In Shopping and Performance Max, a badly timed edit costs more than a missed optimization. Think of a target moved twice in one week, a product group rebuilt while the bidder is still settling, or a campaign paused because a seven-day chart looked bad. So my routine is a list of things to look at on a schedule, plus a shorter list of things to leave alone.
Here is the routine I run on e-commerce accounts: what I look at weekly, what waits for a monthly pass, and what I leave alone while a campaign is learning or right after I changed it. Take the logic and adapt it. Your cadence depends on your budget, your conversion volume and how fast your catalog changes.
Principle 1: separate “is it running” from “is it improving”
There are two different jobs here. The first is making sure nothing is broken. The second is making the account better. They need different rhythms.
The first job is frequent and cheap. Merchant Center status, billing, and whether impressions and clicks still look like last week. If something breaks, every day of delay costs money, so these checks do not wait for the weekly slot. Google lists five product statuses: Under review, Processing, Approved, Limited and Not approved, and says the Needs attention tab is where you check for issues with a product (Check product visibility and status in Merchant Center, as of 2026-11-18). When a campaign goes quiet, the order I follow is in Google Shopping with no impressions.
The second job, improvement, is where discipline matters. Every improvement is a change, and every change makes the data harder to read. So the weekly slot is for looking and deciding, and a change only follows when the evidence is clear enough to justify it.
Principle 2: know what restarts learning, and what does not
Before you edit anything on a campaign that uses Smart Bidding, it helps to know what Google says triggers a learning period. Google lists three triggers for the “Learning” status. The bid strategy was recently created or reactivated. Campaigns, ad groups or keywords were added to or removed from it. Or the conversion goals or actions you optimize to changed substantially (About Smart Bidding learning, as of 2026-11-18).
The same page says the “Learning” status never lasts more than 7 days, and ads keep running during it. Smart Bidding usually converges to a new goal within 1 to 2 conversion cycles (same page). Budgets, targets and targeting settings do not trigger the status. Google says updates to them take effect in the next auction (same page).
I read that in two ways. First, structural edits are the expensive ones. Adding a campaign to a portfolio, swapping the conversion action or reactivating a paused strategy all reset the clock, so I batch them and do them deliberately. Second, a setting that does not trigger learning still has a cost. Google also advises avoiding multiple target changes within a single conversion cycle unless there is a clear business need (same page). A target that moves three times in a fortnight gives you no clean read on any of the three moves.
Principle 3: judge over conversion cycles, not over days
The page on Smart Bidding recommends evaluating performance over 2 conversion cycles or a 2-week window, whichever is longer, and accounting for conversion delay (same page). The Performance Max evaluation guide also tells you to account for conversion lag. It warns that the average return of a single channel, especially over a short period, does not show the full picture (Evaluate Performance Max Results, as of 2026-11-18).
In practice this sets my weekly rhythm. A week is long enough to spot a break in a pattern and too short to conclude that a bid or a target is wrong. When conversions lag clicks, a week of data includes conversions from clicks that happened earlier. So the weekly pass is for reading signals, and the verdict on a change comes later.
What I look at weekly
Whether the account is limited by budget or by rank. The search impression share columns split the lost share into two causes. Google defines Search lost IS (budget) and Search lost IS (rank) as the percentage of time your ads were not shown on the Search Network, due to insufficient budget or to poor Ad Rank in the auction (Get impression share data, as of 2026-11-18). Shopping shows it on the Campaigns, Ad Groups and Product Groups pages (same page). Each cause needs a different fix. A profitable campaign that loses share to budget is a budget conversation. One that loses share to rank points at bids, targets or product data. I check which one it is before I change either.
