- Google can only bid on the business value you send. Decide whether your campaigns should maximize revenue, expected customer value or contribution margin.
- Performance Max supports campaign-level negative keywords and up to 50 search themes per asset group. The limit is not a target.
- Google has announced that the automatic upgrade of Dynamic Search Ads to AI Max will begin in February 2027.
- Derive values for inquiries from close rates and contribution margin instead of assigning arbitrary points.
More clicks do little good if they don't lead to relevant inquiries or profitable orders. If you want to improve Google Ads, first check what the account counts as success, where the budget is spent and which results actually pay off.
These 100 tips show you where to start: account structure, bidding, search terms, ads and measurement. Begin with the points that solve your biggest current problem. You don't have to change everything at once.
Set up your account and goals properly
1. Separate campaigns for a good reason
Separate campaigns when different budgets, regions, offers or business goals need their own controls.
For closely related services, separate ad groups are often enough. A highly granular structure is not a sign of quality.
For each existing campaign, write down why it stands on its own. If there is no clear reason, consider merging it. Keep campaigns separate, for example, where business units are required to have their own budgets.
2. Analyze brand searches separately
People who search for your company already know you.
Evaluate these searches separately from generic terms. Otherwise, cheap brand conversions can hide the fact that actually winning new customers costs much more.
Compare cost, revenue and the share of new customers for brand and generic searches. A dedicated brand campaign and suitable brand exclusions in Performance Max make the separation easier. Check in your shop or CRM whether additional customers are actually being won.
3. Choose the right value to optimize for
Google can only bid on the business value you send.
If products have different margins, the same revenue does not mean the same profit. So decide deliberately whether your campaigns should maximize revenue, expected customer value or contribution margin.
Work with your finance department to define which costs go into the calculation. Send these values consistently and document what they mean. A conversion value reported as contribution margin must not be read as revenue in a report later on.
4. Review bid strategies for shared goals
A portfolio bid strategy can bring several supported campaigns together under one shared goal.
This is an option when the campaigns deliver comparable results and, for example, pursue the same target cost per inquiry.
Check close rates and order values before combining campaigns. Two regions with similar profitability can fit together. Clearly different customer values, on the other hand, call for separate targets. A shared budget is a separate decision and not automatically part of the portfolio strategy.
5. Choose auto-applied recommendations deliberately
Google Ads can apply certain recommendations automatically.
These include changes that affect your targeting, ads or bids. So check which categories are turned on in your account and what effects they can have.
Review the change history regularly. Only accept automations whose effects you can understand and monitor. For strategic changes, it should be clear which problem they are meant to solve and which metric you will use to judge success.
6. Give each campaign type a job
A Search campaign for specific purchase queries does a different job than a video campaign that introduces a product that needs explaining.
Before launch, decide what a campaign should contribute.
Assign a suitable success metric to each job. For direct lead generation, that might be the cost per qualified inquiry. For a product launch video, you also need a sensible way to assess later demand and conversions.
7. Identify new customers based on reliable data
If you want to bid higher for new customers, your measurement has to tell new and existing customers apart as reliably as possible.
Check your customer lists and how new buyers are flagged before you set an additional new customer value.
Regularly compare a sample against your actual customer base. Watch out for duplicate records and buyers with several email addresses. Derive the additional value from repeat purchases and profit instead of entering an arbitrary bonus.
8. Only use suitable conversion goals for bidding
Check for each campaign which actions are actually used for bid optimization.
A page view, a form start and a qualified inquiry carry different weight. They should not be treated the same just because they can all be measured.
Go through the goals of your most important campaigns one by one. Also keep custom goals in mind: they can use secondary actions for bidding. What matters is not just the setting of the individual conversion action, but also how it is used in the campaign goal.
9. Evaluate efficiency and growth together
A strict target ROAS can limit delivery.
Loosening it can enable additional sales, but can also lower their average profitability. So evaluate both the ratio of value to cost and the absolute profit.
Before making a change, write down how much additional contribution margin would be needed. If €500 in additional advertising brings only €300 in additional contribution margin before ad costs, the result gets worse. Rising revenue alone does not justify higher spending.
10. Understand attribution as allocation
Data-driven attribution distributes measured conversions across the ad interactions it considers.
It helps account for earlier interactions that would get no credit if you only looked at the last click.
Use the model comparison to reveal larger shifts. However, this is not yet solid proof of additional advertising impact. If you want to shift larger budgets based on the analysis, back up the decision with a suitable test where possible.
11. Define a test question before major changes
Before a major change, formulate a specific hypothesis.
For example: a new landing page lowers the cost per qualified inquiry. Before launch, decide which metric is decisive and what budget you will use for the test.
Change only the factor you are testing, if possible. If the bid strategy, offer and landing page all change at once, the result is hard to attribute. The required run time depends on data volume, fluctuations and the time between click and conversion.
