In advertising and data analysis, the first step is understanding the terms. This page brings together the most common ones. For each term, as well as the definition and formula, we add a short “In practice” note, because numbers do not speak for themselves: interpretation is what matters.
Working with CloudAD: common questions
The 8 questions we are asked most often before working with a client.
What services does CloudAD offer?
End-to-end marketing services built around data: GA4/GA360 implementation and health checks, data consulting, Data Studio dashboards, digital advertising management and optimisation, Google Business Profile integration (Cloud My Business), corporate training, and consulting on data and AI applications.
Is there a charge for the first consultation?
No. In the first consultation our consultants will do their best to answer your questions and offer advice. You only pay if you later decide to use our consulting services.
What should I prepare before a consultation?
We suggest preparing: the analytics and advertising tools you currently use (GA4, GTM, ad accounts), the problem you want to solve or the results you expect, and your website URL. The more specific the information, the more precise our advice can be.
What size of business is CloudAD suited to?
We have worked with businesses ranging from SMEs to listed groups, across media, e-commerce, finance, transport and technology. The scope of our services is adjusted to your size and needs: a corporate website with simple traffic should not pay for complexity it will not use.
We already have an in-house marketing or data team. Do we still need CloudAD?
Many of our clients have in-house teams. Our role is usually to add technical depth (GA4 event design, BigQuery, the data layer) or an independent third-party view (health checks, acceptance testing). We can also handle only planning and training, with your team taking over execution.
What role does CloudAD play in a project?
We start from the decisions the business needs to make and support the team through data transformation. First we make sure tracking and data sources are correct, then we organise GA4, advertising and operational data into evidence that supports judgement, so that data is used in everyday decisions. CloudAD is a Google Marketing Platform Sales Partner, and the only Taiwan-headquartered company among the GA360 sales partners in Taiwan.
How does a project usually work?
It usually starts with a free consultation or a Data Health Check to confirm the problem and scope, followed by a matching service proposal. Delivery has clear acceptance criteria, and implementation projects also include documentation and training.
How do you handle the security and confidentiality of client data?
We have experience serving highly regulated industries such as finance and insurance, and are familiar with de-identification and personal data protection practices. Every engagement is covered by a non-disclosure agreement, and data access is managed on the principle of least privilege.
Ad performance metrics
The 10 numbers that appear most often in campaign management reports, and what they really mean.
CPC Cost per click
Ad spend divided by the number of clicks: the average cost of each click. Formula: CPC = total spend ÷ clicks.
In practice | A low CPC does not mean cheap. If the wrong people are clicking, even a very low CPC is wasted money. Always read CPC together with back-end conversion data.
CPM Cost per mille (per 1,000 impressions)
The cost of every 1,000 impressions. Formula: CPM = total spend ÷ impressions × 1000.
In practice | CPM reflects how competitive the audience is and the quality of the creative. It is naturally higher in peak seasons or for narrowly targeted audiences. Compare it with your own historical benchmark; the absolute value on its own is easy to misread.
CTR Click-through rate
The share of people who saw the ad and clicked on it. Formula: CTR = clicks ÷ impressions × 100%.
In practice | A high CTR means the creative grabs attention, not that it is good quality: clickbait creative gets a very high CTR and very poor conversions. Compare it with the conversion rate to know whether you are attracting the right people or the wrong ones.
CPA Cost per action
The average cost of one conversion (a purchase, a sign-up or a lead form submission). Formula: CPA = total spend ÷ conversions.
In practice | A rising CPA is not necessarily a campaign problem. It may be caused by a landing page redesign, stock levels or seasonality. First confirm that conversion tracking is not broken, then go back and review the ads.
CPL Cost per lead
The average cost of acquiring one lead, commonly used for B2B and high-value services.
In practice | Read CPL together with lead quality. If cheap leads rarely convert into sales, the real cost of acquiring a customer is actually higher. We recommend tracking down to the level of cost per qualified lead.
ROAS Return on ad spend
How much revenue each dollar of ad spend brings back. Formula: ROAS = revenue from ads ÷ ad spend.
In practice | What counts as a good ROAS depends on your margin structure: for a product with a 30% gross margin, a ROAS of 3 may still be losing money. Different attribution methods also produce very different numbers, so align attribution settings before comparing.
