Reading your bid recommendations
What each BidSmart recommendation shows, why it was proposed, how to judge whether to trust it, and how to action or dismiss it, all before anything executes against your ad account.
Each BidSmart recommendation tells you which product it's about, what change it proposes, and, in plain English, why. It shows the numbers behind the call: the product's own performance, its current target, and the proposed one. Nothing happens until you approve it, so a recommendation is an argument to weigh, not an action to fear. This guide explains how to read that argument and decide.
What a recommendation is
A recommendation is a proposal, not an action. It says: here is a product, here is a change I think would help it, and here is why. It sits in your approval queue doing nothing until you decide. So reading a recommendation well means weighing an argument: not bracing for a change that's already happened.
What each recommendation shows
Every recommendation lays out the same fields so you can judge it at a glance:
| Field | What it tells you |
|---|---|
| Product | The exact product (title + ID) the change applies to: recommendations are always product-level, never a blanket campaign move. |
| Proposed change | What would happen: e.g. raise target from 380% to 410%, move to incubator, graduate to main campaign. |
| Current state | Where the product sits now (its current target or campaign) so the change is in context. |
| Reason | A plain-English explanation of why the change was proposed. |
| Evidence | The product's own numbers: spend, conversions, conversion value and ROAS over the analysis window. |
| Confidence | How much data backs the call. Thin-data products carry lower confidence and often point at the incubator instead. |
The reasons you'll see most
Recommendations cluster into a few recognisable arguments:
- "Clearing its target comfortably: room to grow." The product returns well above its current target, meaning Google is being told to bid more cautiously than the product can bear. The proposal lowers the target slightly to let it win more traffic. This is the classic "money left on the table" case: see product-level tROAS analysis.
- "Missing its target: pull spend back." The product consistently returns below target and is draining budget that stronger products could use. The proposal raises its target to throttle it down.
- "Too little data to bid on efficiently." A new or thin-data product with too few conversions for a reliable target. The proposal moves it to the incubator to build data cheaply.
- "Proven: ready to graduate." An incubator product has cleared its data threshold and earned a place in the main tROAS campaign.
- "Budget is throttling this campaign." Not a per-product bid change but a pacing signal that a capped budget is losing you impressions.
Deep dive How to judge a recommendation before you approve it
A recommendation is BidSmart's best read of the numbers. You bring the context the numbers can't hold. Judge each one against these questions:
- Is the evidence window representative? Check the conversions and spend behind the call. A product with 40 conversions in the window gives a target you can lean on; a product with 5 gives a hint, not a verdict: and BidSmart will usually flag that with lower confidence or an incubator suggestion rather than a target change.
- Do you know something the data doesn't? Seasonality, a margin the feed can't see, a hero product you keep loud for brand reasons, a supplier issue that made last month unrepresentative. If you do, reject: that's the whole point of the approval step.
- Is the step size sane? Proposed target moves are deliberately small (typically 10–20% relative) so Google's bidding doesn't reset its learning. If a change feels timid, that's correct: approve a sequence of small steps over weeks rather than reaching for one big jump. See Target ROAS explained for why big jumps backfire.
- What's the downside if it's wrong? Because every executed change records its before/after state and is reversible, the honest answer is usually "small and undoable." That should lower the bar for approving well-evidenced, small-step changes: and it's exactly why approval-first isn't a burden.
Approving does not commit you to anything permanent. It moves one small, recorded, reversible step: and you can watch the result before approving the next.
Actioning a recommendation
For each item you have three choices, and none of them is irreversible:
- Approve: the change moves to execution against Google Ads, with its before-state recorded. Batch several approvals together if you like.
- Reject / dismiss: the change is removed from the queue and your account stays exactly as it is.
- Leave it: do nothing; it waits. An untouched queue means an unchanged account.
Approved items only ever execute after you sign them off. See using the approval queue for the batching and audit-trail mechanics.
Frequently asked questions
Does dismissing a recommendation delete it forever?
No. Dismissing removes it from your current queue. If the underlying performance still supports the change on the next analysis pass, a similar recommendation can resurface, with fresh numbers.
Why is the proposed target change so small?
By design. BidSmart moves in small relative steps so Google's bidding never triggers a full re-learning period. A recommendation is one step, not the final destination, approve a few in sequence rather than one big jump.
A recommendation looks wrong for a product I know well. What do I do?
Reject it. Your knowledge of a product's seasonality, margin quirk, or strategic value is exactly what the approval step is for. Rejecting teaches nothing gets executed and keeps your account as-is.