12. Evaluate recommendations by business value
The optimization score can point to possible improvements.
But it does not show whether your account is profitable. A higher score is therefore not a good enough reason to apply a recommendation.
For every suggestion, ask what additional results to expect and what costs it creates. For a budget increase, for example, you need a plausible business case. The fact that a campaign could spend more does not mean it should.
Steer Performance Max and Demand Gen with purpose
13. Add relevant audience signals
Audience signals give Performance Max hints about suitable customers.
Use buyers or qualified leads, for example. However, the signals do not limit delivery to the people you add.
Where possible, separate valuable customers from contacts with no recognizable business value. Check updates, consent and sufficient list size. One clearly named, well-maintained data source is more useful than several similar lists of unclear origin.
14. Check the landing pages actually used
With final URL expansion turned on, Performance Max can choose other relevant pages on your website.
This can open up additional opportunities, but can also lead to pages you did not intend for the campaign.
Start by checking the landing pages with the highest spend. Do prospects find the advertised offer there and a sensible next step? Exclude unsuitable sections, and if you use a fixed page selection, also check the settings and product links from Merchant Center.
15. Revise assets with purpose
Text, images and videos should communicate the offer accurately and clearly.
Use the asset reports to spot possible weaknesses. However, the metrics of a single asset do not cleanly isolate its contribution from the other parts of an ad.
Before replacing anything, note which problem you want to fix: a poor crop or an unclear benefit, for example. Make targeted changes and then also assess how the asset group develops. Replacing assets every four weeks on a fixed schedule is not a goal in itself.
16. Test Shopping on its own with Standard Shopping
If you want to examine Shopping independently of the rest of the Performance Max inventory, consider a Standard Shopping campaign.
Performance Max without your own additional assets is not a reliable setup for Shopping only.
Before the comparison, define products, target regions and business goals. Take overlaps with existing campaigns into account. Otherwise, only the attribution of sales may change, while the total number of orders and the profit stay the same.
17. Exclude irrelevant searches deliberately
Performance Max supports campaign-level negative keywords and the corresponding lists.
The exclusions apply to Search and Shopping inventory. Only terms that never fit your business belong at the account level.
Derive exclusions from what you actually offer. For an office furniture retailer, free craft templates can be clearly irrelevant. “Used,” on the other hand, should only be excluded if the retailer does not sell used furniture. After major changes, check the remaining demand.
18. Provide your own videos for key formats
Check which automatically created videos Performance Max uses.
Add your own versions if they show product, benefit and brand more clearly. Design landscape and portrait formats deliberately for each placement.
Start with a simple sequence: show the offer, explain one specific benefit and name the next step. Read the text overlays on a smartphone. If the message is hard to understand there, simplify the video before using it.
19. Test Demand Gen with a clear offer
Demand Gen can make sense if you want to explain an offer visually or reach prospects before they run a specific search.
Plan a separate test budget for it and a clear job within your customer acquisition.
Check in the account which channels are actually selected, including the Display Network. Send the ad to a suitable landing page. Beyond reach and clicks, assess whether suitable prospects come in and what it costs to reach a qualified result.
20. Test lookalike audiences with good seed data
Lookalike segments in Demand Gen are built on a seed list.
If possible, start with people whose value you know, such as buyers or qualified leads. All website visitors are not automatically the best basis.
Compare the available similarity settings against your business metric. Also check whether audience expansion changes the targeting you want. Greater similarity to existing customers guarantees neither better inquiries nor cheaper new customers.
21. Investigate overlaps with Search campaigns
If your Search campaigns lose volume while Performance Max grows, check the affected searches.
An identical, eligible exact match keyword generally takes priority. Budget, approvals and other conditions can still limit delivery.
So first check whether the Search campaign can participate without restrictions at all. Only then add further exclusions. Two campaigns moving in opposite directions does not, on its own, prove that one campaign is taking valuable demand from the other.
22. Derive search themes from real customer inquiries
Search themes give Performance Max additional hints about relevant demand.
Up to 50 themes are possible per asset group. The limit is not a target: a few precise themes can be more useful than many similar terms.
Use wording from search reports, customer inquiries and sales conversations. A provider of storage systems could, for example, pick up specific requirements for small warehouses. Search themes add to automated targeting; they do not limit the campaign to these searches only.
23. Use the channel report for troubleshooting
The channel report shows how delivery and measured results are distributed across different Google channels.
Use it to spot unusual shifts. Keep in mind different interaction types and conversions that come in later.
Compare an unusual period with the period before it. Were videos added, goals changed or promotions launched? Check such connections before you write off a channel or assume that new audience signals will solve the problem.
24. Only plan for store visits when the requirements are met
A linked Business Profile is not enough to measure store visits.
