Google Ads – how to create campaigns that convert?

An effective Google Ads campaign starts with a strategy: a precisely defined post-click goal, measurable KPI and a budget that will provide enough data for the algorithm to learn. Skipping this step means that even very good ads can drive visits without sales or leads. In this part, you will organise your priorities: which goal to choose in Google Ads, how to set success thresholds (CPA/ROAS) and how to calculate a starting budget for 30–50 conversions per month. Then you will move on to keywords: where to source them from, how to distinguish transactional intent from informational intent, and how match types and negatives translate into traffic cost and quality. This will help you build a solid foundation for Search campaigns and, where justified, Shopping/Performance Max with a feed. Read on if you want to base your decisions on data, not on “gut feeling”.

Strategy and goals for Google Ads campaigns: how to set KPI and a starting budget?

You will set up a Google Ads campaign strategy correctly when you first define what should happen after the click and which KPI will confirm it. The goal can be a purchase, lead, phone call or store visit, and only then do you choose the campaign goal in Google Ads (Sales/Leads/Traffic), because this affects the bidding strategy and the available campaign types. For a user at the “I’m comparing” stage, answers to questions about price, reviews and delivery are key, rather than general claims such as “best”. In B2B, offers like “audit/quote/consult” often deliver conversions better than “buy now”, because they make the first step easier and shorten the path to contact.

Set KPI in three thresholds so you can immediately see when the campaign “breaks even”, when it can be scaled and where the ceiling for maximisation lies. For e-commerce, you can assume ROAS: minimum 400%, target 600%, stretch 800%, and in lead campaigns CPA: minimum 120 zł, target 80 zł, stretch 60 zł. Such ranges organise budget decisions and the intensity of optimisation, instead of ending up with nervous adjustments every few days. Without a clearly defined CPA/ROAS, it is difficult to choose bids consciously and assess whether conversion growth remains profitable.

Calculate your starting budget “from the bottom up”, meaning from the number of conversions needed for learning, not from how much you “want to spend”. For the algorithm to have something to work with, aim for 30–50 conversions per month per campaign or bidding strategy, because otherwise learning will be slow and results unstable. If the expected CPA is 80 zł, the test budget is 2400–4000 zł per month (80 zł × 30–50). This approach reduces the risk of drawing conclusions too early from too small a sample.

Choose the campaign type according to the goal and intent, because different formats “capture” different types of demand. At the start, for purchase intent, Search usually works best (queries such as “buy”, “price”, “shop”) as well as Shopping/Performance Max with a feed, while in local services Search with location and call extensions works well. Treat demand-generation channels such as YouTube/Discovery as support, not the sole source of leads, if stable conversion is what matters to you. Query mapping to TOFU/MOFU/BOFU also helps, because you will set up texts and bids differently for “[product] price” than for “how to choose [product]”.

Set your test priorities so that you first move the levers with the biggest impact on financial results. First test keyword match types, intent, the landing page and the offer, and only later the micro-elements, because differences in keyword selection can change CPA by 30–200%. At the same time, check industry restrictions and policies if you operate in areas requiring verification or with restrictions (e.g. financial, medical, supplement services), to avoid disapprovals and delays. Finally, simplify the post-click decision: instead of many options, show 3 variants (Basic/Pro/Enterprise) and clearly describe “what’s included” to reduce friction in the conversion path.

1. 01Define the KPI after the clickWhat should happen? Purchase, lead, phone call.
2. 02Match the campaign goalSales, Leads, Traffic. It affects strategies.
3. 03Adapt to the userStage “I’m comparing”? Give price, reviews, delivery.
4. 04Low entry threshold (B2B)Audit, quote, consultation. Shortens the path.
5. 05KPI in three thresholdsReturn, scaling, maximisation. You can see the effects.

The key is to define the goal precisely, match the message to the user’s stage and set logical KPI thresholds.

Keyword research: how to effectively match keywords to user intent?

