International
Most Google Ads problems in the UAE are not keyword problems. They are rhythm problems. Thin conversion data leaves the algorithm guessing, volatile results provoke the manager into action, and every substantive edit sends the bid strategy back to recalibrate. Round two then starts from worse information than round one. Over-optimising, accepting whatever Google suggests, and feeding the account bad data turn out to be one failure caught at three different points in the same cycle.
This bites harder here than elsewhere. Small search volumes, long B2B consideration cycles, and a buying culture where serious enquiries arrive by phone call, WhatsApp or a direct email to a salesperson leave a great many UAE accounts below the conversion volume that Smart Bidding and broad match quietly assume. Settings that work on a high-volume US ecommerce account do real damage on a Dubai lead-gen account.
Why Does My Google Ads Campaign Get Worse Every Time I Edit It?
Because substantive edits restart the bid strategy's learning period, and the campaign spends that window recalibrating rather than performing. Google's guidance is that calibration can take up to three weeks or one to two conversion cycles, faster where there is already plenty of conversion data. Performance wobbles while that happens. Most managers read the wobble as a problem to solve, edit again, and restart the clock.
Google names two triggers: creating or reactivating a bid strategy, and changing a bid strategy setting. That covers most of what gets done on a Monday morning. Practitioners have put usable thresholds around those categories, and while the percentages below are rules of thumb rather than published Google figures, they hold up consistently across accounts.
| Edit | Restarts learning | Do this instead |
|---|---|---|
| Switching bid strategy type | Yes | Commit to one strategy for a full conversion cycle before judging it |
| Changing tCPA or tROAS by more than roughly 15 to 20% | Yes | Move the target in smaller steps, one cycle apart |
| Budget swings of roughly 20 to 30% or more | Yes | Scale in increments under 20% across several days |
| Pausing then restarting after several days | Yes | Reduce budget rather than pausing when you need to slow spend |
| Changing the conversion action or counting method | Yes | Settle your tracking before launch, not after |
| Adding or removing audience segments | Yes | Batch targeting changes into one deliberate edit |
| Restructuring ad groups | Yes | Restructure once, then leave it alone to accumulate signal |
| Ad copy edits | No | Safe to test continuously |
| Budget nudges under roughly 20% | No | Safe for gradual scaling |
The safe column is short. Copy changes and small budget adjustments are close to the only levers with no calibration cost, which is uncomfortable for anyone who reports weekly on optimisation activity.
How Long Is the Google Ads Learning Period?
Up to three weeks or one to two conversion cycles, according to Google, and faster where conversion data is plentiful. The important word is cycles. Because calibration is measured in conversion events rather than days, a campaign producing three conversions a week and one producing thirty sit on completely different timelines while the calendar treats them identically.
This is where low-volume accounts get stuck. Google does not publish a minimum, but the working figure among practitioners is around 15 conversions a week, below which campaigns routinely stay in learning for three weeks or longer and some never fully exit. If that describes your account, waiting patiently will not fix it, because the problem is volume rather than duration. No amount of restraint manufactures conversion data that isn't there.
Sales cycle length compounds it. A UAE B2B prospect might click an ad, read a case study, sit through a meeting and enquire ten days later. Where the attribution window is shorter than that gap, conversions land outside the window they belong to, and the algorithm calibrates against a distorted picture of what worked.
Once the Learning status clears, give it another cycle. If results are still erratic thirty days later, stop blaming learning. In our experience, the cause is usually one of four things, and each leaves a different fingerprint.
| Likely cause | What it looks like | Where to check |
|---|---|---|
| Target set too aggressively | Impression share collapses, delivery throttles | Bid strategy status, "limited by target" flag |
| Flat value signal | Every conversion looks identical to the algorithm | Conversion values, whether any are differentiated |
| Conversion action mismatch | Conversion count healthy, revenue flat | Which actions are set as primary in Conversions |
| Insufficient volume | Learning status clears then returns | Weekly conversions per campaign against the 15 mark |
What Happens When You Apply Every Google Ads Recommendation?
Not every recommendation will benefit your campaign, so read each one on its merits and resist clearing the whole tab in a sitting. The common failure here is not an absent manager. It is an attentive one who opens Recommendations, sees a score of 68%, and works down the list clicking Apply because a higher number looks like a job well done.
Every recommendation involving bid strategy, targeting or budget restarts the learning period described above. Accept five in one session and the campaign is not recalibrating once. It is being reset repeatedly by someone who believes they are improving it. The account never stabilises long enough to produce the data that would show whether any of it helped.
Budget is where the manual route diverges from the automated one. Google's auto-apply recommendations implement changes across ads, keywords, match types and bid strategies depending on which bundles are switched on, and Google states plainly that auto-applying will not increase your budget. Budget-increase recommendations sit on the Recommendations page instead, waiting for a person to accept them. A human can raise spend in one click where the automation cannot.
Neither route is inherently wrong, and the tracking and disapproved-ad prompts are often worth acting on. The damage comes from treating the list as a chore to clear, on the assumption that Google's suggestion and your commercial interest are the same thing.
What Does Your Google Ads Optimisation Score Actually Measure?
How closely your account matches Google's recommendations, which is not the same as how well it converts. The detail sits in Google's own documentation: an account reaches 100% by applying or dismissing all recommendations. Dismissal counts toward the score.