Among other things, Google requires suitable locations and sufficient data volume. So first check whether the corresponding conversion action is available in your account.
Also distinguish between direction requests, calls, and store visits. These actions carry different meaning and should not all get the same value by default. A direction request shows interest in your location, but it is not yet a confirmed purchase.
25. Control brand suitability with the right exclusions
Define which content or environments are unsuitable for your brand.
Then check which inventory, content, or placement exclusions are actually available for each campaign type.
Start with clear-cut cases and check the scope of the setting. Not every exclusion applies to all Google surfaces. Keep lists manageable and document their purpose so they can be maintained in a traceable way later.
26. Review search terms by cost and relevance
Sort the search terms report by cost and start with the most expensive queries.
Separate clearly wrong search intent from searches that fit in principle but deliver weak results. The two cases call for different actions.
For the wrong intent, an exclusion can help. For relevant queries, you should also check the ad, offer, landing page, and measurement. Account for delayed conversions and the limits of the report: not every individual search query is shown, for privacy reasons.
27. Build asset groups around related offers
Group products or services that share the same text, images, and landing pages.
For example, one group for running shoes and one for hiking boots can make sense. The structure should allow ads that are easy to understand.
Assign each group suitable landing pages and creatives. If a text could stand equally for any of your offers, make it more specific. Grouping only by assumed audiences does not create a firm separation, because audience signals do not limit delivery accordingly.
28. Use seasonality adjustments for unusual changes
Seasonality adjustments can prepare Smart Bidding for significant, short-term changes in conversion rate.
Google cites one to seven days as the typical use case. Normal seasonal trends do not need to be announced separately each time.
Estimate the change based on comparable promotions. What matters is the conversion rate, not just an expected increase in revenue. More orders can also come from more visitors. So check whether the promotion actually changes the share of visitors who buy.
Manage bids and budget profitably
29. Consider value-based bidding when values differ
If orders or inquiries differ in economic value, optimizing for conversion value may fit better than optimizing for the number of conversions alone.
For this, the values you send must reflect the intended difference in a traceable way.
Before switching, compare some reported values with shop or CRM data. Check currencies, outliers, and possible double counting. A bid strategy does not correct faulty values; it uses them for its decisions.
30. Derive your target ROAS from comparable data
Use a period with reliable values and enough completed purchase processes.
Take promotions, price changes, and shifts in your product range into account. A target based on an unusually strong promotion month may not fit your normal business.
If a campaign barely serves, check whether the ROAS target fits past performance and current demand. Treat the most recent days with caution if purchases often come in later. Otherwise you are reacting to results that are still incomplete.
31. Choose between budget and efficiency target
“Maximize conversions” can fit if you want to achieve as many relevant conversions as possible within a set budget.
Target CPA adds an intended average cost target. This value is not a cap for each individual conversion.
Derive your target CPA from your economics. For inquiries, this includes the rate at which they turn into customers. Set the budget so that you can also absorb temporary fluctuations, and check the quality of the results achieved.
32. Don't automatically optimize new campaigns for clicks
If conversions are your goal and measurement works, conversion-based bidding is an option from the start.
Google does not require a conversion history for target CPA. There is no general rule that you must first collect 30 conversions with a click strategy.
“Maximize clicks” can make sense for a limited traffic test, but it optimizes for clicks. Decide beforehand what insight you want to gain. Ideally, start with a clear offer and a few unambiguous goals so the results can be assessed.
33. Plan price promotions and budget together
A discount can bring more sales and at the same time lower the profit per order.
So before a promotion, check how much contribution margin remains after the discount and additional fulfillment costs.
Then determine how much room you have for advertising. Only adjust targets if the calculation supports it. Align the duration and end of the promotion with ads, product data, and landing pages so that outdated offers are not still being advertised.
34. Cushion tracking outages with data exclusions
If conversion data is missing or was reported incorrectly because of a technical error, a data exclusion for Smart Bidding can make sense.
First check the cause instead of treating a normal drop in demand as a measurement problem.
Take the time between click and conversion into account. Click days before the outage can be affected too. Document the period, campaigns, and devices. Data exclusions change the information used for bidding, but they do not repair historical reports.
35. Use shared budgets deliberately
A shared budget can suit supported campaigns that pursue the same business goal and are allowed to share funds flexibly.
First check whether such an allocation matches your internal guidelines.
If budgets are fixed per region or business unit, a shared allocation may be unwanted. Also keep budget and bid strategy separate: a portfolio strategy manages bids, while the shared budget pools the available spending limit.
36. Spot ineffective bid adjustments
Smart Bidding takes many signals into account on its own and does not use many manual bid adjustments.
So check which settings apply to your specific strategy before you assess their impact.
If mobile devices perform worse, first look at the landing page and conversion quality. A form that is hard to use is a different problem from a lower customer value. With target CPA, device adjustments can change the device-specific target; other strategies have other rules.