Effective keyword research involves collecting keywords from several sources and assigning them to intent and the correct landing page. In practice, use Google Keyword Planner, the “Search terms” report (if you already have traffic), Google Search Console and autocomplete in the search engine. For e-commerce, add feed data (product names and attributes), because users often search by model, e.g. “Dyson V15 filter”. This helps you identify queries more quickly that genuinely match the offer and have conversion potential.

Matching keywords to user intent is based primarily on distinguishing transactional and informational queries. Transactional keywords answer “where can I buy / how much does it cost / is it available” and usually deliver the lowest CPA, so they often should form the core of the campaign. Informational keywords can work if you direct users to a guide with a strong CTA (e.g. “book a measurement”, “download the checklist”), because otherwise it is easy to burn budget on traffic without intent. Intent and page relevance win first, and only then do “nice” ad messages.

Use keyword match types in a way that lets you control traffic quality, especially at the start of a campaign. Exact match is less “exact” today than it used to be, so check the search terms report and negatives every 2–3 days at the beginning. Broad match makes sense when you have plenty of conversions and good signals (audiences), because otherwise it can expand traffic into peripheral queries and raise CPA. This approach helps maintain stable results before the algorithm and the account collect enough data.

Negative keywords create a quality filter that cuts out intents that almost never convert. Prepare account-level negative lists (e.g. “free”, “used”, “OLX”, “jobs”, “course”, “instructions”) and separate negatives dedicated to campaigns. If you sell subscription software, add the negatives “crack” and “torrent”, because such queries usually do not end in a purchase. Well-planned negatives protect budget faster than any later bid optimisation.

Cluster keywords into the structure by intent and landing page, not solely by linguistic similarity. When you have separate pages for “air conditioning installation” and “air conditioning service”, split those keywords, because the user is looking for a different service and a different price point. Before launch, check CPC estimates and volume in Keyword Planner so you do not build a campaign on keywords with 10 searches per month. If the niche has CPC of PLN 20–40 and little traffic, plan for long-tail queries (“[product] for allergy sufferers”, “[service] with on-site visit”), so the budget has somewhere to “catch” clicks.

Use audience signals to reinforce intent, not to guess demographics. Add audience observations (In-Market, Custom Segments) and check which groups lower CPA, because that gives you concrete optimisation directions. Run the brand campaign separately, because it usually has the cheapest CPC and the highest conversion rate, but it should not cannibalise the budget of non-brand campaigns — it is worth measuring incrementality, especially if you have strong SEO. Competitor keywords can work, but they often have a poorer Quality Score and a higher CPC, so it is wiser to build comparative messaging (“alternative to…”) and direct users to a comparison page to better match intent.

Campaign account structure: optimal split by products, services and margins

The optimal split of a campaign account consists of separating campaigns according to business logic (products/services) and differences in margin, seasonality or LTV. This way each category gets its own budget and does not compete for the same limit with an offer that is completely different in profitability. If one product has a 10% margin and another 40%, keeping them in one campaign encourages the “pushing out” of cheaper conversions, which may not be the most profitable ones. Profitability is safest to manage when budgets reflect real business priorities.

You will set up the ad group structure correctly when one group answers one user question and leads to one, ideally dedicated, landing page. Combining topics such as “interior design” and “turnkey finishes” in one group forces a generic ad, lowers CTR and relevance, and as a result increases CPC. This split also makes the next decisions easier: which queries and landing pages deliver results and which need separate messaging. In practice, this means consistent matching: intent → ad group → ad → landing.

Set targeting so that you pay for users who can actually use the offer, especially for local services. If you only serve a travel area, choose the location setting “Presence: people in or regularly in your targeted locations” instead of “interest”, and narrow the reach (e.g. a 15–25 km radius) so you are not funding clicks from outside the market. Align your schedule with the hours when conversions happen (e.g. 8–20), and for phone leads remember that missed calls raise the real CPA even if “conversions” are visible in the system. The most common budget leak in structure is traffic outside the service area or outside the hours when you handle enquiries.

