amazon does not block it, and neither do we.
2Why one word is never a phrase
A negative phrase blocks its words inside any longer search. With one word that is the whole vocabulary: negative phrase “japanese” would block “japanese bonsai tree”, “japanese maple kit”, “japanese garden gift” — usually the best traffic in the ad group. What actually failed was the bare one-word search, and negative exact stops exactly that. So a one-word term is exact everywhere, even in a broad ad group.
3Where it goes
Ad-group level, never campaign-wide. Your campaigns hold broad, phrase and exact ad groups side by side — 1,589 keyword texts here are EXACT in one ad group while running BROAD/PHRASE in another of the same campaign, so a campaign-wide negative would block a keyword you chose to bid on.

Ticked by default: the keyword's own campaign — its own ad group included, because its sibling keywords keep catching the term after you pause it. Every other campaign that runs the same product is listed under them, unticked, with one click to take them all.
4ASIN targets and Sponsored Display
A product/ASIN target has no match types at all — the negative is simply this ASIN, excluded, and it goes to the ad groups that target products. Sponsored Display has no negative of any kind in Amazon's API, so those rows can only be paused, and the button says so.
5NOT BUILT YET — negatives from the SEARCH TERM report
Everything above negates a keyword you chose. The other half — the actual searches Amazon matched you to — is not judged at all yet. The rule agreed for when we get to it (owner 2026-08-22):

A search term is offered as a negative when it has 10+ clicks and has NEVER produced an order. Nothing else. A term that DOES convert is left alone however expensive it looks — you cannot bid a search term down, so the only lever is a permanent block, and blocking something that sells throws the sales away with the cost. Fix those at the keyword's bid instead.

The click number will be editable like every other, and the destination will follow the same ad-group rules as section 3. Nothing here is judged or pushed today.