37. Calculate the profitability of additional spend
The average CPA does not automatically show what additional customers cost.
So when you raise the budget, compare the additional spend with the additional qualified results. €1,000 in extra cost for ten more customers equals €100 in additional ad cost per customer.
Use periods that are as comparable as possible. If prices, promotions, or other channels change at the same time, attribution becomes uncertain. In that case, treat a simple before-and-after comparison as a guide, not as exact proof of impact.
38. Use value rules for proven differences in value
Conversion value rules can reflect economic differences by location, device, or audience.
They make sense when additional value is not yet included in the reported conversion values. A higher conversion rate alone does not justify an uplift.
Document which data the weighting is based on. For example, a demonstrably higher order value in one customer segment can be relevant. Review the rule again if customer groups, prices, or close rates change significantly.
39. Document changes one by one
For important changes, note the date, starting point, action, and expected effect.
This helps you put developments into context later. A simple spreadsheet is enough if it is maintained consistently.
Choose the size of a budget change based on your data and what you can afford. There is no general percentage rule that reliably prevents a new learning period. Avoid several unfounded corrections in a row just because a single day was weaker.
40. Distinguish between daily budget and monthly limit
The average daily budget is not a rigid daily limit.
For most campaigns, charged daily costs can be up to twice this budget. If the budget stays the same for a full month, a monthly limit of 30.4 times the daily budget usually applies.
Check the budget report and take changes during the month into account. Campaign launches, adjusted budgets, and special settings affect the calculation. Set up regular spend monitoring so deviations become visible early.
41. Send qualified leads back from your CRM
A completed form says little about whether an inquiry fits your offer.
Define with your sales team when a lead is qualified and when an actual deal is closed.
Send these results to Google Ads regularly and check import diagnostics and attribution. Start with a few unambiguous status values. Avoid carelessly double-weighting the form and the later deal as equal successes.
42. Judge brand advertising by its contribution
A brand campaign can help you control how you appear in searches for your company.
But also check organic visibility, competitors, and the actual cost of this additional presence.
Separate purchase-ready queries from support, job, or other search intent. A target impression share can support a visibility goal, but it does not guarantee the top position. Set an economically justifiable framework for budget and bids.
Manage keywords and search terms cleanly
43. Test broad match keywords in a controlled way
Broad match keywords can open up additional search queries.
Combine a test with reliable conversion measurement and a suitable Smart Bidding strategy. Keep your offer and landing pages clear.
Start with a clearly defined topic area and a set cost limit. Compare qualified results and search terms with your previous approach. Smart Bidding can also buy irrelevant clicks; automation does not replace oversight.
44. Pay attention to meaning with phrase match
Phrase match is not limited to a fixed word order.
Delivery can also cover phrasings with a similar meaning. So check the actual search queries, not just how your keywords are spelled.
Look for recurring misinterpretations. If a commercial cleaning service mainly gets job applications, the campaign needs a clearer boundary. Besides exclusions, also check whether the ad and landing page clearly describe the service offered.
45. Use exact match keywords for important search intent
Exact match can help you link key search intent closely to one ad and landing page.
Still, this option is not a strictly literal limit. Close variants and searches with the same meaning can also be matched.
Start with terms for which you have a compelling offer. Check demand, cost, and lead quality. A narrow term is not valuable just because it is specific.
46. Maintain exclusion lists according to their scope
Separate terms that are irrelevant in general from exclusions for individual offers.
A word can be irrelevant for one campaign and central to another. So only clearly universal exclusions belong at the account level.
Name lists by their purpose and check where they are used. Also keep match types in mind: exclusions do not automatically cover all synonyms or singular and plural forms. Document larger changes so accidental blocks remain traceable.
47. Evaluate recurring patterns, not individual flukes
An n-gram analysis groups recurring words or word sequences in search queries.
This lets you spot patterns that are missed when you look at individual search terms.
Evaluate cost and results across enough cases and read some queries in context. Words like “reviews” or “test” do not always stand for the same intent. An exclusion should be based on a solid pattern, not on a few weak clicks.
48. Choose the location option that fits your business
For local services, targeting people in or regularly in the target area can be a sensible starting point.
For hotels or travel offers, on the other hand, people interested from outside the target area are often relevant too.
Compare the locations of your leads with your actual service area. An inquiry from another city can fit if the job takes place in the target area. What matters is whether you can provide the requested service.
49. Use Quality Score for troubleshooting
Quality Score, from 1 to 10, is a diagnostic metric and is not used directly in the auction.
Its components still help you identify weaknesses in expected click-through rate, ad relevance, and landing page experience.
Start with important keywords that have below-average component ratings. If the ad does not match the search intent, work on the message and the assignment first. Then judge the improvement by qualified results and cost, rather than just raising the number.
