Should I use smart bidding for shopping campaigns?

Automated bidding adjusts what you pay for each auction using signals no human can react to in time: the query, the device, the time of day, the location, the browsing context and how those combine. Nobody managing bids by hand can match that.

What it cannot do is know things you have not told it. It optimises towards the conversion data you send. If that data is wrong, incomplete or counting the wrong thing, the system will efficiently pursue the wrong outcome, and it will do so faster than manual bidding would.

So before switching, verify three things. Purchases are tracked once per order with the correct value and currency. Micro conversions like newsletter sign-ups are not being counted as purchases in the same column. And the conversion delay in your category is understood, because a product people take a week to decide on will look unprofitable to a system reading a short window.

Also check the campaign gets conversions regularly rather than in occasional clusters. A campaign producing a handful of orders a month gives the system almost nothing to learn from, and the automation will behave erratically.

Three inputs, and they are all yours to provide.

Conversion data with values. Bidding towards revenue requires the revenue to be sent with each conversion. A campaign reporting conversions without values can only optimise for count, which pushes spend towards cheap products.

Enough volume in one place. Signals are learned per campaign, so splitting a modest budget across many campaigns starves each of them. Fewer, larger campaigns generally perform better under automation than a detailed structure that made sense under manual bidding.

A sensible target. The system will try to hit what you ask for. Ask for something unreachable and it will simply reduce spend until it can, which shows up as a campaign that stops delivering.

Beyond that, keep the feed clean and the product set deliberate. Automation decides how much to bid, not which products deserve the money. A campaign containing your entire catalogue including thin margin clearance items will happily spend on the wrong things at an efficient cost per click.

The strategy choice is yours too. Maximise clicks is for gathering data, cost per action targets are for a stable order cost, and revenue targets suit stores with varied price points.

There are situations where handing bidding over is the wrong call, and it is worth naming them rather than treating automation as always correct.

A brand new campaign with no conversion history. There is nothing to learn from, and a strict target on day one usually results in almost no delivery. Start with a cost based approach, gather data for a few weeks, then switch.

Very low volume campaigns. If a campaign produces a few orders a month, the automation is working from noise, and a simple bid you control is more predictable.

Strict cost ceilings. If you genuinely cannot pay more than a set amount per click, whatever the projected value, manual bidding enforces that in a way targets do not.

Diagnosing a problem. When results have gone strange, moving a campaign to manual for a short period removes one variable and makes the rest readable.

Outside those cases, manual bid management on Shopping is mostly a way to spend hours reproducing a worse version of what the system does automatically. The valuable human work has moved to the feed, the product selection, the structure and the targets.

Switching bid strategy restarts a learning period, so plan it rather than doing it on a whim.

Pick a quiet trading period. Changing strategy the week before your busiest season is how a good year goes wrong.

Set the initial target close to what the campaign has actually been achieving, not to what you wish it achieved. If the campaign has been delivering a certain return, start there and tighten gradually. Starting with an ambitious target collapses delivery immediately, and the campaign never gathers the data it needs to improve.

Change one thing. Do not switch bidding, restructure the campaign and change the budget in the same week, because you will not know which change caused what.

Leave it alone for two to three weeks. Delivery is unstable during learning, and reacting to the first few days produces a permanent learning phase.

Keep the budget stable through the transition. Large budget swings restart learning as well.

Write down the date of the change and what the previous performance was, so the comparison afterwards is against a recorded baseline rather than a memory.

Four things, weekly.

Delivery. Is the campaign spending its budget. A campaign that suddenly spends far less than before is usually being throttled by a target it cannot reach.

Search terms. Automation changes which queries you appear for. Read the report and keep the negative keyword list current, because the system will find cheap traffic that converts occasionally and is not what you want long term.

Product level spend. Check that the money is going to products with margin. Automated bidding optimises the reported value, and if all products report the same margin to the system, it cannot know that one of them earns you nothing.

Actual profitability. Compare gross profit after ad spend against the period before the switch, not just the reported return. A campaign can improve its reported number while shifting spend to lower margin products.

And resist tinkering. Every meaningful change resets learning. Set a rhythm of reviewing weekly and changing monthly unless something is clearly broken, and the account will settle into something you can actually read.