GOOGLE SHOPPING CAMPAIGN FROM ZERO: THE FIRST 14 DAYS

Most guides on setting up a google shopping campaign stop the moment you click "Publish". That's the least interesting part. The account either works or falls over in the two weeks that follow, and almost nobody writes about what those days actually look like. This is the sequence I run on every new build, day by day, because a google shopping campaign launched from zero needs a completely different rhythm than one you're scaling. The goal of the first 14 days isn't profit. It's clean data and a structure you can trust.

I've launched enough of these to know where they go wrong, and it's rarely the setup screen. It's the days after: the panic edits, the premature product splits, the switch to automated bidding before there's anything to automate. So I treat the fortnight as three deliberate phases. Days 1 to 3 are setup and hygiene. Days 4 to 9 are a protected learning window where I largely leave the thing alone. Days 10 to 14 are the first real optimisation pass, once the account has finally told me something. That cadence is the whole method.

What follows is the practitioner version, not the documentation version. It's the order I do things in, the things I refuse to touch too early, and the honest reasoning behind each call. If you're building your first Shopping campaign, or your tenth but on autopilot, this is the fortnight that decides whether the account has a foundation or a mess to unpick later.

Days 1–3: setup, feed hygiene and a deliberately simple launch

The first three days are about getting a clean base live, nothing clever. Before I touch Google Ads at all, I confirm the Merchant Center link and that products are actually approved, because an unapproved feed means there is nothing to advertise no matter how well the campaign is built. Feed hygiene here saves days of confusion later: I check titles, product types and any disapprovals before spending a penny. If you want the wider setup context, my walkthrough on the full Shopping campaign setup process covers the account-linking steps in detail.

The launch itself

To create the campaign I click New campaign, choose a Sales or Website traffic objective, select Shopping as the type, then pick the Merchant Center account and country of sale. This is the one screen where I slow down, because you can't change the linked account or country later without rebuilding. I deliberately choose Standard Shopping over Performance Max for a first build: I want visibility into search terms and product performance that a black-box campaign won't give me in week one.

One campaign, one ad group

The structural decision that matters most is restraint. I launch a single campaign with a single all-products ad group. No pre-emptive product splits, no ten-way segmentation. You don't yet know which products or queries earn their keep, and splitting early just spreads a small budget thin across branches that all learn slowly. Structure should follow evidence. I set a modest daily budget sized to collect real data over 14 days, start on Manual CPC or Maximize Clicks, and launch. That's it for day one.

Days 4–9: the learning window I refuse to disturb

This is the phase where discipline earns its keep, and where most people sabotage themselves. Between roughly day 4 and day 9 the campaign is gathering its first signals, and the worst thing I can do is interfere. So I largely leave bids and structure alone. No nudging CPCs daily, no splitting product groups on a hunch, no swapping bid strategy because Tuesday looked flat. Every structural edit resets learning and adds noise to data that is already thin.

That doesn't mean I ignore the account. I check two things daily and act only on those. First, disapprovals and feed errors, because a product that drops out of the feed is invisible regardless of bids. Second, the search terms report, which I read every day without touching it yet. I'm building a mental list of the queries that are clearly irrelevant, the ones I'll block at the end of the window. Understanding what you can and can't control here starts with knowing how keywords work differently in Shopping, since there's no positive keyword field to lean on.

By the end of day 9 I usually have a clear picture: which products get impressions, which queries waste money, and whether the budget is pacing sensibly. Crucially, I have that picture because I resisted editing, not in spite of it. A week of hands-off observation on a new Shopping build is worth more than a week of frantic tweaks, every single time.

Days 10–14: the first real optimisation pass

Now the campaign has earned its first changes. The order matters. I start with negatives, because cutting waste is the cheapest win available: I take the irrelevant-query list I built during the learning window and add it as negatives, trimming the searches that were quietly draining budget. This is usually where efficiency jumps first, before I touch a single bid. My standard approach to building that list mirrors the wider Shopping negative keyword method I use on every account.

Split only the winners

With waste cut, I look at product performance. Now, and only now, do I split. The products pulling meaningful clicks and conversions get lifted into their own product groups or, if they're strong enough, their own campaign so I can control their budget separately. The long tail stays in the all-products group. This is structure following evidence exactly as intended: I'm building around what the data proved, not what I guessed on day one.

By day 14 I have a campaign with waste trimmed, winners isolated, and two weeks of clean conversion data. That last part is what unlocks the next decision. Everything up to here has been about earning the right to optimise; the fortnight ends not with a finished account but with a foundation solid enough that ongoing Shopping campaign optimisation has something real to work with.

The bidding call at day 14 (manual to automated)

The question I get most about a new build is when to switch to automated bidding, and my answer is: not on day one, and often not until the end of these 14 days. A brand-new campaign has no conversion history, so a strategy like Target ROAS has nothing to learn from and simply guesses. That's why I start every new Shopping build on Manual CPC or Maximize Clicks, gathering clicks and conversions cheaply while the account builds a track record.

At day 14 I reassess. If the campaign has generated enough conversions to give an algorithm something to optimise toward, I'll consider moving to Smart Bidding. If it hasn't, I stay manual and give it more time, because switching into automated bidding on thin data just hands the guessing to a machine instead of doing it myself. The threshold isn't a fixed number so much as a judgement about whether there's a real signal yet. For the mechanics of that transition, Google's own reference on how Shopping ads and campaigns work is the documentation I keep open while I decide.

The temptation to automate early is strong because it feels like progress. In practice it's the fastest way to make the first fortnight's data meaningless. Earn the switch. The whole point of the first 14 days is to reach day 14 with enough evidence that the bidding decision is informed rather than hopeful.

The mistakes that quietly kill a new campaign

Almost every failed first fortnight I've inherited traces back to the same handful of errors, and none of them happen on the setup screen. The first is over-building on day one: ten product groups, granular splits, elaborate structure before a single click has landed. It feels thorough and it's actively harmful, because it starves every branch of data and slows all of them down. Start simple, split later.

The second is tinkering during the learning window. Daily bid nudges and structural edits between days 4 and 9 reset learning and turn already-thin data into noise. The third is the premature automated-bidding switch I just described. The fourth, and most avoidable, is ignoring the feed: a Shopping campaign can look flat purely because products are silently disapproved, and no amount of bid work fixes a feed problem. I check approvals before I ever question bids.

If I compress the whole 14 days into one principle, it's this: build the smallest sensible thing, protect the learning window, then let the data, not your instincts, drive every change after day 10. A google shopping campaign structured that way reaches day 14 with a foundation. One built on impatience reaches day 14 with a mess, and you spend the next month unpicking decisions you made before you had any evidence to make them.