When we looked underneath the campaign total, the answer became less obvious. Some targets were still producing orders. Others were consuming clicks without producing sales.
Protect, Reduce, or Remove.
Context: Noon marketplace, sponsored products campaignsApplies to: Products with existing sales historyPart of: The Saddl advertising diagnostic series
The framework in one sentence
A weak campaign-level ROAS does not mean every target gets the same response. Break the campaign into its targets and search terms, identify what is still producing orders, what is consuming spend without sales, and where lower bids or negative keywords are more appropriate than switching everything off. That distinction — Protect, Reduce Budget, Reduce Bid, or Remove — is the decision the data should drive.
Spend is rising. Orders are not keeping up. ROAS is well below target. The instinct is obvious: pause it.
In one real Noon advertising campaign, a 14-day optimisation window showed AED 836.86 in ad spend, 11 orders, and an observed ROAS of approximately 0.6x — against a campaign target of 2.5x. The campaign clearly needed attention. But when we looked underneath the campaign total, the answer became less obvious.
Some targets were still producing orders. Others were consuming clicks without producing sales. Another needed a lower bid rather than necessarily being removed. The useful decision was not "pause the whole campaign." It was: what should we protect, what should we reduce, and what should we remove?
This is that decision, walked through target by target.
The campaign-level picture was clearly weak
Over the 14-day period reviewed, the Noon campaign showed:
Campaign Signal
Observed Result
Context
Implication
Ad spend
AED 836.86
14-day window
Material budget committed at below-target efficiency
Orders
11
Across all targets
Some conversion happening — but not evenly distributed
Observed ROAS
~0.6x
Campaign blended
Well below campaign objective — action warranted
Campaign target
2.5x
This campaign only
Not a universal Noon benchmark — context-specific
Decision direction
Reduce Budget
Campaign level
Reduce exposure; do not automatically remove all targets
At campaign level, this was not a marginal miss. The observed return was materially below the target being used for this campaign. That justified action. But "action" does not automatically mean turn everything off. A campaign is an aggregation of multiple targets and search behaviours. Some may be inefficient. Some may still be useful. Those need to be separated before making a broad decision.
Same campaign. Two very different search behaviours.
Inside the same campaign, two targets told very different stories. The same blended ROAS combined both of them:
Producing Orders
"water bottle"
Orders5
Cost per order~AED 30
Sales✓ Converting
Decision applied
Protect / Continue testing
Consuming Spend
"milton water bottles"
Clicks27
SpendAED 31.52
Sales0 sales
Decision applied
Investigate for negative / exclusion
Across five campaigns, "milton water bottles" recorded 27 clicks, AED 31.52 in spend and zero sales. The campaign average combined both behaviours. The operating decision didn't.
The phrase target "water bottle" had generated 5 orders at approximately AED 30 cost per order. That does not prove the target should automatically receive more budget. But compared with the rest of the weak campaign, this target was still doing something valuable — it was producing orders. That makes it a candidate to protect or continue testing carefully, rather than remove simply because the overall campaign ROAS was poor.
Now compare that with "milton water bottles." Across five campaigns, the term recorded 27 clicks, AED 31.52 in spend, and zero sales. The useful question here is no longer "should the campaign stay on?" It becomes: why are we continuing to pay for this search behaviour?
A third target needed a bid change, not a pause
Between the extremes of a converting target and a zero-sales term, there is a third type of PPC decision — one that neither protects nor removes, but adjusts:
Target phrase
"water bottles" — phrase match
Observed ROAS
Observed
~0.47x ROAS
Below campaign objective of 2.5x
The target may still have relevance. The problem may be that the business is paying too much for the traffic relative to the return currently being generated. That does not automatically require a pause.
Bid recommendation
AED 1.97
→
AED 1.48
≈ 25% lower bid
This preserves the possibility of demand while reducing the price paid for that traffic. Now the operator can observe what happens at the lower bid. If it remains weak, another action may be needed. AED 1.48 was the recommended bid — not an established optimal.
The four decisions inside a weak campaign
A weak campaign can contain several types of inefficiency. They should not all receive the same response:
01
Protect
When to use
A target is still generating useful orders and deserves continued testing, even inside a weak overall campaign.
"water bottle" — 5 orders, ~AED 30 cost per order. This target earned continued testing while the campaign budget was reduced overall.
02
Reduce Budget
When to use
The overall campaign is consuming too much spend relative to the result. The objective is to reduce exposure while the campaign is underperforming.
~0.6x observed ROAS against a 2.5x target — a strong signal that the campaign-level budget needed attention, without removing all targets.
03
Reduce Bid
When to use
The traffic may still be relevant, but the current click price is difficult to justify given the return being generated.
AED 1.97 → AED 1.48 for "water bottles" phrase match showing ~0.47x ROAS. Preserves the possibility of demand at a lower cost.
04
Exclude / Negative
When to use
A term repeatedly consumes spend without producing useful results and evidence has had time to accumulate.
27 clicks, AED 31.52 spend, 0 sales for "milton water bottles" — a much stronger negative candidate than a term with just a few non-converting clicks.