50. Clear unnecessary variants out of keyword lists
You don't need to add every spelling and every word combination individually.
Keep variants where they allow meaningful control or analysis. The length of a keyword list is not a problem in itself.
Check terms with the same intent, ad, and landing page for unnecessary repetition. Document larger cleanups and then check delivery. Don't remove important variants just to make the account look as small as possible.
51. Prepare the switch from DSA to AI Max
Google has announced that the automatic migration of dynamic search ads to AI Max will begin in February 2027.
Before then, save successful search terms, landing pages, and exclusions as a baseline for comparison.
Test the new configuration with clear rules for landing pages and text. In particular, check new page assignments and automatically generated messages. Other legacy features follow different timelines, so check the notices for your specific campaign.
52. Test competitor terms separately
If you want to use searches for other providers, plan a separate budget and a suitable landing page for them.
Explain clearly what your offer does for these prospects.
Evaluate the campaign separately from generic terms and your own brand. Check whether these contacts are actually looking for an alternative or just want to reach the other provider. Also clarify exactly how other companies' trademarks are used in the ad and on the landing page.
53. Don't rush to add singular and plural separately
Positive keywords can cover singular and plural variants.
Adding them separately is therefore not always necessary. What matters is whether it brings a practical benefit for targeting, ads, or analysis.
Don't infer differences from blanket assumptions about purchase intent. Check the actual search queries. For negative keywords, on the other hand, you need to account for the relevant forms separately if both singular and plural should be excluded.
54. Use locations only for services you actually provide
Location insertion can give the ad a local connection.
Only use it if your offer is available in the locations concerned. An automatically inserted city must not create false expectations.
Check several possible ad variants and the corresponding landing pages. Make it clear whether you work on site, deliver, or advise remotely. Also add a sensible default text for cases where no suitable insertion is possible.
55. Filter out unsuitable prospects in the text
State key requirements as early as possible, such as “businesses only,” a minimum order quantity, or a realistic starting price.
This gives prospects important information before they click.
Use the most common reasons for unsuitable inquiries reported by sales. Then check whether their share goes down. A lower click-through rate can be acceptable if the right results become more profitable as a result.
Use your own data and audiences wisely
56. Update customer lists regularly
Customer lists can support reaching existing customers, exclusions, and audience signals.
For that, it must be clear who belongs on each list and how current the data is. An outdated list can undermine the intended control.
Define how updates happen and who is responsible. For example, distinguish between active contract customers, one-time buyers, and former customers. Automate the sync if manual maintenance becomes unreliable; a daily upload is not necessary for every business.
57. Collect data for a clear purpose
Your own customer data is useful when it serves a specific task.
For each source, describe whether it is meant to attribute closed deals, reach existing customers, or serve as an audience signal, for example.
Check usage rights, consent signals, and technical transfer. More data does not automatically lead to better campaigns. Freshness, formatting, and business relevance matter more than the largest possible, unspecific collection.
58. Add GA4 audiences selectively
Linking Google Analytics 4 with Google Ads can make sense if you want to use audiences or additional reporting from it.
It is not a general requirement for conversion tracking with Google Ads.
Start with a specific question and check the events it requires. Predictive audiences are only available when the requirements are met. An audience built on unreliable events carries that uncertainty into your campaign work.
59. Observe a few relevant audiences
With the “Observation” setting, you can analyze audiences in Search campaigns without restricting reach.
Choose segments whose results could actually lead you to a decision.
For example, start with existing customers and a few relevant groups. Only look at differences once there is enough data. Keep overlaps in mind: several observed audiences are not automatically independent customer groups.
60. Base demographic exclusions on results
Don't exclude age groups based on a hunch alone.
Check costs, qualified results, and the amount of data available. The “Unknown” category can also contain a relevant share of demand.
Before excluding a group, read through some of its actual inquiries or orders. A few bad cases are not a reliable basis. If your offer has mandatory restrictions, however, the targeting must be right from the start; estimated demographic characteristics do not provide complete protection.
61. Use audience expansion deliberately
Check whether optimized targeting or another expansion is allowed to serve outside your selected audience.
This is especially important if a campaign should only reach past visitors or existing customers.
When testing, document whether expansions are allowed. Otherwise you will compare supposedly separate approaches that reach similar people. Decide based on additional qualified results whether broader delivery makes sense for your offer.
62. Connect life events with a suitable offer
A life event can be a plausible reason for advertising.
For a furniture store, for example, this could be a move. First check which suitable segments your campaign type actually offers.
Connect the targeting with a concrete benefit, such as a furnishing solution for the new home. Compare the results with an alternative without this restriction. Assignment to a segment is an estimate, not confirmed proof of the event.
63. Interpret custom segments correctly
Search terms in custom segments can describe relevant needs.
On Google-owned properties, such segments can reach people who searched for these or similar terms. On other properties, entries are interpreted differently, for example as interests.