Base device and demographic decisions on data, not on “default” settings. Before lowering bids on mobile, check whether the landing page loads quickly and whether the form is convenient (often UX is to blame, e.g. a form that is too long), because refining the page can lower CPA more than cutting traffic. Demographics are often incomplete or inaccurate, so only exclude them when you have hard data (e.g. 500+ conversions and clear CPA differences), and in B2B the query intent often matters more than age. For as clean data as possible at the start, in Search campaigns you usually switch off the “Display Network”, and when working at scale use Google Ads Editor and MCC to implement changes and standards more quickly.

1. 01Logical splitProducts/services, margins, seasonality.
2. 02Separate budgetsPrevent competition for the limit.
3. 03Profitability managementProfitability over volume.
4. 04One group, one goalOne question, one page.

Profitability is safest to manage when budgets reflect real business priorities.

Creating effective ads: how to use RSA and ad assets?

Effective Google Ads ads are built when you use RSA and ad assets to match the message as closely as possible to the user’s intent and to shorten the path to contact. In RSA, prepare 10–15 headlines in several variants (benefit, feature, proof, offer, urgency), and pin only when necessary (e.g. legal requirement or a fixed name), because too much pinning limits testing. Tailor the message to the query: for “[product] price” the user expects specifics (e.g. a price range or “from £X” with conditions), and for “best [product]” proof works better (e.g. a rating and number of reviews) than a simple quality promise. RSA works best when you give the algorithm different “angles” of argument, while keeping consistency with the search intent.

The CTA should result from the decision stage, because a different call to action works in e-commerce and a different one in high-ticket services. Instead of “buy now”, in services “book a 15 min consultation” or “get a quote in 2 min” is often more effective, because it lowers the entry barrier at the first step. In e-commerce, CTAs such as “Check availability” and “Delivery tomorrow” can lift CTR, because they remove doubts about timing. At the same time, the differentiators must be measurable (e.g. “dispatch within 24h”, “14-day returns”, “installation in 72h”, “service 7 days/week”), because vague claims like “high quality” rarely answer the question “why should I click?”.

Add ad assets right away, because they genuinely increase the ad’s footprint and make it easier for the user to choose the next step. Most often it is worth implementing:

• sitelinks (e.g. “Pricing”, “Reviews”, “Delivery and returns”),
• callouts and structured snippets,
• call and location assets in local services, which shorten the path to contact.

In practice, these elements not only improve CTR, but also lead the user to the right information (e.g. pricing or reviews) without having to search the site. This is particularly important when the user is comparing offers and needs “proof” and specifics before leaving a lead.

Test creatives methodically, because only then can you be sure what genuinely improves CPA/ROAS. Check one hypothesis at a time (e.g. “price in the headline” vs “delivery time”) and assess the results based on business outcomes by ad group, not solely by the “Ad strength” metric. Keep consistency in mind: if the ad promises “20% off your first order”, the landing page should immediately show the terms and how to use the code, without having to search for them, because inconsistency increases bounce rates and lowers Quality Score, which ultimately raises CPC. If a campaign is “not delivering”, first check whether the promise in the ad matches what the user sees after clicking.

In Performance Max, prepare a minimal set of assets so you do not hand over creative quality to automated generation. The recommended package is 5–10 headlines, 5 descriptions, 15 images in different ratios and 1–3 short videos (e.g. 10–20 s), because a lack of video usually ends up with weaker materials being generated. In e-commerce, the Merchant Center feed is crucial: without correct titles and attributes, the algorithm matches queries and placements less effectively. At the same time, keep an eye on policies: avoid superlatives without proof (“best”, “No. 1”) and check the reasons for disapprovals, because sometimes the problem lies with the landing page itself (e.g. no privacy policy or unclear prices).

Landing page and conversion path: how to match the page to the query?

You match the landing page to the query when, after clicking, the user immediately gets an answer to their intent (e.g. “calculator”, “quote”, “delivery times”, “comparison”). Instead of sending all traffic to one general page, prepare landings for specific needs, because this shortens the time to decision and increases the conversion rate. Such a split works particularly well when queries differ in expectations around price, scope of service or contact method. The closer you are to what the user is asking in the search engine, the less friction there is in the conversion path.

