GA4 for small businesses: what it is and how to set up basic events

Open your GA4 account in another tab. Find the answer to one question: how many enquiries or orders did the website bring in last week? If after half a minute you can see only “Users” and a visits chart, the answer may not be to take an analytics course. The site may simply not be sending GA4 the actions the business depends on.

Google Analytics 4, or GA4, collects data about where people came from and what they did on a website. A useful small-business setup starts with three questions: what counts as a result, when should it be recorded, and how will you check that a GA4 record matches a real event? Installing a tag does not answer those questions by itself.

We will follow the path through a hypothetical online sports nutrition store. It has a catalogue, a cart, Nova Poshta delivery and Viber and Telegram buttons. This is an educational example, not a GrandMa client case. It shows where a click ends and an order begins.

The main unit GA4 measures is an event. Opening a page can create page_view, viewing a product can create view_item, starting checkout can create begin_checkout, and a completed order can create purchase. You can add parameters to an event, such as the product name, its identifier, amount or currency. Google lists recommended names and parameters so standard store actions can be represented consistently in reports.

Sessions, users and events answer different questions. “Users” estimates the number of people, “Sessions” counts visits, and “Events” records actions. One person might open five product pages and start checkout twice. So 100 product views do not mean 100 buyers. Universal Analytics is often remembered for goals and pageviews. In GA4, you first define important actions as events, then mark some of them as key events.

Return to the open tab. In Reports, find Realtime to check current traffic. If the Life cycle collection is published, Acquisition → Traffic acquisition helps you inspect traffic sources, and Engagement → Events shows event names and counts. Google describes the Life cycle collection. An administrator can change the navigation, so a missing menu item does not prove that data collection is off. The interface names here follow Google’s English documentation; other language settings may use different labels.

For now, ignore the temptation to explore every report. Find one event that corresponds to a real order. If there is none, the next task is in the website setup, not the chart.

In our hypothetical store, someone opens a protein product, adds it to the cart, starts checkout, chooses delivery and places an order. They may also message the store on Telegram instead of buying through the site. These are different steps with different meanings for the business.

This is a plan, not a requirement to implement all seven rows on day one. Start with purchase and a few steps that help locate where buyers drop off. If the store does not take online payments, agree what “purchase” means in your business model: an order created, an order paid for or an order shipped. Do not call the first event a sale if payment comes later. Track those statuses in the store admin or CRM as needed, and reconcile them with GA4 separately.

A unique transaction_id is especially important for purchase. Google explains how it can help prevent the same purchase from being counted twice in a web stream. A blank ID or the same ID for different orders distorts the data. An order number can serve as an ID if it contains no customer personal data.

Start with Google’s official setup sequence: an Analytics account, a GA4 property, a web stream for the site and a data collection tag. A web stream has a Measurement ID beginning with G-. Compare it with the tag on the site so data does not go to a test property or a contractor’s property. The business should control access to the property, while the specialist gets the role needed for setup.

You can add the tag through the store platform, directly in the code or using Google Tag Manager. Choose a controlled method and check for duplicates. If the platform already installs GA4, adding another tag in Tag Manager may send page_view or purchase twice. For a store, that is worse than lacking a polished dashboard: it reports sales that never happened.

Enhanced measurement in a web stream can collect page views, outbound clicks, form interactions and other basic actions without changes to the page code. It is a useful start, but its form_submit does not automatically mean a manager received an enquiry. An outbound click records a link visit, not a message sent in a messenger. Check each needed scenario on the actual site. A button that works through a script may need its own trigger.

GA4 does not automatically collect ecommerce events such as add_to_cart and purchase for an arbitrary store. The platform or a configured integration has to send them. Google Tag Manager can hold tags and triggers, but it cannot guess when the store confirmed an order or create valid items and transaction_id values without data from the site. Give the developer or integrator a precise task: which event happens, when it happens, which parameters it needs and how to prevent duplicate sends.

A short implementation brief could say: “After an order is successfully confirmed, send one purchase to our GA4 property. Pass a unique order ID, the amount in hryvnias, currency UAH and the purchased items. Do not send the event after a payment error or when the confirmation page is reopened. Show the test in DebugView and the same order in the store admin.” This is not ready-made code for every platform. It is an acceptance checklist the owner can use, even without opening Tag Manager.