Group terms around a specific need and align the ad with it. A software provider could use typical questions about replacing an existing solution. This does not give you a precisely defined list of everyone who searched for a particular term.
64. Check app placements based on their quality
For Display campaigns, examine which apps generate costs and useful results.
A gaming app is not automatically worthless. Conversely, cheap reported conversions do not prove that suitable customers come from there.
If results look unusual, check some of the actual contacts. Are the inquiries complete and can you reach them, or are only brief events being triggered? Exclude clearly unsuitable environments instead of treating all apps as equal.
65. Test in-market audiences against alternatives
In-market segments can be a sensible starting point.
But only a comparison shows whether they work better for your offer than other audiences. The segment name does not guarantee an imminent order.
For the comparison, use offers and creatives that are as similar as possible. If audience, price, and message change at the same time, a difference is hard to explain. Evaluate qualified results and take into account the volume you can reach in each case.
Make ads and landing pages easy to understand
66. Write headlines with different jobs
Use the up to 15 headlines of a responsive search ad for the offer, benefits, proof, and calls to action.
Several nearly identical keyword variants often use the available space less effectively than different relevant messages.
Read possible combinations out loud. Each headline must make sense on its own and fit with the others. Add up to four descriptions and only use benefits and proof that actually apply to the specific offer.
67. Pin only the messages you need to
Pin content when its position is required for factual or legal reasons.
Let other parts combine as freely as possible. Pinning all text greatly reduces the possible ad combinations.
Don't rely on positions that don't always show for mandatory statements. Also check automated features: with AI Max, final URL expansion and URL inclusions can affect whether pinned assets are used. So check the entire configuration.
68. Add suitable images to search ads
If your account meets the requirements, image assets can make your offer easier to understand.
Use images that match the advertised service and meet the current requirements. In particular, check crops and how they display on small screens.
Choose images that carry information. For a technical product, this can be a clear product shot; for a service, a real usage situation. An interchangeable image does little if it makes neither the offer nor the difference from competitors easier to understand.
69. Mark up promotions with the correct terms
Use promotion assets for promotions that are actually available.
Align discount, time period, code, and landing page with each other. Also check that the promotion reliably disappears after it ends.
Go through the purchase path yourself. Does the code work, are the right products included, and is any minimum order value clear? An attractive notice does little if the prospect cannot follow the offer after clicking.
70. Make sitelinks lead to concrete next steps
Sitelinks should offer additional orientation.
For example, link to pricing, important product categories, or a relevant consultation page. Use descriptions to make the difference between these destinations clear.
Give each link its own job. Someone looking for prices needs different information than someone who needs advice. Google decides how the ad is actually displayed; a larger ad space or a specific performance effect cannot be guaranteed.
71. Distinguish callouts from snippets
Callouts are suitable for features such as “in-house customer service” or “ships from stock.”
Structured snippets, on the other hand, list values for a predefined category, such as brands, models, or services.
Choose details that make a decision easier. A software provider could list available modules, a retailer the brands it carries. Avoid vague statements like “high quality” if they neither explain nor prove what sets the offer apart.
72. Review AI suggestions before publishing
Generated text and images can help you develop new variants.
Before using them, check product features, prices, spelling, and image details. The output must match your real offer and your website.
Decide who gives the subject-matter approval. Don't adopt certifications or performance promises that the system has merely added because they sound plausible. Then evaluate selected variants using your usual business metrics, not by the number of ads generated.
73. Use ad strength as a pointer
Ad strength shows possible gaps in the amount and variety of your text.
It is a guide and does not directly affect whether a responsive search ad is eligible to serve. Business performance has to be assessed separately.
Don't change a well-performing ad just because of its rating. Check which specific improvement is being suggested. An additional meaningful variant can make sense; a weaker wording just for a better rating does not help your customer.
74. Align ad and landing page with the same offer
Open your most important ads and check the entire path on a smartphone.
Are the offer, price or service, and the next step clear right away? Does the form or checkout process work?
Start with specific obstacles such as contradictory information, missing details, or unnecessary questions. Prioritize pages that already receive a lot of paid visitors. There, an improvement affects a relevant share of your current spend.
75. Test keyword insertion with every variant
Keyword insertion uses the keyword in your ad group that triggered the ad, not automatically the search term that was entered.
So check which wordings can actually appear in the ad.
Read the entire ad with every possible keyword and set a sensible default text. Pay attention to grammar, length, and brand terms. Use fixed text if the possible insertions lead to inconsistent or confusing messages.
Use YouTube with a clear job
76. Connect attention and offer early
Use the ABCD framework as a checklist: grab attention, show the brand, connect with the viewer, and name the action you want.
Show early who the offer is for and why it might be relevant.