A perfect score can therefore mean an account has been carefully optimised, or that somebody worked down the list clicking dismiss. The number cannot distinguish between them, which makes it a poor proxy for account health however reassuring it looks in a report. It is the same problem we wrote about with AI visibility scores: a metric that measures activity gets read as a measure of outcome.
Google does publish supporting evidence, reporting that advertisers who raised account-level optimisation score by ten points saw a median 14% lift in conversions. Read each recommendation on its merits and treat the score as a prompt to investigate.
Why Doesn't Broad Match Work on a Small Budget?
Because broad match buys exploration, and on a small budget that exploration comes out of the money that should be buying known intent. Broad match today reaches considerably wider than it did five years ago, while plenty of advertisers still work from a mental model built in 2021.
That gap is about to widen again without anyone opting in. Google has confirmed that from September, campaigns using Dynamic Search Ads, automatically created assets, or the campaign-level broad match setting will be automatically upgraded to AI Max, which adds keywordless matching on top of broad match. Worth knowing before it happens: final URL expansion is switched on by default when AI Max is enabled and has to be unchecked manually, so Google will also start choosing which of your pages the traffic lands on.
It performs reliably when four conditions hold at once. The checks in the third column are the ones we run first on an account audit.
| Condition | Why it matters | How to check |
|---|---|---|
| Smart Bidding with sufficient data | Without it, there is no auction-level intelligence steering the expanded reach | Weekly conversions per campaign |
| Clean conversion tracking | Broken tracking means the algorithm expands toward the wrong outcome, confidently | Conversion actions marked primary, duplicate counting |
| Maintained negative keyword list | Inherited lists do not cover queries broad match surfaces today | Date of last negative added |
| Healthy conversion volume | Signal cannot be separated from noise below a certain threshold | Conversions per campaign per week |
Miss one and you leak spend. Miss two and the algorithm optimises confidently toward a corrupted definition of success. Google's search terms report makes this harder to catch than it once was, since a share of query data is withheld for privacy reasons, so the waste you can see is not all of it.
How Does Bad Conversion Data Compound in Google Ads?
Because today's bad queries become tomorrow's inputs. Irrelevant broad match traffic does not simply waste this month's budget. It fills the search terms report that your next round of keyword expansion, copy testing and landing page decisions gets built from.
The damaging version runs like this. Broad match surfaces a query from someone with no intention of buying. A form fires and registers as a conversion. Smart Bidding logs the win and bids harder on similar queries. Conversion volume climbs, the report looks healthy, and the account is now learning quickly in the wrong direction.
Attribution sits underneath all of it. Google Ads optimises toward the conversions it receives, and in a typical UAE lead-gen account, a large share never reach it. Phone calls can be tracked but often are not, because call conversions were never configured. Walk-ins and showroom visits need offline conversion import that nobody set up. Direct emails and Instagram messages to a salesperson land outside every system. And WhatsApp passes no referrer data at all, so those enquiries are structurally invisible rather than merely unconfigured.
The effect is the same regardless of channel. The algorithm learns from whoever happens to fill in a form, and the campaign sits artificially below the volume threshold, which is how an account with healthy real demand ends up stuck in permanent learning. WhatsApp is the widest of these gaps in this market, and we covered how to close it in WhatsApp lead tracking for Google Ads. The same principle applies to every other channel: if a conversion happens somewhere Google Ads cannot see, it has to be sent back deliberately.
How Often Should You Optimise a Google Ads Campaign?
Set the frequency by conversion volume, not by the calendar or by how overdue a report feels. A campaign generating fifty conversions a week can absorb weekly attention. One generating five cannot, and putting both on the same schedule guarantees the second never finishes calibrating.
For a low-volume UAE account, the sequence matters more than the frequency.
- Fix the conversion signal before touching bid strategy. Audit which conversion actions are counting, set attribution windows that match your real sales cycle, and get every off-platform conversion imported: calls, WhatsApp enquiries, walk-ins and CRM-closed deals. This is the same discipline that separates developers who convert enquiries from those who lose them. A campaign optimising toward a handful of real conversions beats one optimising toward plenty of fake ones.
- Consolidate the structure once, deliberately. Fragmented ad groups each hold too little data for the algorithm to learn from. Restructuring restarts learning, so do it in a single move and then leave it alone long enough to accumulate signal.
- Maintain negatives continuously. Broad match and AI Max expand your query set whether or not anyone is watching, and the search terms report will not show you all of it. Build from what you can see and assume the invisible tail is wider. If your campaigns fall into the September auto-upgrade, get the list in order before it lands.
Then hold a review cadence you can actually keep, and resist editing between reviews unless something is genuinely broken. That restraint is the strategy. The trap closes when a manager, watching a campaign that cannot exit learning because its conversions are too thin, keeps intervening to fix it, and each intervention guarantees another cycle of the same. Applying a batch of recommendations feels like the opposite of neglect, and produces the identical result. Breaking out starts with the conversion data, not the bid strategy.
If you want to know where your own account sits in that loop, what is being auto-applied, what keeps resetting, and where the conversion signal is breaking, that is what an audit is for. It surfaces something more specific than an optimisation score ever will.