ROI Return on investment
How much profit the total investment (including ad spend, staff and production costs) brings back. Formula: ROI = (revenue − total cost) ÷ total cost × 100%.
In practice | ROAS counts only ad spend; ROI counts all costs. Use ROI when reporting to management and ROAS when reviewing with media buyers. Mixing up the two is a common cause of arguments in meetings.
CVR Conversion rate
The share of people who visited or clicked and then completed the target action. Formula: CVR = conversions ÷ clicks (or sessions) × 100%.
In practice | When CVR drops, break it down by traffic source first. Often the website has not got worse; the new traffic simply has different intent. Only by looking at sources separately can you find the real cause.
Impressions and reach Impression / Reach
Impressions are the total number of times an ad is shown; reach is the number of unique people who saw it. One person seeing an ad three times = 3 impressions, 1 reach.
In practice | Impressions divided by reach = frequency. A frequency that is too high means the same people are being bombarded repeatedly. It usually comes with a falling CTR and is an early sign of audience fatigue.
Remarketing Retargeting
Showing ads again to audiences who have previously interacted with the brand (visited the site, added to cart, watched a video).
In practice | Remarketing usually shows the best-looking results, but it harvests existing intent. Putting the entire budget into remarketing means not acquiring new customers, and the audience pool gets thinner over time.
AI search terms
The way people search is changing. These 8 terms determine whether your brand appears in AI answers.
SEO Search engine optimisation
A set of practices that help a website rank better in the organic results of search engines such as Google.
In practice | SEO is not dead, but traffic is splitting across entry points: more and more questions are answered directly by AI. The focus is shifting from “what position do we rank” to “will we be cited”.
AEO Answer engine optimisation
Optimising content to become the “direct answer”, including featured snippets, voice search and AI Overviews. The key is to write questions and answers so that machines can use them directly.
In practice | The core AEO tasks are FAQ structure, definition paragraphs and structured data. A page that cleanly answers one question has a chance of being chosen as the answer.
GEO Generative engine optimisation
Optimising content so that generative AI such as ChatGPT, Perplexity and Gemini cites or mentions it in answers. The concept was formally proposed in 2023 by a research team at Princeton University.
In practice | Success in GEO is measured by citations, not rankings. Numbers, sources, clear definitions and named first-hand experience are the four kinds of material AI is most likely to cite.
LLMO / AIO Large language model optimisation / AI optimisation
LLMO refers to optimising for citation by large language models; AIO is the broadest umbrella term, covering search, generative and recommendation AI. The practices behind these four terms (AEO, GEO, LLMO, AIO) overlap by around 70 to 80%, and the industry has no agreed definitions yet.
In practice | Rather than worrying about which term to use, put the effort into making content “worth citing by AI”: verifiable, sourced and clearly structured. That works whichever term you use.
AI Overviews
AI-generated summary answers at the top of Google search results, which cite several web pages as sources.
In practice | When an AI Overview appears, clicks on traditional rankings are diluted. What you can do is aim to become one of the sources it cites, which takes the same approach as AEO.
llms.txt
A site guide file placed in the website’s root directory and written for AI crawlers, similar in purpose to a sitemap for search engines.
In practice | An emerging convention that costs very little to put in place; the CloudAD website already has one. It does not guarantee citations, but it makes it easier for AI to understand your site.
Structured data Schema markup
Marking up web page content in a machine-readable format using the schema.org standard, for example FAQs, articles, organisations and products.
In practice | Structured data is the shared foundation of SEO and AEO. Marking up FAQs, HowTos and case study results first gives the highest return.
E-E-A-T
Google’s framework for assessing content quality: Experience, Expertise, Authoritativeness and Trustworthiness.
In practice | E-E-A-T matters even more in the AI era, because engines have to decide “whom to cite”. Named authors, real case study figures and clear company information are all signals machines use to judge credibility.
Google data tools
From GA4 to BigQuery: the full chain that takes data from collection to use.
GA and Universal Analytics Google Analytics / Universal Analytics
Google’s family of web analytics tools. The older Universal Analytics was fully shut down in July 2024, and its data can no longer be accessed.
