Clients supported

Tracking was set to "click on the confirm booking button," so it counted even when the form returned an error. I changed it to count only when an element that appears solely on the confirmation screen is shown.

The thank-you page URL used in the key event condition was different from the actual URL. I fixed the condition so inquiries are counted correctly.

LINE counts people who arrived from that URL, while Google Ads counts only people who clicked the ad. I matched records one by one by date and time to pin down the conditions behind the gap.

Revenue was being recorded at "begin checkout," and purchase completion wasn't set as a goal. I made purchase completion the only event that carries revenue.

The booking form was inside an iframe on a different domain. I added GTM to the iframe as well, and where that still didn't work, had the vendor send a dataLayer event on completion.

You can still track the thank-you page, but you lose sight of where converting visitors came from. Cross-domain tracking needs to be set up first.

I had unique class names added to each link and image, then tracked each button separately in GTM.

Firing on the click also counts people who left without seeing the next page. If the only path is the button, tracking the next page load is enough and simpler to set up.

I look at site changes made around the same time and bounce rate by device. In this case, bounce rate jumped on smartphones only, right after the home page was changed.

Break it down by traffic channel into sessions, revenue and purchase rate. If the purchase rate hasn't changed, the issue is in the traffic, not on the site.

Compare Search Console queries by landing page against the previous year and split them into "both ranking and impressions fell," "demand fell" and "only ranking fell." That tells you which pages to fix.

Page title dimensions include pages viewed later in the visit, so they don't show where people entered. Use the "Landing page" dimension.

You can see it in GA4's "AI Assistant" channel. Visits through apps and similar routes often aren't classified, so assume the real number is higher.

Common factors include a structure machines can read easily, FAQs, and credibility through clearly cited sources. Decide on 20 to 30 questions you want to appear for, ask AI tools regularly, and measure the effect by the share of times you're mentioned.

It works if you set rules, such as giving it last month's delivered report as a model and having it label guesses as "hypotheses." A person still checks the result against the tables at the end.

You avoid API limits, thresholds and sampling, and you can reproduce segments and past conversions in SQL. I make clear up front that you'll need someone who writes SQL and that usage-based charges apply.

E-commerce tracking and the BigQuery link are prerequisites. To look at individual members, hash the login ID and send it as a user ID, then shape the data in SQL before connecting it to a dashboard.

People who first visited within the period and came back are counted in both. GA4 also defines "new" differently for the dimension and the metric, so decide which one you'll use first.