KW/ASIN optimization

×
a question raise it lower it stop it a human decides nothing to do every white box is a live setting — change it and press Save at the bottom
Every figure in this tree is the house number. A keyword marked 🎯 special — or one whose product is marked 🚀 launching, anywhere in its ad group — runs the same branches with the target ACOS and the fair chance raised by the two percentages in section 5. The impressions rule is the same for both.
1The two rescues
A keyword above the 150% pause line goes to the “check yourself” list instead of being paused when either of these is true:
Still new — it has fewer than clicks in its whole life — the all-time slice, the same one the trial uses; until it has loaded this half of the rescue does not apply at all — AND it was opened at a HIGHER bid than the normal opening bid for a product at its price (the normal opening bids are table 2). Meaning: it started out expensive and never got a fair try at a normal bid — a human should look before it is killed.
A good run — within the last months (12 at most — that is how far back the page keeps the history) it sold more than units at an average ACOS under %. Meaning: it has proven itself before — killing it over one bad stretch would be hasty.
2Starting bid by product price
What a NEW keyword should open at — used by the “still new” rescue. Leave a bid blank and that band is simply not used.
price from ($)
up to ($)
starting bid ($)
3MAX BID — the ceiling, both ways
The LOWER of these two ceilings. Either may be unknown; whatever is known applies. It is the ceiling a raise never passes — and since 2026-09-01 the line a bid is cut back to when it ends up above it (the “Is the bid ABOVE its MAX BID?” question on the chart). Click any Bid in the table and the card names BOTH figures and marks the one that is actually the ceiling, so you always know which of the two boxes below to change.
1 · % above Amazon’s own suggested-high bid (Amazon offers one for Sponsored Products only).
2 · the max bid for its product’s price range:
price from ($)
up to ($)
max bid ($)
4How a raise or a cut is sized
Once the tree says raise or lower, this is the arithmetic that picks the number.
New bid = Bid × ( Target ACOS ÷ Current ACOS ) Example: $1.20 at 44% ACOS with a 30% target → 1.20 × 30/44 = $0.82.
One push moves the bid at most +% up or −% down, and a raise never goes past the MAX BID (section 3). Meaning: a long way down settles over a few days rather than in one jump, and the cell tells you when a ceiling held it back. These two are the safety limit on EVERY push — the rules’ own and the ones you type in by hand — and they are checked again on the server, so a stale tab cannot get around them. Type a bid past the limit yourself and the page says by how much and asks you to confirm; confirmed, it is sent and the audit row records that you overrode it.
5Special keywords — launches, ranking, going after a competitor
Some words you want to be less profitable on: a product still earning its first reviews, a keyword you are buying rank on, an ASIN you are taking sales from. Mark them 🎯 on the table (or mark the whole product 🚀 as launching) and they are judged on these two numbers instead of the house ones. Everything else is unchanged — the MAX BID, the no-visibility raise, the push limits and the budgets all work exactly as usual.
Target ACOS — raise it by % The ordinary target ACOS is — right now, so a special keyword is judged at —. The other three ACOS lines move by the same amount, so you type one number and cannot end up with an inconsistent set: —
Fair chance — raise it by % The ordinary fair chance is —× the product price, so a special keyword may spend —× the price before it is stopped. This is still a limit, just a bigger one — a special keyword is stopped later, never “never”. And — — without that, a launch word is still stopped after spending a tenth of the price with under two clicks.
Leave both empty and nothing is special. Empty, or 0, means “no increase” — a keyword you have marked is then judged exactly as every other keyword is, and its mark says so when you hover it. Nothing is ever sent to Amazon by marking or unmarking — it only changes how the rules read the word.
6Other settings
Numbers that belong to no branch of the tree. They lived in ⚙ Settings → Colour rules until 2026-08-12, when that window was cut back to colour and nothing else (owner) — they moved here rather than being deleted, because each one is still read by something.
The click window — every click since the last bid change counts, and nothing is capped; only clicks older than the age limit at the top of the tree are retired. The old “gold standard” cap kept only the newest 50 clicks and threw the rest away. It was measured on 2026-08-20: it changed the answer for 30 of 1,191 keywords, and the oldest data since any bid change was 53 days — so the cap was discarding good evidence for almost no protection. Age is what makes a click stale, not its place in a queue.
Table size — the table shows the top rows of the current sort, with one “show all” link after them.
Ideal optimize window — closes at :00 AM US Pacific time. A bid changed through this page after ALL of tonight’s data landed and before this hour counts its change day as the new bid’s first day — judged again tomorrow, so you can optimize daily. Changed later, the mixed day is skipped and the word returns in ~2 days. The badge in the top bar shows the live state; blank switches the window off. Read by the nightly server build — a change applies from the next refresh.
Paused with potential — flag a paused campaign or keyword that sold more than units at an ACOS under % over the last 180 days. Feeds the panel’s “reactivate” list. It suggests nothing by itself — a human decides.
7Target ACOS per product NEED TO CONNECT!
Planned, not wired: today the one 30% target applies to every product. The plan — each product gets its OWN target, from its real margin (the catalog’s COGS + fees, the same figures “Worse performance” already uses):
target ACOS = the product’s break-even ACOS × %  ·  never below % or above %
A fat-margin product may pay more per sale than a thin-margin one. Nothing reads these boxes yet — the whole tree still runs on the single target.
8Critical clicks per product NEED TO CONNECT!
Planned, not wired: today the 25-clicks line is the same for every product — but a product converting 3% needs far more clicks to prove itself than one converting 20%.
critical clicks = × the clicks one sale usually needs (1 ÷ the product’s own conversion rate)
Nothing reads this box yet — every word is still judged at the flat 25.
8Good to know — why the page works this way
What this flow deliberately skips. Anything not ENABLED; a never-sold word whose product has never sold either (no price, so no trial can be sized); and a bid already at its MAX BID, which cannot be raised further. Always PAUSE, never ARCHIVE — a pause can be undone at Christmas.
Everything here is refreshed once a day, all at the same time. Yesterday’s clicks, spend and sales land between 1:01 and 5:50 AM PT; the bids and on/off states are copied from Amazon at 12:02 AM PT, BEFORE the numbers arrive. Both then stay fixed for the rest of the day — so the ACOS you see and the bid you see always describe the same moment. That is deliberate: a bid refreshed hourly next to an ACOS refreshed daily would tell you to cut a bid that had already been cut.
A bid changed in Amazon’s console shows up tomorrow morning, and that is safe, because every suggestion here is a target, not a step: the line says “set it to $0.50”, never “cut it by 40%”. If a colleague already made the change and you apply the same suggestion again, you land on exactly the same bid — nothing compounds. Bids changed with the panel’s own Update bid button are known immediately.
Why the clicks can only be counted in whole days. Amazon reports keyword performance one row per day, with no hours in it. So a bid changed in the middle of a day can never have that day’s clicks split at the moment of the change — the count starts from the first FULL day on the new bid. Knowing the exact minute of a change would not make the numbers any better.