In practice | Making decisions from report screenshots from the Universal Analytics era is like driving with a map that is no longer published. Historical data that was not backed up cannot be recovered now.
GA4 Google Analytics 4
The current version of Google Analytics. Its tracking model is built around events, it works across websites and apps, and it can be linked to BigQuery free of charge.
In practice | Installing GA4 does not automatically give you the right data; you only get that if it is set up correctly. Event planning, conversion definitions and the data retention period are the three things most often overlooked.
Further reading | 7 settings to configure when you open a GA4 account
GA360 Google Analytics 360
The paid enterprise version of GA4, offering higher data limits, an SLA service guarantee, unsampled reports and deeper integration with Google Marketing Platform (GMP).
In practice | It is only worth evaluating when traffic is large enough for reports to start sampling, or when you have compliance or SLA requirements. CloudAD is the only Taiwan-headquartered company among the GA360 sales partners in Taiwan.
GTM Google Tag Manager
A tool for managing website tracking tags in one place, so marketing teams can deploy GA4, ad pixels and other tracking without changing the website’s code.
In practice | GTM makes deployment faster, but it also gets you faster to the point where nobody knows who installed what. Containers need naming conventions and version records, or a year later you will be doing archaeology.
Further reading | Installing GA4 and the Facebook pixel quickly with GTM
Data layer dataLayer
The standard structure for passing data between a web page and GTM, organising page information (products, amounts, membership status) into a machine-readable format.
In practice | The quality of the data layer design determines the quality of all your tracking. An undocumented data layer is a minefield: every site redesign risks breaking it.
BigQuery
Google Cloud’s data warehouse service. GA4 can export raw event data here free of charge for advanced analysis with SQL or AI tools.
In practice | BigQuery holds the complete version of your GA4 data: unsampled and not subject to the interface’s data retention limits. If you want AI to analyse your traffic data, this is where to start.
Data Studio formerly Looker Studio
Google’s free data visualisation tool, which turns data from multiple sources such as GA4, advertising platforms and spreadsheets into interactive dashboards. Its name has changed more than once: it was renamed from Data Studio to Looker Studio in 2022, then renamed back to Data Studio in April 2026. Existing reports and permissions are not affected.
In practice | A dashboard’s value is not in how good the charts look, but in whether the people reading it know what to do next. With the wrong metrics, even the most beautiful report is just wallpaper.
UTM parameters
Tagging parameters added to the end of a URL (utm_source, utm_medium, utm_campaign and so on) so that GA4 can identify traffic sources.
In practice | Without consistent UTM naming rules, the same campaign ends up under five different names in your reports. We recommend creating them consistently with a builder.
Attribution Attribution models
The rules for assigning credit for a conversion to each touchpoint along the path, for example last click or data-driven attribution (DDA).
In practice | Change the attribution model and the results for the same campaign change too. Before comparing performance across platforms, make sure they use the same attribution, or you are comparing apples and oranges.
Consent Mode Google Consent Mode
Google’s privacy mechanism that adjusts tracking according to each user’s cookie consent status, and uses modelling to fill in conversion estimates when consent is not given.
In practice | Privacy regulation will only get stricter. Consent Mode is a buffer between “less data” and “compliance”; the earlier you deploy it, the more solid the basis for modelled estimates.
We will keep adding terms. If your team often gets stuck at “we have the numbers, but we don’t know what they mean” when reading reports, that is usually not a terminology problem but a problem with the framework for interpreting data. If that sounds familiar, talk to CloudAD.
Common questions and solutions
The practical questions we are asked most often about data tracking, attribution and implementation.
Conversions in our GA4/GA360 reports do not match the orders in our website back end, and sometimes they suddenly drop sharply. What are the common causes, and how do we fix it?
For e-commerce businesses in Taiwan, the most common causes include events firing more than once (duplicate purchase events), referral information lost across domains or on payment pages, UTM tags that are not standardised, and internal, payment and logistics traffic that has not been excluded. CloudAD’s Data Health Check goes through the GA tracking code, the event and conversion design and the referral exclusion list item by item. We first align definitions at the data source, then confirm which numbers in the reports and the back end can be compared and whether the differences are reasonable.
I already compile ad platform and website data by hand in spreadsheets, but the numbers never match and it takes a lot of time. Is there a more reliable approach?
