Takeover stabilizes the account, and this cadence keeps it stable. One light daily check catches fires, one weekly working session makes the changes, and a monthly review resets the strategy. Almost every decision is rules based, so two people running this SOP on the same data should land on the same changes.
Daily Checks
The daily pass stays short. We watch spend pacing against the budget, look for Amazon's out of budget flag on campaigns that are making money, and check stock on the best sellers, because a stockout on a hero product tanks conversion and skews every metric downstream. When a best seller drops out of stock we shift spend to in stock products, then shift it back on replenishment, and we log stock status in the performance narrative so nobody mistakes a supply issue for an advertising issue.
Major event windows turn the daily check into a full cadence. At fixed checkpoints we monitor performance, raise budgets on top performers that run out within the account cap, and fix uncompetitive bids on winners showing no impressions. We also pull an end of day snapshot of spend, sales, and ACoS, then compare it to the baseline and to the prior event days. For events we shorten the analysis window to seven days instead of thirty, because event traffic moves too fast for a monthly average to mean anything.
Weekly Session Three Bucket Sort
Once a week we work off the last 30 days of data. Every keyword and every campaign goes into one of three buckets against the target ACoS, which defaults to 20 percent when nothing better is established. A Top Performer sits at or below target with sales, so it gets pushed. A Bloated target sits above target but still sells, so it gets eased off. A Bleeder has spend with zero sales in 30 days, so it gets cut. This one sort drives the bids, the budgets, and the negations that follow.
Weekly Bid Optimization
Bid moves start from the keyword's recent cost per click, and the step size is fixed by its price tier. If a keyword clicks above three dollars, we move it by 30 cents either way. For clicks between two and three dollars, the move is 20 cents. If the click is between one and two dollars, the move is 10 cents. For clicks between 51 cents and a dollar the bid rises by 5 cents or falls by 7, and for clicks between 30 and 50 cents it rises by 3 cents or falls by 5. Top Performers get raised, and Bloated targets and Bleeders get lowered.
Two golden rules overrule the ladder. Top Performers never get lowered, and Bloated targets or Bleeders never get raised, so when a step points the wrong way the bid stays put. We also stay inside Amazon's suggested bid range, so a good keyword never goes below the low end and a bloated one never goes above the high end. The rule of thumb underneath it is simple. We never underbid what sells and never overbid what does not.
Weekly Budget Management
We set budgets in the same session, because bids set the cost per click while budgets set how much a campaign can spend in a day. When a Top Performer is capped, meaning it carries the out of budget flag or its average time in budget sits under 100 percent, its daily budget rises by about 20 to 25 percent, and if it is still capped and still profitable the following week it goes up again. If a Top Performer never hits its cap it stays put, since more budget cannot help a campaign that is not spending what it already has. A Bloated campaign gets its bids fixed first, and only if it is still over target on the next pass does its budget come down by about 20 percent. Bleeders drop to the minimum or get paused outright once they have spent for 30 days with zero sales.
Budget only rises on profitable campaigns that actually hit their cap, and only falls on inefficient ones. We watch TACoS as the top line health gauge and aim to hold it under about 30 percent per product, and a climbing TACoS usually means bloated campaigns are carrying too much budget and need tightening first.
Weekly Harvest and Negation
We work the search term report weekly on a 30 to 60 day window, since search terms need a little time to collect sales. A term is a winner when it has orders, ideally two or more, and its ACoS sits at or below target. Winners get promoted into a manual campaign as exact match keywords with a starting bid near their recent cost per click, and then bid management takes over. The moment a term is promoted it also goes in as a negative exact in the automatic or broad campaign it came from, so the two campaigns do not bid against each other for the same shopper.
Negation runs in the same session. We negate a term when it has meaningful spend with zero orders, meaning roughly five dollars or more than twice the value of a sale, when it is clearly irrelevant to the product regardless of spend, or when it was just promoted. Negative exact is the safe default. Negative phrase is reserved for the rare case where a whole family of terms sharing a phrase is bad, and the negative goes into the specific campaign where the term appeared rather than account wide. One hard exception holds. A term with orders at a good ACoS never gets negated no matter how odd it looks, because sales win.
We log every weekly change with the date, the target, the old value, the new value, and the reason, because the log is how we see what is actually moving the account.
Monthly Review and Deep Clean
By the third business day of each month we post the prior month's actuals against the forecast, and a variance beyond plus or minus 10 percent triggers an investigation. The account gets rescored on current data, a score swing of two or more points forces a fresh look at the growth projection, and every standing assumption gets checked and updated. Only forward months ever change, so the original forecast stays visible and the variance tracking stays honest.
On a longer cycle we run a structural deep clean. Enabled campaigns with lifetime spend and zero sales get paused. Campaigns running above 40 percent ACoS across the recent windows get paused. Campaigns with zero impressions get paused. A catch all campaign is the standing exception and never gets paused. Search terms that burned more than ten dollars over 90 days without converting get negated in their campaigns, and previously converting targets get consolidated into dedicated low bid campaigns so their history keeps working without the waste.
What Runs Automatically
The bucket definitions, the bid step table, and the golden rules are fully mechanical, so they fire the same way every week without judgment calls. The automation platform holds the standing ACoS thresholds and the dayparting schedule between sessions, and around events we tighten those automations before traffic surges and reset them afterward, with the original settings documented first so every change can be reversed. Judgment stays reserved for the monthly strategy review, the reinvestment decisions, and the escalation when a brand's goals start conflicting with their own numbers.
The takeover process that sets this up lives in Our account takeover SOP, and the engine that runs the mechanical pieces sits on the dev and automation service page.
← Back to the process overview