Shopify conversions and GA4 purchases differ
The symptom
In your experiment report, the conversions from Shopify and the purchases from GA4 differ significantly. In some cases the variation is ahead in one source and behind in the other.
Two data sources never measure exactly the same thing, so a deviation is not automatically an error in your setup. What matters is that you can identify the cause and that the relative effect between original and variation behaves similarly in both sources.
Why Shopify and GA4 show different values
If you use Shopify Conversion Data (the "Use Shopify Data" setting in the Advanced Setup of the Tracking Setup), the data for your experiment comes from two sources: visitors are imported from GA4, while conversions and revenue come directly from Shopify. The two sources differ in who they measure and when the data arrives. Varify Realtime Tracking, by contrast, measures visitors and conversions from a single source. The following possibilities cover the most common reasons for the deviation.
How to find the cause
Possibility 1: Different consent bases
Shopify records every completed order, regardless of whether the visitor agreed to tracking. GA4, on the other hand, only records visitors who gave consent. The visitors from GA4 and the conversions from Shopify therefore have different bases. As a result, the conversion rate in Shopify can look higher than it really is, and the purchases in GA4 come out lower.
To check the setting:
- In your Shopify store, open "Settings > Customer events".
- Open the Varify custom pixel (you can find the setup in the article Shopify Conversion Data).
- Under "Customer privacy", check the "Permission" setting. With "Not required", measurement happens without consent; with "Required", only with consent.
→ With Varify Realtime Tracking, you decide in the Tracking Setup yourself whether Varify measures with or without user consent. Visitors and conversions then share the same basis.
Different consent rates distort the absolute conversion rate, but not the relative uplift between original and variation. So focus on the uplift (see "Which data source should you base your decision on?").
Possibility 2: Duplicate events in GA4
Shopify counts each order exactly once. In GA4, events can arrive more than once, for example through test orders, bot traffic or events that fire multiple times.
→ Keep the "Exclude Duplicate User Events" filter enabled. This keeps such events out of the evaluation.
→ If you disable the filter, the number of purchases measured in GA4 goes up. This can distort the result significantly.
Possibility 3: Time delay of the GA4 data
Shopify conversions are available almost in real time. GA4 data typically arrives 12 to 24 hours later, and in rare cases it can take up to 48 hours. If you compare the two sources too early, GA4 is still missing data.
→ Compare the sources over complete days, and don't draw final conclusions from the last one or two days yet.
→ Varify Realtime Tracking shows your results in Varify right away, without delay.
A paused experiment keeps collecting data, because visitors can still complete purchases after the pause. Only archiving the experiment ends data collection permanently.
Possibility 4: Sample too small
With few conversions per variation, the values of both sources fluctuate strongly. Random differences then carry a lot of weight.
→ Let the experiment keep running. As a rule of thumb for reconciling Shopify and GA4, we recommend about 1,000 conversions per variation. Only then do random fluctuations largely even out. To learn how Varify calculates significance, read the article How Varify calculates the significance.
→ If many visitors don't accept tracking, less traffic is available, and the experiment takes correspondingly longer.
Which data source should you base your decision on?
Both sources have their strengths. This overview helps you classify them:
| Question | Recommended source | Reason |
|---|---|---|
| Which variation wins on conversions and revenue? | Shopify | Shopify also records orders without consent and delivers the data almost in real time. |
| How do individual segments (e.g. channels) behave? | GA4 | Segmentation is based on the GA4 data. |
| How do new and returning customers behave? | GA4 | The "New vs. Returning" dimension follows the GA4 definition. |
Also evaluate the relative uplift instead of the absolute conversion rate. With a large enough sample and evenly distributed visitors, the uplifts of both sources are in a similar range. An uplift of about 3% in one source then corresponds to roughly 2 to 4% in the other.
Besides purchases, also look at revenue and at the performance of individual products and variations. You may also want to factor in returns. A higher conversion rate doesn't automatically mean more revenue.
If the deviation remains
- Let the experiment run until it reaches about 1,000 conversions per variation, then compare the sources again.
- If the values are still far apart, check your tracking again using possibilities 1 to 4. With an A/A test, you can test both sources against each other under comparable conditions. If your A/A test shows a significant result, the article A/A test shows significant results will help.
- If you don't want to rely on GA4 as your data basis, use Varify Realtime Tracking. It measures visitors and conversions from the same source, independent of GA4, and shows the results in Varify without delay. Besides purchase and revenue, you can measure further goals such as add to basket, checkout started and page views. You set how consent is handled in the Tracking Setup.
- If you need further help, contact our support team.
Accounts created since September 17, 2026 can activate Varify Realtime Tracking directly in the "Tracking Setup". Existing customers get the feature on request at [email protected].
Conclusion
Deviations between Shopify and GA4 usually come from different consent bases, the duplicate filter, the delay of the GA4 data or a sample that is too small, not from an error in the test. Use Shopify as the reference for conversions and revenue and GA4 for segments, and evaluate the relative uplift instead of the absolute values.