Campaign budget optimization

×
a question raise it lower it nothing to do every white box is a live setting — change it and press Save at the bottom
1Good to know — why the page works this way
One number here changes no decision: the Budget cell tints blue when the heaviest day stayed under % of the budget. That is a colour only — since 2026-08-22 a budget that is never reached is deliberately left alone, because moving a cap nothing touches changes nothing.
These numbers ARE the settings. Change one, press Save, and the attention panel re-writes its budget lines with it — the suggested new budget in the sentence is the same figure that lands in the Update box. Leave a “by __ %” blank and the line still names the action without suggesting an amount. Budgets and the Update buttons speak the campaign’s local currency; Amazon-side caps still apply — a raise stops at ×2 per step and a cut is floored at the $1 daily minimum.
What this flow deliberately skips. Paused campaigns (they can’t spend), and any campaign that spent nothing in the window — there is no heaviest day to judge.
Everything here is refreshed once a day. Yesterday’s spend and sales land between 1:01 and 5:50 AM PT; the bids, budgets and on/off states are copied from Amazon at 12:02 AM PT, BEFORE the numbers arrive. Both then stay fixed for the rest of the day, so what you see is always one consistent picture — never yesterday’s ACOS next to a budget from a different moment.
A change made in Amazon’s console shows up tomorrow. That is safe, because every suggestion here is a target, not a step: the line says “set it to $40”, never “cut it by half”. If someone already made the change and you apply the same suggestion again, you land on the same number — nothing compounds. Changes made with the panel’s own Update button are known immediately.

Audiences & retargeting — rules not decided yet

×
These Sponsored Display campaigns aim at people, not search words: shoppers who viewed a product (last 7 / 14 / 30 days), bought it before, or browse the category.
Today they ride through the general target rules wherever they have data.
Which audiences earn their bid — and by what test — has no agreed rules yet.
This window is reserved so the group is not forgotten. Nothing extra is judged or pushed from here yet — the rules land here once they are decided, never guessed.

More to optimize — the waiting list

×
None of these are built. This window only makes sure they are not forgotten — nothing here is judged or pushed. Roughly in order of value:
1 · Placement extra bids — the top-of-search % and product-pages % each campaign adds on top of its bids. The settings and the top-of-search share are already collected daily.
2 · Search-term harvesting — read what shoppers actually typed (collected daily): promote the terms that convert into exact keywords, negate the money-wasters.
3 · Auto → manual graduation — move an auto campaign’s discoveries into manual campaigns (its own reserved window above).
4 · Audience optimization — which retargeting audiences earn their bid (its own reserved window above).
5 · Bidding strategy — dynamic down-only / up-and-down / fixed, per campaign. Already collected daily.
6 · Budget shifting between campaigns — take from the losers, give to the capped winners. Today every campaign is judged alone.
7 · New-keyword discovery — terms we rank and convert on organically (from the SQP and rank-tracker data) but do not advertise.
8 · Match-type balance — the same word in broad / phrase / exact competing with itself.
9 · Dayparting — by-hour bids and budgets. Deliberately rejected 2026-08-06 (it breaks “everything counts since the last bid change”); would need its own careful design first.

Campaign Breakdown — Yesterday

↕ drag to resize · double-click to fill the screen