Your first Lookout

A Lookout watches one part of your advertising and proposes a change when what you described happens. It proposes. Everything it finds lands in the Publish queue and waits for you, exactly as if you had queued it by hand.

This page builds one from a template and follows it all the way through.

The six parts

Every Lookout is the same shape, and it reads like a sentence. The middle column is the word you write if you are working in text rather than in the builder — it is the same rule either way.

There is exactly one then. A Lookout that could do several things at once would be a rule you cannot predict, and predictability is the entire product.

Build one

Open Lookouts and press New lookout. It opens a page of worked starting points — the templates — with the three blank starts beneath them: the builder, the editor, and paste. Most rules take a template or the builder; a rule somebody pasted, or one that nests, wants the editor.

Take the visual builder and start from a template.

1. Pick the grain

Choose target. You are looking at individual keywords and product targets — the level at which money is actually spent.

The other grains exist for other questions: campaign and ad_group for structure, search_term for what shoppers typed, product for products, and gap for rows that do not exist — a design missing from a marketplace it should be in.

2. Set the scope

Leave it at all markets unless you have a reason. A rule scoped to one marketplace silently ignores the others, and that is easy to forget six weeks later.

3. Choose the trigger

Take weekly.

One thing to understand: triggers are floors, not clocks. "Weekly" means not more often than weekly — the extension's background worker sleeps when Chrome is idle, so a run happens the next time the browser is awake after the interval has passed. Days are counted in Los Angeles report days, because that is the day boundary Amazon's own reporting uses.

4. Say what has to be true

This is the rule. Something like:

• spend over the last 30 days is at least $10, and
• orders over the same 30 days are zero

Read it back before moving on: targets that have taken ten dollars and returned nothing in a month. That is a sentence about money you can defend, which is the test a condition should pass.

5. Say what to propose

Choose lower the bid, and set it to a percentage rather than a fixed amount. A flat cut treats a four-cent bid and a four-dollar bid the same way.

Preview it first

Press Preview. It shows exactly what a run would propose, and it is not an approximation: preview and run call the same function, so a preview cannot lie about what a run would do.

You will usually see two lists. The proposals, and — more useful the first time — the rows the Lookout held back and why.

Run it

Press Run. Proposals land in the Publish queue as Pending.

Nothing has reached Amazon. Open Publish, read the rows, approve the ones you agree with, delete the ones you do not. A Lookout that proposes ten things and has three approved has done its job — the other seven were a conversation.

When it holds something back

It often will, and that is the feature. Three guards can stop a proposal even when your condition matched:

• Its price moved. If the product's price changed inside the period the rule is judging, the ACoS it would act on is measuring two different prices. The window is derived from your own rule rather than a fixed quiet period.
• Ads are still feeding it. Advertising is feeding the product, so thin organic sales are not the whole story.
• Out of season. A seasonal product judged outside its season.

My Lookout proposed nothing walks through every reason a row can be held.