This question arrives in almost the same shape every time: where do I start with data analysis for marketing, and what do I learn first, Excel or Google Analytics or Python?
The answer nobody likes: none of those is the first step. The first step is knowing the question you are trying to answer. People who learn the tool first end up able to produce a report. People who learn the question first end up able to decide something.
The difference shows up immediately at work. Someone sends you a report with twenty numbers in it and not one sentence saying what to do tomorrow. That is not a gap in the tool. It is a gap in the fact that nobody asked a question first.
Start with a question that needs a decision
Write down one question in this shape: if the answer is this, I will do that, and if the answer is the other thing, I will do something else. If you cannot complete the second half, it is not an analysis question, it is curiosity.
A real example from a small business: are short videos bringing customers, or just views? If customers, I spend more on them. If views, I stop and move the time somewhere else. That is an analysis question, because there is a decision on both sides of it.
This is also what makes learning fast. You only learn the part of the tool you need to answer the question, instead of all of it. A month spent that way is worth more than a year of courses.
Learn the definitions before the tool, and this is not theory
The biggest mistake in marketing data work is not arithmetic. It is somebody reading a metric they have misunderstood. These metrics have published official definitions, and reading one takes five minutes and saves months of wrong conclusions.
Take an example a lot of people get wrong: engagement rate and bounce rate in Google Analytics 4. Google's own documentation says a session counts as engaged if it meets any one of three conditions: it lasts longer than 10 seconds, or it has a key event, or it has 2 or more screen or page views. Bounce rate is simply the percentage of sessions that were not engaged.
Now think about what that number is. A site where people read a long article on one page for two minutes with no key event: engaged, because it passed 10 seconds. A site where somebody opened two pages in five seconds and left: also engaged. The metric measures far less than its name suggests.
The same page also notes that events such as first_visit, first_open and session_start are excluded from the engaged sessions calculation, even if you have marked them as key events. That detail changes a real report, and only people who read the documentation know it.
So the rule: before you conclude anything from a metric, read its definition at the tool's own source. Not from an article, and not from a course, because articles and courses go stale and definitions change.
Learn to read before you learn to build
Plenty of people start their analytics learning by building a dashboard. That is the wrong order. The ready-made reports already exist, and Google's documentation on reports says there are two kinds: overview reports, which summarise data about a topic, and detail reports, which provide more in-depth information about an area of interest. It lists 35 pre-made reports, among them traffic acquisition, user acquisition, pages and screens, purchase journey and checkout journey.
Take three of those in your first month: traffic acquisition, pages and screens, and events. Between them they answer most of a small business's questions: where people came from, what they read, and what they did. A custom dashboard waits until you know what the ready-made ones are missing.
There is also a real difference between the user acquisition report and the traffic acquisition report, and the documentation keeps a separate comparison page for exactly that. If you conflate them, you are comparing the source that first brought a user in with the source that brought this particular session, which is a classic reason a monthly report says something untrue.
The tools to learn, in order
This order is based on what returns value fastest, not on what looks advanced.
First: spreadsheets. Excel or Google Sheets, and only four things in them: sorting, filtering, pivot tables, and VLOOKUP or XLOOKUP. Those do about 80 per cent of day-to-day marketing analysis work, and most people with two years of experience still have not got pivot tables properly.
Second: the analytics platform on your own site, and reading it rather than building in it. Third: the ads interface where your money actually goes, because those are the numbers you will be held to. Fourth, and optional: SQL, if your work genuinely involves a database. Python comes last, and if you are in small business marketing you may not need it at all.
Do not learn a tool before you have a question that needs it. A tool learned without a question gets forgotten. A tool learned while answering one sticks.
Four numbers to start with in a small business
Do not try to track everything. Start with four and understand them properly:
Weekly enquiries, which is the number most small businesses actually judge themselves by. The share of enquiries that became customers. Cost per enquiry, if you are spending on ads. And the channel each enquiry came from, which is the one that lets you move budget.
Those four fit in a simple table with no advanced tooling at all. And if you are not collecting them today, learning Python will not fix your problem, because the problem is not analysis. The problem is that there is no data to analyse.
How to practise without a big account
Practice does not need an account with a million visits. It needs a question and real data, even a small amount of it.
Take your own project's page, or a friend's, pull the last thirty days of numbers, write down one question, and try to answer it in a single page. That page has to end with a sentence saying what to do next week. That is the exercise, and you repeat it four times.
Then compare your prediction with what happened. What actually teaches data analysis is having written the prediction down before seeing the result, because only then do you find out whether you were understanding or rationalising. That is the whole skill.
The order that comes before all of this is in how to learn digital marketing from scratch, and if you are not sure what to begin with at all, what to learn first in marketing is the better starting point. If your question is about time, the month-by-month table is in how long it takes to learn digital marketing.
The part numbers do not solve
This has to be said plainly: data analysis does not tell you who your market is. That gets decided by people and questions, not by a dashboard, and I went through it in how to identify your target market.
Analysis also does not fix a weak offer. If the offer itself is unconvincing, all the numbers will do is tell you so with more precision. That is useful information, but it is not a solution.
If you are learning with no budget, the free route in full is in how to learn marketing without courses. Data analysis is one of the cheapest things to learn, because the official documentation is free and your own small project's data is already available to you.
One last practical note. The part people get wrong most often is reading the Google Analytics report itself, and the traps that keep recurring in it. Our friends at KF Agency wrote that up in detail in how to read a Google Analytics report correctly, which is in Arabic. If you are starting now, read it before your first report.
All of the above in one sentence: analysis is a question skill, not a tool skill. The tool takes two weeks. The question gets learned every time a number surprises you.