A weekly routine that actually works

Amazon advertising rewards consistency more than intensity. A modest routine every week beats an overhaul every quarter, because the account drifts continuously and small corrections compound while large ones destroy your ability to measure.

Here is a routine that fits in about twenty minutes, plus the part that is harder than any of it: knowing when to do nothing.

Weekly — the main pass

Do it on the same day. Not because any day is better, but because a fixed cadence makes the interval between decisions constant. Variable intervals make results uninterpretable — you cannot tell whether a change worked or simply had more time.

Look at direction before detail. Two minutes on spend, sales, ACOS and TACOS against target, this week versus last. You are asking one question: is anything meaningfully different from last week? If yes, find out why before changing anything. Optimising into an unexplained shift is how you end up chasing a competitor's promotion or your own stock-out.

Work the queue. Most of a week's changes are the same clean patterns and can be swept in bulk. What deserves your actual attention:

• Anything held by a gate — the reason is the interesting part
• Strategic trade-offs the engine deliberately will not make alone
• Anything unusually large

Apply approved changes. Download the reviewed files, upload them to the correct Amazon marketplace and inspect the processing report. AutoPPC checks expected state on a later pull. An approval or download alone does not mean the change landed.

Review evaluation status. Check what was verified, kept, reverted or is still awaiting enough settled evidence. Investigate a repeated pattern of reversions, including your targets and data quality. A kept result means the change passed its evaluation rules; it does not prove a causal profit gain.

Every two to four weeks

Placements. They move more slowly than bids and need realized placement data to judge. Monthly is usually right; weekly generates noise.

Budget check. Which campaigns capped consistently, and were they under target while doing it? That pattern is the highest-value thing you will find in a routine check.

Structural findings. Terms owned by several campaigns, hero products without dedicated campaigns, one marketplace shaped differently from the other. These persist rather than resolving on their own, so they can wait — but they should not wait forever.

Monthly

TACOS trend, per marketplace. Use matching total-sales data and comparable windows. Record stock, pricing, promotions and seasonal demand. There is no fixed six-to-eight-week deadline that proves an organic sales effect, and a lower TACOS does not establish that ads caused the change. See ACOS vs TACOS.

Targets still right? Costs change, prices change, competition changes. A target set eight months ago against margins that have since moved is quietly wrong in a way nothing will alert you to.

What is not working. Products where advertising has not moved anything in months. Sometimes the answer is not a bid — it is the listing, the price, the reviews, or the product.

The harder skill: not optimising

If the account is on target and trending well, do less. There is genuine skill in sitting on your hands. Every change resets a measurement clock and adds a variable. A quiet week where you approve four negatives and leave everything else alone is often the correct week.

Do not optimise immediately after anything unusual. A deal, a stock-out, Prime Day, a competitor's promotion — the data from that period does not predict the period you are about to advertise into. Wait for a clean window, or start your window after the event.

Do not judge a change before its attribution settles. Days one to three after any change look artificially bad, because the spend has landed and some of the resulting sales have not been attributed yet. Judging inside that window is the single most reliable way to revert changes that were working.

Do not change settings every week. If you adjust your target continuously, you can never attribute a result to anything — you have been moving the goalposts and the ball at the same time. Give a setting several runs before deciding it is wrong.

Choosing the window

The date range you optimise on is a decision, not a default.

Longer windows are statistically sounder and slower to notice change. Shorter windows react quickly and mistake noise for signal. A marketplace with a third of the traffic needs roughly three times as long to reach comparable confidence.

But the more important rule is where the window starts: after your last significant event. Optimising on 30 days that include a four-day deal means pricing normal weeks using promotional conversion rates. Your bids will be too high and you will not know why.

The mindset that helps: you are not optimising based on what happened. You are optimising for what is about to happen. Choose the window whose performance best predicts next week.

Two rhythms, not one

Most accounts need both a routine pass and the ability to intervene.

The routine pass covers the whole account, uses a long window, moves conservatively, and runs on a schedule. It keeps things tuned.

Intervention is surgical — one campaign, a short window, a bigger move, right now, because something is visibly wrong. Smaller blast radius, so a wider move is acceptable.

Running only the routine means real problems wait for the schedule. Running only interventions means the account is never systematically maintained and you are permanently firefighting.

When something looks wrong

A sequence that avoids most bad decisions:

1. Is it real? One day is noise. Check several.
2. Is it advertising? A stock-out, price change, lost buy box, review drop or competitor

promotion all look like an advertising problem in an advertising report.

1. Which lever? Capped while under target is budget. Bad ACOS with fine placement

distribution is bids. Fine ACOS with all spend on product pages is placements.

1. One change, then wait. Change three things and you learn nothing regardless of the

outcome.

What AutoPPC does with this

The routine pass runs weekly or biweekly on a schedule you set, with a window per marketplace. Changes settle before being judged, and the day-5 verdict lands in your change history with the ACOS movement attached.

A daily watcher checks live spend and delivery state and raises alerts, so genuine problems surface without you waiting for the weekly run. Alerts never change anything.

Skipping a week costs nothing structurally — proposals recompute from current evidence next run rather than accumulating stale. The only cost is that waste you would have stopped keeps running.