These are operating choices, not automatic rules. The right action depends on the product economics, campaign objective and accumulated evidence.
"Optimise the campaign" is too vague to be useful
Marketplace teams often use phrases such as "optimise PPC," "improve ROAS," "reduce ACoS," or "cut wasted spend." Those goals are reasonable. But they do not tell the operator what to do. A useful optimisation decision should be more specific: which budget should change? Which bid should come down? Which search term should be blocked? Which target should be protected?
The report says: ROAS is 0.6x. The workflow says: reduce the campaign budget, protect selected converting targets, lower certain bids and investigate non-converting search terms. That is actionable.
Why "pause everything" can destroy useful demand
Imagine the campaign is paused immediately because the blended number looks poor. The waste stops — that part is good. But so does every target that was still generating useful orders. The operator has removed the bad and the potentially good with the same action.
Sometimes that is still the right commercial decision — the product may have poor contribution, stock may be constrained, or the campaign may no longer serve a strategic purpose. But that conclusion should come from the full context. It should not happen automatically because one blended number turned red.
The opposite mistake is equally damaging
The other error is keeping the entire campaign running because "at least it is generating orders." That can also hide waste. A campaign can have converting terms that make it appear active while other targets quietly consume large amounts of spend. So the right question is not: did the campaign produce any sales? It is: which parts of the campaign are earning their place?
ROAS still needs product economics underneath it
Suppose a target improves from 0.6x to 2.0x. Is that good? We still cannot answer without context. A product with strong margin before advertising may support a lower ROAS than a product with very thin margin. A campaign used for a launch may deliberately tolerate weaker short-term economics. And a product with high returns may deserve caution even if the advertising itself converts efficiently. So a PPC decision should eventually reconnect to product contribution, marketplace fees, product cost, returns, inventory and business objective. There is no universal number where "above this ROAS = scale" and "below this ROAS = pause."
The campaign objective still matters
There are situations where weaker short-term advertising economics may be deliberate — launching a new product, entering a marketplace, trying to build initial visibility, supporting a strategic SKU, testing a new price, or clearing inventory. The important word is knowingly. If the business says "we are willing to spend at this level for this period because we are trying to achieve this objective," there is a decision behind the number. If nobody knows why the campaign is losing money and it simply continues because "we are getting some sales," that is a different situation.
A negative keyword is not simply a "bad keyword" label
Negative keywords should also be used carefully. A term with zero sales does not automatically deserve permanent exclusion after a few clicks. The operator should consider how many clicks it has received, how much it has spent, whether it is genuinely relevant, whether the product page converts appropriately, and what the campaign is trying to achieve.
In the example above, 27 clicks with zero sales across five campaigns creates a much stronger signal than one or two non-converting clicks. The evidence has had time to accumulate. That is when a negative-keyword decision becomes more defensible.
Bid reduction solves a different problem
A negative removes or restricts traffic. A lower bid says: we still want this traffic, but not at the current price. For the target where ROAS was around 0.47x, the recommendation to move the bid from AED 1.97 to AED 1.48 did not assume the target had no value. It recognised that the current economics did not justify paying the same amount for the click. Now the operator can observe what happens at the lower bid. That is a more controlled decision than immediately removing the target altogether.
Optimisation should be a loop, not a one-time cleanup
The advertising decision does not end when the bid changes or the negative is added. It creates a new question: what happened afterwards? A useful workflow should therefore repeat:
The PPC Optimisation Loop
Step 01
Campaign Signal
Is performance below the intended objective? Identify the weak campaign — where ROAS, spend and orders tell you action is needed.
Step 02
Target Analysis
Which targets are creating orders? Which ones are not? Separate useful demand from waste before making a broad decision.
Step 03
Search-Term Analysis
What did customers actually type before clicking? Where is non-converting spend accumulating? This is often where waste becomes visible.
Step 04
Product Economics
Does the product have room to support this advertising? Reconnect to contribution, fees, returns and inventory before scaling or cutting.
Step 05
Operator Decision
Protect, reduce, exclude or investigate? Which budget should change, which bid should come down, which term should be blocked?
Step 06
Execute
Apply the agreed changes. For Noon, recommendations are reviewed and applied through the team/manual workflow — human judgement remains part of the process.
Step 07 — Always
Monitor & Review
What happened after execution? A recommendation is not an outcome. A bid reduction may look sensible. A budget reduction may look sensible. But the business still needs to see what actually changed before deciding whether to keep, reverse or refine the change.
A recommendation is not an outcome. Only the monitor step confirms whether the change actually improved performance.
The Saddl Operating Workflow
From weak advertising signal to specific action
01
PPC Performance
Campaign signal: where is the weakness? Which targets, terms and bids need attention?
02
Optimiser
What budget, bid, target or search-term changes deserve consideration? Saddl surfaces the signal.
03
Human Review
Does the recommendation fit the product economics and campaign objective? Operator judgement remains part of the workflow.
04
Noon Execution
Apply or coordinate the agreed changes through the operating workflow. Noon changes require team/manual implementation.
