Why your analytics numbers don't match, and which one to believe

Why do two tools show different visitor numbers?

A "visitor" isn't a fact the browser reports. Each tool decides what counts as the same person, and those decisions differ. Here's what each product says about its own counting.

Shopify's update is a good warning. On its rollout days, your sessions and conversion rate could shift with no change in your traffic, and Shopify says so itself. Some sessions with no pageview now count too, and bots it identifies are filtered out. If a chart jumped in late September 2026, read that page first.

How much can a definition alone change the number?

More than most people expect. Take our demo site over 30 days, where the website recorded 3,943 visitors, 4,930 sessions and 8,985 pageviews. Those are three honest traffic numbers, 2.3 times apart, from the same visits.

Now the visitor number itself. Those 4,930 sessions came from 3,943 visitors, so 987 were return visits. A tool that gives everyone a new ID each day would count any return on a later day as a new visitor. If all 987 came on later days, that tool would report up to 4,930 visitors for the same traffic. That's 25% higher, with nothing wrong in either tool.

So compare pageviews first. They have the least room for interpretation. Compare visitors last, and only over the same dates.

What else makes the numbers differ?

• Consent banners. If GA waits for consent, it misses people who decline. Google's behavioral modeling can estimate some of them where it's turned on. A tool without a banner counts them, so here the gap is about who is seen.
• Blockers. Browser extensions and some networks block analytics scripts, and they don't block them all equally. No tool sees what never reached it, whatever it claims about accuracy.
• Bots. Each tool has its own filter. MetricMind drops known bot user agents and data centre traffic. Fathom and Plausible describe their own filters. No two of these lists are the same.
• Hidden rows. GA4's data thresholds can withhold rows in a report. A breakdown may not add up to its total.
• Time zones. A day in one tool can start hours before a day in another. Set both to the same zone before comparing days.
• Unfinished days. Today so far isn't comparable to anything. Use finished days.

Why does each tool credit sales to a different source?

Because they follow different rules for who gets the credit. Three common ones:

• GA4 sessions use non-direct last click. Google's own example: someone arrives from organic search on day 1 and returns directly on day 68, and that second session still counts as google / organic.
• GA4 key events can use data-driven attribution, which shares credit across touchpoints with a model. A sale can be split, so your source totals may not look like counts of people.
• MetricMind gives each visit the source of its entry page and doesn't look back. That day-68 visit stays Direct. You'll see more Direct here than in GA4, on the same traffic.

Ad platforms add their own rules. Google's older article comparing Ads and Analytics, written for the previous version of Analytics, explains that Google Ads can report a conversion on the date of the ad click, while Analytics reports it when it happened. It also says Ads only counts conversions after a click on its own ads. So expect the ad platform to report a different day and a different total, with neither one wrong.

How do you test whether the gap is a real problem?

This takes about 30 minutes and works with any two tools. Do it with both installed on the same pages.

1. Send one marked visit. On your phone, on mobile data, in a private tab, open your home page with ?utm_source=mismatch-test-1 added to the address. Click through to two more pages. Ten minutes later, filter each tool by that source. Both should show one visit and three pageviews. If one shows nothing, its tag is missing, blocked or excluding you, and you've found the problem.
2. Repeat on your main pages. Try a product page, a checkout or contact page, and any page built by a separate system. Tags often go missing there and nowhere else.
3. Work out the daily ratio. Export pageviews per day from both, for 14 finished days. Divide one by the other for each day.
4. Compare the order. List the top 10 pages and top 5 sources in each. The same pages should lead. Direct and search may swap, for the reason above.

Which number should you believe?

No tool has the true visitor count, ours included. Choose one source per question and stick with it.

Trends inside one tool are worth far more than a match between two. If you're moving away from Google Analytics, keep the ratio you found. It's your bridge between old numbers and new ones. And if your numbers fell in both tools at once, the problem is real, so go to the 15-minute routine for a traffic drop.

How MetricMind shortens this

It can't make the numbers match, and it won't try. It doesn't model missing visitors, so if someone blocks the script, you see fewer visitors than really came. What it does:

• Answers show their work. Ask the analyst about a number and you get the query, plus the assumptions it made: which days, which traffic was left out, what counted. Every plan can ask, with a monthly allowance.
• A running period is compared fairly. Up to the same clock time, so this morning isn't judged against all of yesterday.
• Your own rules stay in one place. Excluded addresses, excluded paths and team rules such as what counts as a lead apply to the answers you read.

For the full list of differences with Google's tool, see MetricMind vs Google Analytics 4.