GOOGLE SHOPPING ADS BID STRATEGY: WHAT HOLDS UP

The question I get asked most about Shopping isn't which products to advertise, it's which google shopping ads bid strategy to launch on. And the honest answer is the one nobody wants: it depends on the data you already have, not on which strategy sounds most advanced. Pick wrong and the learning phase quietly eats your first three weeks of budget while the campaign flails.

This is a practitioner's take, drawn from launching and salvaging Shopping campaigns across a dozen accounts, not a rewrite of the help docs. I want to settle the tROAS versus Maximize Conversion Value versus CPC argument the only way that matters: by asking which one survives the learning phase on the account in front of you.

I'll lay out the three real options, explain what the learning phase is actually doing under the hood, then give you my rule for which google shopping ads bidding approach holds up at each level of conversion volume, the migration path between them, and the errors that keep campaigns stuck in learning forever.

One framing to hold onto before we start: in Shopping there are no keywords to bid on, only products, so your entire lever is how you price access to auctions across your product groups. That's a genuinely different discipline from Search. In a text campaign you can lean on match types and negatives to shape demand; in google shopping campaigns the bidding strategy carries far more of the weight, because it is deciding, product by product, how hard to compete. That's why getting the strategy right at launch matters more here than almost anywhere else in Google Ads, and why I treat the first three weeks as the part of the job that actually determines the outcome.

The three options that actually matter

Strip away the marketing names and Shopping really only gives you three meaningful choices. The first is Manual CPC, where you set the maximum cost per click on each product group yourself. It requires no conversions, you control the price of a click directly, and you see exactly what you're paying. It's the only fully hands-on lever left in Shopping.

The second is Maximize Conversion Value, a Smart Bidding strategy that spends your budget chasing the most revenue it can find, with no hard efficiency constraint. The third is Target ROAS, the same value-based engine but bound to a return-on-ad-spend target you set. Google's own documentation frames the Shopping choice cleanly in its overview of automated bidding for Shopping campaigns, and the split it draws is the one that matters: clicks versus value.

Everything else is a variation on these three. Maximize Clicks is really Manual CPC's automated cousin for buying traffic; enhanced CPC layers signals onto a manual base. But the strategic decision, the one that decides whether your campaign survives its first weeks, is which of the three above you launch on. That choice shapes how the campaign learns, as I unpack in my wider guide to how Google Shopping works as a channel.

It helps to separate two families here, because people conflate them constantly. Manual bidding is you setting a maximum cost per click and living with it until you change it. Automated bidding strategies, by contrast, set bids based on the model's real-time prediction of how likely each auction is to produce a conversion or conversion value, adjusting in every single auction rather than once a week when you log in. That difference, a static price you own versus a dynamic price the machine sets, is the whole reason the learning phase exists: a static bid has nothing to learn, an automated one has to gather data before its predictions are any good.

There's a structural point that sits underneath all three, too. In Shopping you don't bid on queries, you bid on your catalogue, sliced into product groups. With Manual CPC you can set a distinct bid at the group level, pushing a high-margin hero SKU hard while holding a loss-leader down. Automated strategies take that control away and bid across the whole structure at once, which is a strength when you have volume and a weakness when you need surgical control. Either way, how you subdivide products, and how clean the underlying shopping feed is, feeds directly into how well any bidding strategy can perform, because the model can only bid on what your feed actually exposes.

Manual CPC's job, then, is narrow but real: it exists to drive traffic predictably and cheaply while you have no conversion history worth optimising against. It won't chase value for you, and it won't get smarter overnight, but it will buy clicks at a price you dictate, which is exactly what a cold account needs before it can graduate to anything automated.

What the learning phase is really doing

Before comparing strategies you have to understand what you're fighting. When you launch or materially change a Smart Bidding strategy, the campaign enters a learning phase: the algorithm is gathering conversion data and calibrating how to bid in each auction. During that window performance is volatile and, crucially, not representative of where the campaign will settle.

The single variable that governs this is conversion volume. A value-based strategy learns from conversions, so a campaign producing 40 a week calibrates fast and stabilises; one producing four crawls and may never fully exit learning. This is the core reason a bid strategy that works beautifully on a mature account collapses on a new one: same strategy, wildly different amount of signal to learn from. The mechanics mirror what I describe for how long the smart bidding learning period lasts across campaign types.

Two things reset or prolong that clock, and both are self-inflicted. Editing budgets, targets or structure mid-learning throws the model back to the start. And launching with a constraint so tight the campaign can't spend starves it of the very conversions it needs to learn from. Understanding that trade-off is most of the battle; the rest is matching strategy to volume.

It's worth being concrete about what "volatile" means in practice, because people panic at day three and undo good decisions. In the first few days a freshly automated campaign will often overspend, then underspend, then swing again as it explores which product groups and audiences convert. Cost per acquisition can look alarming precisely because the model is deliberately testing auctions it isn't yet sure about. None of that early data is a verdict. I've watched managers kill a perfectly healthy strategy on day four because the numbers looked ugly, when the honest read was simply that the campaign hadn't finished learning yet.

The other thing worth internalising is that budget and bid strategy interact, they aren't independent dials. A campaign that is budget-constrained learns more slowly, because it's entering fewer auctions and therefore gathering conversions at a trickle. If you want a value-based strategy to exit learning in a reasonable window, you have to fund it enough to accumulate conversions, then leave both the budget and the target alone while it does. Starving the budget and then blaming the bid strategy for slow learning is one of the most common misreads I see.

Which strategy holds up, by conversion volume

Here's the practical rule I actually use. The right google shopping ads bidding choice is a function of one number: how many conversions the account produces per month. Three bands, three answers.