Search terms. For Performance Max, Google has a report called “Search terms and landing pages for Performance Max”. It only includes data from March 2023 onward (About the search terms report in Performance Max, as of 2026-11-18). The page also says to prioritize conversion data over clicks and impressions when evaluating Performance Max, and that negative keywords for Search and Shopping can be applied at the account or campaign level (same page). I read terms weekly and add exclusions for queries that plainly do not fit what the store sells, and I keep each exclusion narrow. Google’s own Performance Max guidance says to monitor reach and performance after implementing exclusions (Answering Your Top Questions About Performance Max, as of 2026-11-18). So I tighten later.
Budget constraint. The same Performance Max guidance says to make sure the campaign is not limited by budget (same page). Budget is the lever I reach for first, because Google says adjusting daily budgets takes effect immediately and lets campaigns scale without triggering “Learning” status (About Smart Bidding learning, as of 2026-11-18). That makes it my first move when the numbers justify scaling. I still move it in steps and watch a full cycle before moving it again. That part is my own practice.
Anything that broke the pattern. Cost, conversions, impressions and clicks against the previous period. I am looking for a break in the shape, such as impressions dropping to near zero overnight. Normal variance I leave alone.
What I look at monthly
Product-level performance. The Products tab shows information and metrics by product, category, brand, product type or custom label, and Google says it is available at account and campaign level (Navigate the Products tab, as of 2026-11-18). A month is the earliest I consider a product-level read, and for slow sellers I wait longer, because a single product can have too few clicks to say anything. This is where I decide which products deserve more room and which ones carry cost without returns. If the segmentation is by custom label, Merchant Center custom labels covers how I think about structuring them.
Whether the targets still match the economics. Margins, shipping costs and discounting drift. A target ROAS that made sense in spring can sit above breakeven or far below what the business can afford by autumn. I recheck it against breakeven ROAS for e-commerce. Google warns that a Target ROAS set too high may limit the amount of traffic your ads get. After a change, the bidder needs 1-2 conversion cycles to reach the new target (About Target ROAS bidding, as of 2026-11-18). So a target edit is a monthly decision with a deliberate size, and I log the date.
Product data. Titles, attributes and images drift away from how people search. Feed work belongs in the monthly slot because the effect of a feed change also takes time to show, and mixing it with a bid change in the same week hides which one did the work.
Structure. Whether the campaign split still makes sense. Google’s guidance for Performance Max says to avoid creating separate campaigns without a reason, such as different budgets or CPA/ROAS targets, and to segment asset groups by meaningful themes (Answering Your Top Questions About Performance Max, as of 2026-11-18). If you are unsure which of Performance Max and Standard Shopping fits a catalog, Performance Max vs Standard Shopping is the longer read.
What I leave alone
- A bid strategy inside its first conversion cycles. Google says Smart Bidding converges within 1 to 2 conversion cycles. I do not change the target, the budget and the conversion action in the same stretch, because I cannot tell which one moved the result.
- A target I changed this week. The page above says to avoid multiple target changes within a single conversion cycle absent a clear business need.
- Overly tight targets. Google lists overly restrictive CPA or ROAS targets as something that can limit how many auctions a Performance Max campaign enters (Answering Your Top Questions About Performance Max, as of 2026-11-18). I start loose and tighten once the data is in.
- Conversion actions mid-test. Changing which actions a strategy optimizes to is one of the three triggers for a learning status, so I only change it when the measurement itself is wrong.
- Products with thin data. A product with a handful of clicks is not a finding.
A note on what changed and when
Write down every change with its date. Google’s Performance Max guidance points to change history as the report that relates account changes to performance swings (Evaluate Performance Max Results, as of 2026-11-18). A written log kept by you adds the reason for the change, which the report cannot know. When a number moves three weeks later, the log tells you which change to suspect.
When the routine changes
Cadence depends on volume. A store with a few conversions a week cannot be read weekly at the product level. A store with steady daily volume can. The structure of the routine stays the same either way: check that it runs, decide on improvements slowly, change one thing at a time, and let a conversion cycle pass before the verdict.
If you want a second pair of eyes on your own routine, it is part of my Google Ads for e-commerce work. Schedule a call and have your Google Ads account open.