Page speed affects conversions because users drop off when a landing page loads on mobile in 6–8 seconds. In practice, measure the results in PageSpeed Insights and aim for LCP < 2.5 s by reducing heavy scripts, compressing images (WebP) and leaving only the essential tags. It is a simple way to improve results without raising bids in the campaign, because traffic stops “leaking” before it sees the offer. If the landing page is not delivering on mobile, the problem is often not the ad, but the page’s performance and UX.

Set up the form so that it requires only the data needed to take the next step, instead of full qualification right from the start. It is worth asking yourself how much information is actually necessary to call back; often a name, phone number and one choice option are enough. Every additional field can reduce conversions by a few to several per cent, so in leads test short forms and move clarification to the conversation. When the user’s decision is not obvious, use micro-steps such as “check availability”, “choose a date”, “leave your phone number”, instead of one extensive form.

You can reduce user-side risk when you show trust signals on the page and transparent offer terms, including prices as a quality filter. In practice, reviews (e.g. from Google), client logos, certificates, as well as clearly described return or warranty terms work well. In services, the lack of a price often acts as a brake, so ranges (e.g. “from 299 zł”, “packages from 1500 zł”) often improve lead quality, even if the number of leads falls. In addition, use UTM parameters to maintain basic consistency and, if you have the technical capabilities, to personalise the headline in line with the campaign (e.g. location and delivery-time promise).

1. 01Response to intentDirectly and precisely
2. 02Dedicated landing pagesFor specific needs
3. 03Loading speedCrucial on mobile

Shorten the path to decision by answering precisely what the user is asking about, and optimise speed to minimise friction.

Measurement and attribution: how do you configure GA4 and Google Tag Manager?

You will configure GA4 and Google Tag Manager properly when GA4 collects behavioural analytics, while tag deployments are carried out through GTM and you verify that key events fire only once per conversion. In practice, check events such as purchase, generate_lead or phone_click, because duplication can understate the reported CPA and mislead the algorithm. Thanks to GTM, you do not need to edit the site code every time something changes, which helps keep things tidy and roll out fixes without delay. First get to clean, non-duplicating measurement, and only then assess campaign performance.

Import conversions into Google Ads consistently within one model, because mixing sources and definitions throws reports out of line and makes decision-making in bidding strategies more difficult. In campaigns aimed at bid optimisation, optimisation based on conversions from Google Ads (Ads tag) or import from GA4 usually works better, but the choice should remain consistent across the account. If you care about fuller conversion attribution in the reality of limited cookies, implement Enhanced Conversions, because this often increases the number of attributed purchases or leads. When the data is “dirty” or inconsistent, tCPA and tROAS optimise on a distorted signal.

You will configure call tracking reliably when you enable Google forwarding number and count the conversion based on call duration (e.g. >60 s), rather than on the click itself. This is particularly important for phone leads, where a “click” does not yet translate into a real conversation. If you use a CRM, import offline conversions (e.g. from HubSpot or Pipedrive) to assess lead quality and optimise for those wins, not for all of them. As a result, budget and bid decisions are based on sales or valuable leads, not on enquiries alone.

You will set attribution sensibly when you choose the model (often data-driven) and a realistic conversion window matched to the decision-making cycle. In e-commerce, this window is usually 7–30 days, and in B2B even 30–90 days, because decisions are made with a delay. When the window is too short, upper-funnel supporting campaigns appear unprofitable, although in practice they only “close” the sale later. At the same time, separate conversions into primary and secondary so that the campaign does not optimise for micro-events instead of a purchase or form submission.

Carry out tracking diagnostics regularly, using Tag Assistant, GTM debugging and comparing the data with the store system (e.g. Shopify, WooCommerce, Magento). If discrepancies in the number of transactions exceed, for example, 5–10%, fix tracking first (e.g. cross-domain, payment gateways, purchase duplication), because otherwise campaign optimisation becomes a waste of budget. For analytical consistency, stick to the UTM standard (source=google, medium=cpc, campaign=search_nonbrand_pl, content=adgroup_nazwa, term={keyword}), so you can compare results in GA4 and reports without guesswork. Looker Studio is useful for reporting with business metrics (cost, revenue, ROAS, CPA, conversion rate, impression share) and a split into brand and non-brand.