For messenger buttons, write a separate requirement: what exactly you count, what to call it, and whether one action could trigger both an automatic click and a custom event. Otherwise, one attempt to message the store may appear as two “contacts.” After testing, keep the event plan with the analytics access details so the next contractor does not have to reconstruct the logic by guessing.

A key event in GA4 is an event that matters to the success of the business. In our store example, purchase is a candidate. Product views and cart additions help diagnose the path, but should not automatically be counted as results. For a lead-based business, generate_lead may be a key event when an enquiry has actually been created and entered into the working process. qualify_lead means the team has confirmed a next status using its own criteria. GA4 cannot decide whether a customer is a good fit.

Do not mark a Viber click as a key event just because people often buy there. Some will open a chat and send nothing. If you need to count conversations, you need separate confirmation from the messenger or CRM and a reliable way to connect it to the source visit. Until then, label the metric “messenger clicks” and do not substitute it for enquiries.

The route for setting key events can change as the GA4 interface changes. The rule that matters is: check the event trigger first, then mark the event as key. A key-event label makes a faulty event more visible in a report; it does not fix it.

Place a separate test order. Beforehand, open Google Tag Assistant or Tag Manager’s preview mode. They can enable debug mode for your device. Then in GA4, open Admin → Data display → DebugView. Google documents this route. Events should appear in the order you completed them: product page, cart, checkout, purchase.

Select purchase in DebugView. Compare transaction_id, currency, amount and product with the store admin record. Return to the confirmation page or refresh it: reopening the page should not create another purchase. Separately test a failed checkout and a cancellation before completion. Neither should send purchase. Realtime can confirm incoming data, but DebugView is more useful for inspecting parameters during a test.

Then check what happened outside the browser. Is the order in the store system? Did its confirmation arrive? Was it created only once? We have described a case where a form said “thank you” but the enquiry never reached the team. A browser event does not prove delivery to a manager.

If DebugView is empty, check four things: the right property and G- ID, debug mode for your device, whether the tag fired, and whether privacy or consent settings allow it. Google notes that under some client-side restrictions or without consent for analytics cookies, events may not appear in DebugView. Do not solve this by collecting data against a visitor’s choice.

Do not send a person’s name, email, phone number, message text or those details in a URL as ordinary event data. Google Analytics policies prohibit sending personal data that identifies someone. For checking the setup, you need the action name, a technical order ID without personal data and valid non-personal parameters.

If your site serves visitors in the European Economic Area, check the cookie and personal data rules that apply to your implementation. Google requires relevant consent signals for certain measurement and ad personalization use cases in the EEA. Consent Mode passes a person’s choice to Google tags. It is a technical mechanism, not a replacement for a proper banner and policy. Google explains how the banner and Consent Mode work together. The European Commission also offers visitors a choice about analytics cookies on its own sites.

When someone declines consent, uses an ad blocker or has a browser that limits tracking, your numbers may be incomplete. Do not promise a reconciliation “down to the last user.” Check that the system respects the choice and the report explains its data limits. For a business decision, compare GA4 with store orders and CRM records because each system answers different questions.

Return to the GA4 tab. Choose the previous full week and look at three questions:

Compare the same periods and time zones. After changing a tag, record the date so you do not compare a week with a new event setup against an old week as though the conditions were identical. If orders are rare, one extra order makes a percentage chart look dramatic. Before changing the budget, look at the records themselves and the quality of the enquiries.

In the Ratner Beauty case study, we used archived GA4 data to show a channel share: 62.7% of all sessions from 1 July 2025 to 30 June 2026 came from Organic Search. That is a verified traffic-source metric for a specific period. The case study does not publish sales or revenue, so I do not infer purchases from that share. Your own “sessions from search” do not replace confirmed orders in a report either.

If you need a plan for what to measure before configuring tags, read our separate article on a GA4 measurement plan. For B2B ads, we explain Google Ads conversion quality separately. If events are already coming in but the team cannot use them to make a decision, GrandMa’s analytics service starts by connecting events, enquiries and business questions.