Near-zero conversions: start on CPC

On a brand-new account with no conversion history, I don't hand the keys to any value-based strategy, because there's nothing for it to learn from. I start on Manual CPC (or Maximize Clicks) to buy clean data as cheaply as possible, control the cost per click, and watch what the feed matches on. This is the foundation phase: I'm not chasing ROAS yet, I'm building the conversion record that makes automation viable later.

Building volume: Maximize Conversion Value

Once the account is reliably producing roughly 15-30 conversions a month, I move to Maximize Conversion Value with no ROAS target. It's the gentlest entry into value-based bidding: it spends the full budget chasing revenue without a hard constraint choking it, so it exits learning faster than a tight target ever would. I let it run uncapped for a couple of weeks and read the ROAS it naturally lands on.

Stable volume: Target ROAS, launched low

With genuine, stable volume I add target ROAS, but I launch the target deliberately low to encourage scaling, let it spend, then tighten it gradually once the campaign is scaling. Google's own Shopping guidance says the same thing: launch with a low ROAS target, then adjust after the campaign scales. Starting on an aggressive target is the fastest way to trap a campaign in permanent learning. When I optimise a Shopping campaign, this launch-low-then-constrain sequence is the part that most reliably separates campaigns that scale from campaigns that stall.

There's no official published minimum for Shopping the way there once was, but in my accounts I don't hand a campaign a ROAS target until it's reliably producing on the order of 15 to 30 conversions a month, and ideally more. Below that the target simply has too little signal to bid against, so it either throttles spend to nothing or lurches around unpredictably. If a client's catalogue is too fragmented to hit that threshold in any single campaign, I consolidate products rather than accept thin data, because one campaign carrying enough conversions will always outperform three campaigns each starved of them.

A worked example from a real account

To make this concrete: I took over a homewares store doing maybe eight conversions a week, previously stuck on an aggressive ROAS target and spending almost nothing. I moved it to Maximize Conversion Value uncapped, roughly doubled the daily budget to give the model room, and left it untouched for two weeks. It settled around a 380% ROAS on its own. Only then did I add a target ROAS just below that, at 350%, to give it headroom to keep scaling. Spend tripled, revenue more than tripled, and the account finally exited the permanent-learning trap it had lived in for months. Nothing clever happened, I just matched the strategy to the data and stopped interfering.

The migration path between strategies

The strategies aren't rivals to choose between once; they're stages you move through as data accumulates. The path that has held up for me is CPC to Maximize Conversion Value to Target ROAS, and the discipline is in the timing of each switch, not the switches themselves.

I only leave CPC when I have enough conversions that Smart Bidding has something to work with. I only move from Maximize Conversion Value to Target ROAS after the uncapped strategy has stabilised and shown me a realistic ROAS to anchor to, and I set the first target just below that number rather than at some aspirational figure. Each transition triggers a fresh learning phase, so I space them out and never stack a strategy change on top of a budget change.

This is also where Smart Shopping's ghost still misleads people. That old automated black box is gone, folded into Performance Max, which is a different animal from a manual Standard Shopping bid strategy; I draw that line clearly in my comparison of smart Shopping versus Performance Max. On a Standard Shopping campaign you own the migration path deliberately, one calibrated step at a time.

A question I get constantly is whether you can skip the middle stage and go straight from Manual CPC to target ROAS once you finally have conversions. You can, but I rarely do, and the reason is risk management. Jumping straight to a hard efficiency constraint on a campaign whose value patterns you haven't yet observed is a guess dressed up as precision. Maximize Conversion Value uncapped is the reconnaissance step: it shows you the ROAS the account naturally produces before you commit to enforcing one. Skipping it means setting a target blind, and a blind target is almost always wrong in the direction that throttles spend.

The same discipline applies when performance dips months later. The instinct is to yank the target up to "protect" efficiency, but a sudden tightening throws the strategy back into learning and usually makes things worse before they get better. If I need to change a mature target I move it in small steps, a few percent at a time, with days between adjustments, so the model re-calibrates gently rather than lurching. Google shopping ads bidding rewards patience and punishes panic, and the migration path is just that principle applied to the account's whole life, not only its launch.

The mistakes that keep campaigns stuck

Almost every stalled Shopping campaign I inherit is stuck for one of a handful of reasons, and they're all avoidable. The most common is launching Target ROAS with a target that's too high for the account's real economics: the bid strategy can't find auctions it judges profitable enough, so it throttles spend to almost nothing and never gathers the data to improve. The fix is counterintuitive but reliable: lower the target, let it scale, then constrain.

The second is fiddling. Changing budgets, targets or product-group structure mid-learning resets the clock every time, so a campaign that's edited weekly is permanently in learning and never shows its true performance. I make a launch plan and then leave it alone for the full window. Patience isn't a virtue here, it's a setting.

The third is a feed problem masquerading as a bidding problem. When a campaign won't spend or won't convert, managers reach for the bid strategy first, but very often the real issue is upstream: disapproved products, thin titles, missing identifiers or a stale shopping feed that gives the model almost nothing to bid on. No bidding strategy, manual or automated, can rescue a catalogue the algorithm can't read. Before I touch a single target I check the feed is healthy, because bidding is the last mile of a chain that starts with product data. Fixing the feed frequently makes the "bidding problem" disappear entirely.

The fourth is treating early automated numbers as a verdict, which I covered above but which deserves repeating because it costs the most money: killing a strategy in its first week guarantees you never learn what it would have done. If you can't commit to leaving a strategy alone for its full learning window, you're not ready to run it, and you should stay on a simpler approach until you can.

If I compress the whole google shopping ads bid strategy question into one line, it's this: match the strategy to the conversion data you actually have, launch value-based bidding loose and tighten it later, and stop touching the campaign while it learns. Do that and the learning phase becomes a stage you pass through, not the place your budget goes to die.