Optimisation and automation: how do you scale campaigns without losing control?

You can scale campaigns without losing control when you base decisions on data, implement changes in stages and make sure automations do not disrupt the learning phase. At the start, when there is little data, a sensible choice may be to gather the first conversions with Manual CPC with eCPC or Maximise Clicks with a CPC limit, and only later switch to tCPA or tROAS. Set tCPA/tROAS targets realistically on the basis of the last 14–30 days, because an overly aggressive target can suffocate spend and reduce impression share. The most common mistake in scaling is tightening the target so much that the campaign stops spending its budget and loses volume.

Optimisation should stem from the search terms report, because that is where you can see what really converts and what should be cut with negatives. Review search terms every week, add exclusions and separate out new ad groups for high-performing intent, instead of putting them all into one basket with a general category. At the same time, work on Quality Score and Ad Rank, because better ad relevance and landing page relevance often make it possible to lower CPC without increasing bids. In practice, a consistent keyword in the headline, a matched landing page and a high CTR in the ad group translate into a lower cost per acquisition.

You retain control over performance when you drill down to the segment level (device, location, time, audience list, match type), rather than relying solely on averages for the whole campaign. This lets you quickly spot simple levers, such as locations or times of day with a clearly better CPA, which are worth scaling up. Leave automated rules mainly for “hygiene” (pauses and alerts), and avoid automation such as aggressively raising budgets, because big jumps can destabilise performance. Increase the budget gradually (usually by 10–20% every few days), keeping an eye on CPA/ROAS and impression share instead of making abrupt changes.

Scaling will be more predictable if you have a cyclical optimisation plan that organises the work and shortens response time to problems:

• daily: spend and error checks,
• weekly: queries, negatives and creatives,
• monthly: account structure and bidding targets,
• quarterly: tracking, feed (Merchant Center), lead quality in CRM and landing page.

Such a rhythm reduces situations in which you “optimise” the campaign while the real leak is outside the interface (for example in measurement, the feed or lead quality). At the same time, it makes pacing easier, because you can see more quickly whether budget growth is going hand in hand with maintaining profitability. If results fluctuate, the plan helps you get back to the source of the problem instead of multiplying random tweaks.

Combining Search, Shopping and Performance Max campaigns: avoiding cannibalisation

You will avoid cannibalisation if you separate brand and non-brand and clearly define the role of Search, Shopping and Performance Max in generating demand. Run brand in Search separately from non-brand, so the brand campaign does not “eat” the budget of activities aimed at new users. Treat Performance Max with feed as strong sales support, but keep it under control through brand exclusions (if needed) and audience signals. If PMax reports an excellent ROAS while simultaneously taking brand traffic, check incrementality, because it may simply be capturing purchases that would have happened anyway.

Consistent use of the formats means that Search captures intent from queries, while Shopping/PMax use the feed to match products and placements, instead of competing for the same easy conversions. When you see that channels are starting to overlap, refine the split: separate budgets, separate goals and clear reporting segments. In practice, controlling the pace of scaling also helps, because sudden budget increases in one campaign type can disrupt traffic distribution and make results harder to compare. The best approach is to scale in parallel: gradually increase the budget and at the same time broaden reach through new keywords or products, rather than “pumping” only one channel.

You will maintain the profitability of the campaign mix if you regularly segment results and base decisions on what is actually delivering CPA/ROAS in a given traffic type. Compare brand vs non-brand and the impact of individual campaigns on cost and revenue, instead of measuring the account with one averaged value. When you notice that Performance Max is “redirecting” budget towards the easiest conversions, check whether this is not happening at the expense of acquiring new users in non-brand Search. Such oversight allows you to keep your finger on the pulse and maintain a balance between short-term efficiency and a stable flow of new demand.