05
Monitoring
Did spend, orders and efficiency actually change? The review loop confirms whether the action improved performance.
Saddl does not automatically execute Noon PPC recommendations. The current workflow retains human review and team implementation. Noon advertising capabilities differ from Amazon Ads API-supported actions.
0.6x
The campaign metric told us where to look. The target and search-term data told us what to do next.
AED 837 spent. 11 orders. 0.6x ROAS against a 2.5x target. That clearly required action. But when the campaign was decomposed: one target had produced 5 orders, another search term had produced 27 clicks and zero sales, and another target had 0.47x ROAS with a recommendation to reduce the bid from AED 1.97 to AED 1.48. Those were not the same problem. So they did not receive the same solution.
What to check before pausing a weak PPC campaign
Campaign objectiveWhat is the campaign supposed to achieve? Profit, launch, visibility, market entry, clearance or growth? A weak ROAS means something different under different objectives.
Campaign-level trendIs performance weak over a meaningful period or only a few days? Avoid reacting to noise — the 14-day window here provided enough signal to justify a considered response.
Target-level performanceWhich targets are producing orders? Which ones are not? The campaign average can hide both useful demand and waste simultaneously.
Search-term performanceWhere are clicks accumulating without sales? Are irrelevant or low-value terms consuming budget? 27 clicks and zero sales across five campaigns is a much stronger signal than two non-converting clicks.
Cost per orderSome targets may still produce orders but at a cost that does not work for the product. ~AED 30 cost per order may be acceptable for one product and unworkable for another.
Bid levelDoes the target need to be removed, or could a lower bid improve the economics? A 25% bid reduction may preserve useful demand while reducing the price the business pays for that traffic.
Product economicsWhat contribution exists before and after advertising? ROAS cannot answer this alone. A 2.0x ROAS may be good or bad depending on the product margin and operating context.
ReturnsAre customers keeping the product being advertised? Efficient acquisition into a product-quality problem is not necessarily good growth — it compounds a return cost problem.
InventoryDoes the product have enough stock to support more demand? Do not optimise a campaign to scale a SKU that cannot remain available — that creates a different operational problem.
What happened after the last change?Optimisation is iterative. Always compare the recommendation with what actually happened after execution. A recommendation is not an outcome until it has been monitored.
Frequently asked questions
Should I pause a campaign if ROAS is below target?+
Not automatically. A weak campaign-level ROAS should trigger investigation. Some targets may still be producing useful orders while others are causing most of the waste. Review the campaign at target and search-term level before deciding whether the entire campaign should be stopped.
What is the Protect / Reduce / Remove framework?+
Protect / Reduce / Remove is a decision framework for PPC optimisation that separates the four actions available when a campaign underperforms: protect a target that is still generating useful orders; reduce the campaign budget; reduce a bid where the target may still have value at a lower cost; and exclude or negative a search term that has accumulated evidence of non-performance. The right action depends on what's actually happening at target and search-term level, not on the blended campaign ROAS alone.
When should I reduce a bid instead of pausing a keyword?+
A bid reduction may make sense when the target remains relevant but the current cost is too high relative to the return being generated. Pausing removes the traffic completely. Lowering the bid keeps the target active at a lower cost while you continue evaluating performance.
When should I add a negative keyword?+
A search term deserves investigation for negative treatment when it repeatedly consumes clicks or spend without producing useful results, particularly when the term is irrelevant or evidence has accumulated over time. Do not make the decision from one or two clicks alone. In the example above, 27 clicks with zero sales across five campaigns creates a much stronger signal.
Is 2.5x a good Noon ROAS target?+
Not universally. The 2.5x figure in this example was the target used for this particular campaign. The right target depends on the product economics, campaign objective, marketplace conditions and business strategy. It should not be treated as a Noon-wide benchmark.
Does Saddl automatically apply Noon PPC changes?+
No. Noon recommendations are reviewed and implemented through the agreed team/manual workflow. Marketplace execution capabilities differ, and human judgement remains part of the decision. Saddl surfaces the signal and the optimisation recommendation — the operator applies or coordinates the Noon changes and monitors what happened afterwards.
Saddl Account Review
Don't pause the red number. Find what's making it red.
Review advertising performance alongside product economics and marketplace context to identify where spend deserves protection, reduction, investigation or a different operating decision.
The advertising examples above use anonymised operating data from a real UAE Noon advertising environment. The campaign example covers a 14-day optimisation window and includes approximately AED 836.86 in spend, 11 orders and observed ROAS of approximately 0.6x against a 2.5x campaign target. Search-term and target examples represent observed performance and optimisation recommendations within the available data.
The recommendation to reduce a campaign budget, lower a bid or investigate a negative keyword should not be interpreted as evidence that the change subsequently produced a quantified saving, ROAS improvement or profitability improvement. ROAS is an advertising-efficiency metric and should not be treated as equivalent to product contribution or accounting profit. The 2.5x target shown here was specific to the campaign context and is not presented as a universal Noon advertising benchmark. Noon advertising recommendations require operator review and team/manual implementation within the current workflow rather than being represented as fully automated campaign execution.