Ads learn from whatever they were told is a lead. If your account marks the wrong thing as a lead, the ads diligently bring the wrong people: those who read your site for a long time, not those who leave a contact. A business owner does not need to know how to set up the account, but does need to know what is recorded there as “leads”. It takes ten minutes to check, and below is where to go and what to look for.
A note on accuracy. The menu names come from the Google Ads help page in English; in another language or interface version they may differ slightly. The word “conversion” here simply means “an action on the site that the account counts as a result of the ads”.
1. Main actions: there is a contact afterwards
The main ones are a submitted form, a phone call, an order and a purchase. After each of them you are left with a person you can call back or ship goods to. These are the leads.
What to check yourself. Open Google Ads and go to the “Goals” menu → “Conversions” → “Summary”. That is the list of every action the account counts. Next to the main actions the type should say “Primary”. According to Google’s help, primary actions are the ones used for bidding (meaning the ads learn from them) and they show up in the “Conversions” column.
To change the type, find the goal that contains the action, click “Edit goal” and open the “Conversion action optimization” section. There are two options: “Primary” and “Secondary (observe only)”.
2. Supporting actions: interest, not a lead
A click on the phone number, the “Contacts” page, an item in the cart, two minutes on the site, a click on a messenger. The person is interested but has not left you anything yet. The type that fits these actions is “Secondary (observe only)”: per Google’s help, they appear in the “All conversions” column and are not used for bidding. The account sees them, but the ads do not learn from them.
A separate word on clicks on a phone number and a messenger. Our position: any click action is risky for two reasons. The same click can be counted twice, and a person may click and never get through. So we build on what is harder to distort by accident: submitted forms, call tracking (a separate phone number for ads that counts calls; for example Binotel or Ringostat) and offline conversions (data about real sales from your CRM, passed to the account).
There is also an honest nuance, and practitioners differ here. Some say that without call tracking a mobile click on the number can stay primary: otherwise the ads see no calls at all, and Google does not measure calls from ads in Ukraine itself. Others advise removing it so the ads do not learn from a stand-in instead of the real action. Our line: if the number click stays primary, once a month compare its count with how many calls you really received. Besides, practitioner chats describe a case where WhatsApp clicks were recorded in the account but the business found no actual enquiries in the messenger. That is why this click is also easier to keep as a secondary action.
A disagreement we do not hide: if purchases are few, can you temporarily train the ads on a supporting action, such as the cart? Some practitioners say yes: with five purchases a week the ads will not gather enough data, so at the start they learn from the cart and are then switched to the purchase. Others object: the supporting action pulls the ads toward people who add to the cart but do not buy, and one wrong setting breaks the automation. Our position: by default no, supporting actions stay for observation. An exception is possible at the start, when volume is small, but with a written plan for when exactly we switch back to the purchase.
3. Ads look for what they were told to look for
If “time on site” is set as primary, Google will look for people who stay on the site a long time, not for people who buy. It follows the task literally.
An example from the market. A furniture store from one public case study (published 23 May 2024; December 2023 – February 2024 was compared with April–November 2023). There, “site activity over 120 seconds” was removed from the primary actions, and the value of a click on the phone number was lowered from 60 UAH to 1 UAH, so that a click counts for little next to a purchase. The result: ROAS (how many hryvnias of sales per hryvnia of ads) rose from 541% to 1148%, roughly from 5 UAH to 11 UAH of sales per hryvnia spent, and purchases went from 9 to 16 per month.
A caveat. At the same time they changed the strategy, doubled the budget and added an asset group. So the effect of the change in goals alone cannot be separated out. This example shows a direction; it does not promise your result.
4. One lead is counted once
The same purchase can be counted twice: directly by the Google Ads tag and through an import from Google Analytics. Then the report shows twice as many leads as there were, and the ads learn from doubled data.
Per Google’s help, conversions created from Google Analytics are secondary by default. So doubling appears when the imported action has been manually made primary while the same action from the Google Ads tag is primary too. Practitioners advise keeping one of the two actions primary and moving the other to “Secondary”.
What to check yourself. In the same list, “Goals” → “Conversions” → “Summary”, look at the names of the primary actions. There should not be two about the same thing, for example “Order” and “purchase”. If there are, one of them should be moved to “Secondary”.
5. Check the report against real life
Take last month. How many leads does the ads report show? How many forms, calls and orders did you actually get: in your email, CRM, the receptionist’s notebook?
A small difference is normal. A person may not have got through, a lead may have arrived twice, someone sent a test form. But if the difference is several times over (the report says 120 leads, while the phone rang twice), something extra is being counted: a click instead of a call, a duplicate, a supporting action among the primary ones. We do not name an acceptable percentage of discrepancy: there is no primary source for such a limit, and every business has its own.
6. For stores: a purchase must carry an amount
If you run a store, the ads should receive the order amount together with the purchase. Without it a phone case for 200 UAH and a sofa for 20,000 UAH look the same to them, and they do not know which buyers to learn from.
What to check yourself. Find the revenue from ads in your advertising report (in our client reports it is the “Revenue from ads” line). It should look like your real sales over the same period. Supporting actions get no monetary value at all, or a symbolic one, as in the example above (1 UAH); otherwise the “cart” will weigh as much as a sale.
10-minute checklist
- In “Goals” → “Conversions” → “Summary”, forms, calls, orders and purchases are marked “Primary”.
- Clicks, the cart, the “Contacts” page and time on site are marked “Secondary (observe only)”.
- Among the primary actions there are no two about the same thing, including one created by a tag and one imported from Analytics.
- The number of leads in last month’s report is compared with real forms, calls and orders: the numbers are close, not several times apart.
- For a store: “Revenue from ads” looks like your sales.
Summary
Ads count as a lead what they were told is primary, and they learn from exactly that. Primary should be actions after which there is a contact: a form, a call, an order, a purchase. Interest without a contact (a click, the cart, time on site) is better left for observation. Count one lead once, check the report against real life and, if you run a store, make sure the purchase carries an amount. What else should be connected to your site we covered in the article on what should be connected to your website, and how to read the report figures themselves in the article the report glossary. To see how we set up search ads, go to the search advertising section.
Want a free audit of your Google Ads account? Send us a request: we will check what exactly is counted as a lead, where the money goes and whether the numbers in your report are true.