Data merged by hand often fails to match because of inconsistent definitions and different time windows. A more reliable approach is to first set up unified metric definitions (a metrics dictionary), then connect data sources such as GA4, Google Ads and Meta to Data Studio or BigQuery to build dashboards that update automatically. CloudAD provides the complete build, from data connection and metric definitions to automated reporting.
We are a financial services company bound by personal data and compliance rules, so we cannot freely add tracking code or send customer data to external tools. How can we do compliant performance tracking and attribution?
We recommend relying mainly on first-party data, combined with consent management, server-side tracking and minimal data transfer, so that identifiable data stays in your own environment. CloudAD helps finance and insurance clients build compliant first-party data tracking architecture and governance processes, so that traceable performance tracking and attribution remain possible while meeting personal data and compliance requirements.
With third-party cookies being phased out, how can we keep improving ROAS steadily and clearly show each channel’s contribution to revenue?
The key is to build first-party data and micro-conversion signals that feed back into the ad algorithms’ learning, then use cross-channel attribution to break down each channel’s contribution to revenue and compile it automatically into dashboards. CloudAD uses machine learning to calculate website assist factors and F1 scores, speeding up ad learning and optimising bids, and also provides cross-channel attribution reports.
Our GA4 events and conversions are all set up, but the numbers look odd. How can we quickly find where the tracking is going wrong?
Common problems include duplicate events, incomplete parameters, conversions counted twice and referrals that are not excluded. Using DebugView, event parameter checks and definition comparisons, CloudAD quickly pinpoints whether the problem lies in the tag implementation, the configuration or the data source, and provides a list of fixes.
What is a GA4 or website data health check, and where can I get one in Taiwan?
A data health check verifies that the data itself is accurate and reliable before any analysis. CloudAD’s website health check covers three areas: GA tracking code diagnosis, website traffic diagnosis and ad performance diagnosis. It identifies implementation and tracking problems and recommends fixes. In the health checks CloudAD has carried out, around 80% of websites had errors in their GA implementation, which is why a health check is the first step before analysis and optimisation.
How can GA4/GA360 events and conversions be brought into Data Studio as automated dashboards, with ad platform performance in the same set of reports?
CloudAD helps design GA4/GA360 events and conversions, connects GTM, BigQuery and each ad platform, and builds automated Data Studio dashboards with permissions, scheduling and data source integration, so that website and ad performance appear in the same report in real time.
I need traceable attribution and automated reporting to prove marketing ROI, and I also want to plan a GA4/GA360 architecture. What kind of consultant should I look for?
This kind of need calls for a consultant with both data engineering and hands-on marketing skills. CloudAD is a Google Marketing Platform Sales Partner, providing GA4/GA360 implementation, BigQuery integration, first-party data governance and automated Data Studio dashboards to help present marketing ROI in a traceable way.
We are a retail brand driving online traffic to physical stores (OMO). We want to publish post content across multiple stores at once and measure results with metrics such as store foot traffic. Do you offer this kind of service?
Yes. CloudAD provides solutions for physical retail channels and OMO traffic analysis. With Cloud My Business, a platform we developed in-house, information for multiple stores is brought together in one place, supporting content publishing for large numbers of stores and the measurement of store-visit metrics.
I am looking for a team that understands both data engineering and hands-on marketing, and can turn complex GA4, advertising and BI data into actionable strategy and put it into practice. Is there one in Taiwan?
That is exactly what CloudAD does. We translate complex GA4, advertising and BI data into marketing strategies that brands can act on directly, and help with implementation and ongoing operation, so that the data is accurate at source and the marketing delivers results.
Which industries and sizes of business can CloudAD help?
CloudAD serves industries including e-commerce, retail, finance, insurance and media, with implementation experience ranging from small and mid-sized brands to enterprise clients. We provide scalable GA4/GA360 data architecture to meet data governance and cross-channel attribution needs.
Where does CloudAD start when helping a business solve its problems?
At the data source: first a Data Health Check to put the data right, then automated dashboards and cross-channel attribution, and machine learning to optimise advertising. As a member of the iKala group, CloudAD can also bring together data and AI resources from the group and its partners to provide integrated services.