Before production, sketch a short sequence. What does the viewer see first, what benefit do they understand, and what should they do next? If you can't answer these questions briefly, simplify the message before filming.
77. Design vertical video on its own terms
Produce or edit a vertical version in which the product, people, and text stay clearly visible.
A simple crop is only enough if the important content is preserved and fits the format.
Check the result on a smartphone with the platform's controls on screen. Have someone uninvolved repeat the message back to you. Pace and music should support the offer; fast cuts are no substitute for a clear message.
78. Make both sound and picture understandable
Plan speech, music, and sound effects deliberately.
Add subtitles or text overlays where they make the message easier to understand. Not every effective video needs a voiceover, but the key information has to get across.
Play the video once without sound and once with sound only. Check which message stays clear in each case. This shows you whether important content is only spoken or whether the audio conveys something different from the images.
79. Match multi-part stories to the format
A video ad sequence can spread different jobs across several ads.
For example, the first ad explains a need, the second the solution, and the third a specific offer. Each step should have a recognizable job.
Check which event triggers the next step: impression, view, or skip. Not every viewer has watched all previous videos in full. So keep individual ads understandable on their own and check whether the budget is enough for the planned sequence.
80. Measure the added impact of video separately
Watch whether branded searches, qualified inquiries, or sales change after a video campaign.
But don't automatically attribute the development to YouTube. Seasonality, promotions, and other channels can also play a part.
Before launch, define which impact you want to examine. Document activities running in parallel. If budget and availability allow, consider a lift study or another suitable controlled test.
81. Check thumbnails at their actual size
A thumbnail has to be understandable at the size it will actually be shown.
Check the subject, product, and text on a smartphone. A face can work, but it isn't a general requirement for a good thumbnail.
Create a few clearly different variants. Shorten text that is only readable at full screen width. Also keep expectations in mind: the video should deliver what the thumbnail promises instead of just driving as many clicks as possible.
82. Use product feeds in supported video formats
Check whether your chosen campaign type can include products from Merchant Center.
Check approvals, prices, availability, and landing pages. The products shown should match the offer being advertised.
Start with a manageable selection of products. A video about a specific model shouldn't lead to a random selection in which that model is hard to find. A linked feed also doesn't guarantee visible product tiles in every placement.
Strengthen Shopping with better product data
83. Keep your product data API up to date
If your store transfers data via API, check whether it already uses the Merchant API.
According to Google, the Content API for Shopping reached its sunset date on August 18, 2026. Without an approved extension, errors may gradually occur since September.
Ask your store or feed provider to confirm which API is in use and how far the migration has progressed. Also check for successful transfers and current prices. The full shutdown is planned for early 2027; classic file feeds are not automatically affected.
84. Write product titles with purchase-relevant attributes
Name the brand, model, and attributes buyers use to tell products apart.
Put important information first and avoid unnecessary repetition. All details must match the specific item.
Use a pattern that fits the product type. For clothing, color and size can matter; for spare parts, model and compatibility. Start with high-revenue or frequently shown products and then check the respective variants.
85. Submit product identifiers correctly
Submit an existing GTIN from the manufacturer correctly.
Don't make up identifiers and don't use the EAN of a similar product. Different requirements apply to items without an assigned GTIN.
Start with the specific identifier issues in Merchant Center. Match variants individually, since size, color, or version can have their own identifiers. Wherever possible, fix the reliable data source so errors don't come back with the next import.
86. Group products with custom labels
Custom labels help you analyze products by your own business criteria.
Suitable examples include margin tiers, season, or bestsellers. Set clear definitions and make sure they are updated reliably.
Check which decision the grouping is meant to enable. A margin label on its own changes neither the reported conversion value nor the bidding logic. An additional campaign only makes sense if it leads to a well-founded different approach to budget or targets.
87. Simplify recurring data fixes with rules
Attribute rules can simplify recurring fixes to product data.
Check which features your Merchant Center setup supports. Every rule needs a clear condition and a clear goal.
Test typical and unusual items before you approve a rule. A blanket addition such as “Sneakers” to all products of a brand would also create incorrect titles. So document which items are affected and how you check the result.
88. Link supplemental data via unique IDs
A supplemental data source can add details that are missing from the main system.
The matching has to work through stable product IDs. After the import, check whether the values reach the right items.
Check the connection again when your product range or IDs change. A small supplemental spreadsheet also needs an owner and regular maintenance. It doesn't replace reliable updates of prices and availability in the primary source.
89. Test promotions in Merchant Center in advance
Before a promotion, check whether the products and target country are eligible for the annotation you want.
Submit the discount, conditions, and, if applicable, the code in good time. Check approval and display.
Afterward, don't just evaluate clicks and orders. Factor the discount into profitability and compare ad costs with the remaining contribution margin. A promotion with more revenue can still deliver less profit.
90. Check AI product images for accurate depiction
With available image tools, you can test new settings and variants.
The real product must remain accurately recognizable. Check shape, color, material, sense of size, and any accessories shown.
Compare the edited image and the original side by side. Pay particular attention to connectors, labels, and parts that aren't included. If the design creates false expectations, choose another variant or use the unedited product photo.
Make measurement and business values reliable
91. Check enhanced conversions technically
Enhanced conversions can improve the attribution of conversions.
Check whether suitable first-party data is processed correctly and sent with the required consent signals. Use diagnostics and a test conversion.
Also check the underlying event: Is a purchase counted more than once, are value and currency correct? Better attribution only helps if measurement is correct. Enhanced conversions don't recover every lost conversion and don't make personal data anonymous.
92. Test consent signals in both states
Test your website after consent is granted and after it is denied.
Check whether the choice is passed correctly to the tags, including the relevant parameters such as ad_user_data and ad_personalization.
Repeat the test after major changes to the banner or tag setup. A visible cookie notice alone doesn't prove correct technical implementation. Consent Mode passes on consent states; it doesn't collect consent itself and doesn't replace a legal review.
93. Use server-side tagging for a specific need
Server-side tagging can improve data control and technical processes.
First define the problem: for example, duplicate events, too many browser tags, or a lack of control over the data passed on.
Decide how you will recognize the benefit, and take operation and maintenance into account. An additional technical layer doesn't automatically eliminate all measurement loss or consent requirements. Compare data quality and page performance before and after the switch.
94. Derive values for inquiries from close rates
For inquiries, use values derived from business data instead of arbitrarily chosen points.
If ten percent of qualified inquiries become customers and their average contribution margin before advertising is €1,000, that initially gives an expected value of €100 per qualified inquiry.
Take other relevant costs into account and use enough closed sales cases. Separate inquiry types if their close rates differ significantly. Update the values as soon as order values or success rates change significantly.
95. Write down what your return metric means
If you send contribution margin as the conversion value, the interface may still show a ROAS metric.
So state explicitly what the numerator contains. Revenue and contribution margin must not be confused.
In this case, a ratio of 100 percent only means that the reported contribution margin equals the ad costs. Fixed costs not taken into account are not yet covered. Only compare periods in which the same value definition was used.
Make checks and reports practical for everyday use
96. Get automatic alerts for broken landing pages
Set up a regular link check for ads, keywords, and sitelinks.
Google's sample script logs errors and sends notifications. Automatically pausing ads is not part of the standard script.
Decide who reviews alerts and arranges any necessary fixes. Distinguish real error pages from short-term server problems. If you add automatic pausing, safeguard it against false alarms and also test the notification path.
97. Adapt alerts to data volume and time of day
Automated alerts can surface unusual costs, missing impressions, or a measurement outage.
Compare similar weekdays and the same point in the day. Small data volumes need different thresholds than large accounts.
Assign each alert an owner and a next step. For missing conversions, for example, check measurement first; for cost spikes, check budget and change history. An alert is useful when it triggers a specific investigation.
98. Check exclusions for unintended blocking
A new negative keyword can also block relevant search queries.
Pay particular attention to general words that appear in different contexts. Keep in mind how negative keywords and their match types actually work.
A negative keyword for “shoes” doesn't by itself block every search containing the compound word “snowshoes”. Test specific cases and check important topics after major changes. This way you spot unintended restrictions before they cost you relevant demand for long.
99. Evaluate placements with the right data
Use the Performance Max placement report mainly to review environments and brand suitability.
It doesn't give a complete overview of costs and conversions per placement. Classic Display reports may contain different performance data.
So check which information a script actually uses before you turn on automatic exclusions. Separate environments that are unwanted for content reasons from placements that perform poorly economically. To make a statement about unprofitable spend, you need suitable performance data.
100. Set up a few meaningful metrics for decisions
Use metrics that support specific decisions: for example, cost per qualified lead, conversion value relative to cost, or contribution margin after advertising.
For each, define which data and definitions are behind it.
Label forecasts as scenarios. Lost revenue can't be reliably calculated from impressions, click-through rate, and cart value alone. At a minimum, conversion rate and sound assumptions about incremental demand are missing. Remove metrics that are regularly discussed without leading to any action.
Understand ads next to AI answers correctly
Ads above and below AI Overviews follow the existing serving and auction rules. Switching to AI Max isn't required just because of this position. Ads directly within an AI Overview are a different matter.
The Google documentation we reviewed lists English-language serving in selected markets for ads within AI Overviews; Germany is not on that list. Check current availability before you base a campaign decision on it.
Start with the most important improvements
For your next optimization round, choose three tasks: one for more reliable measurement, one against unnecessary spend, and one for better ads or landing pages. Note the starting point, the change, and the success criterion. Evaluate the results after a sufficient conversion lag.



