Step 1: Install the app
Install Arktic from the Shopify App Store. After installation, you will be redirected to the Arktic dashboard automatically.
If you are not redirected, open Apps in your Shopify Admin and click Arktic.
Step 2: Enable the app embed
The app embed is the JavaScript that runs on every storefront page. It handles visitor identification, variant assignment, and all event tracking. Without it, no experiments will run and no data will be collected.
- In your Shopify Admin, go to Online Store → Themes
- On your live theme, click Customize
- In the Theme Editor, click the App embeds icon in the left sidebar (the puzzle piece icon)
- Find Arktic in the list and toggle it on
- Click Save in the top-right corner
The app embed only needs to be enabled once per theme. If you later switch your live theme, you will need to re-enable it on the new theme.
Verify it is working
After enabling, visit any page on your storefront. Then return to the Arktic dashboard and open the Get Started page. Step 1 will show a green checkmark once the dashboard receives its first storefront event — this usually takes under 30 seconds.
You can also verify directly in your browser: open the browser console on any storefront page and type window.__SPT_VID__. If it returns a string (your visitor ID), the embed is active.
Step 3: Create your first experiment
Go to Experiments in the sidebar and click New experiment.
Fill in the form:
| Field | What to enter |
|---|
| Name | A short internal label, e.g. “Homepage hero CTA test” |
| Hypothesis | What you expect to happen and why, e.g. “Changing the CTA from ‘Shop now’ to ‘View collection’ will increase clicks” |
| Type | Choose the experiment type that fits your test. See Experiment types if you are unsure. |
| Traffic allocation | Start with 100% to get data faster. Lower this if you want to limit exposure. |
| Segment | Leave blank to include all visitors, or pick a saved segment. |
After creating the experiment, you will land on the experiment detail page. Every experiment has at least two variants:
- Control - the current experience, unchanged. No setup needed.
- Variant B - the treatment you are testing. Click Configure to set it up.
What you configure depends on the experiment type:
| Type | What you configure on Variant B |
|---|
| Theme | Select a duplicate unpublished theme |
| URL Redirect | Enter the destination URL |
| Price (Plus) | Choose the product and set a price adjustment |
| Price (non-Plus) | Choose the source product and price adjustment (app duplicates automatically) |
| Section content | Paste the HTML to inject into the Variant Content block |
| Template | Select an alternate page template |
For Theme tests, you must first duplicate your live theme in Online Store → Themes and make your changes to the duplicate before you can select it here.
Step 5: Start the experiment
On the experiment Overview tab, click Start experiment. The status will change to Running.
From this point, visitors will be assigned to variants and tracked. Assignments are persisted in a cookie for one year, so returning visitors always see the same variant.
Step 6: Wait for data
Check back after a few days. A few important rules:
- Do not stop early. Results are unreliable in the first few days. Wait at least 7 days before drawing any conclusions.
- Do not change variant weights mid-experiment. This invalidates any data collected before the change.
- Refresh results manually if you want the latest numbers before the hourly rollup runs: go to the Results tab and click Refresh results.
Step 7: Read your results
Go to the Results tab. The table shows one row per variant with sessions, orders, CVR, revenue per visitor, and statistical significance.
When p < 0.05, the result is statistically significant. The conclusion banner at the top of the page will tell you which variant is winning.
See Reading results for a full explanation of every metric and how to interpret them.
Step 8: End the experiment
Once you have a clear winner with significant results and at least 7 days of data, go to the Overview tab and click Complete experiment (if the test variant won) or Archive experiment (if the control won or you want to discard).
- Completed - the winning variant is now the default. For theme tests, publish the winning theme manually. For price tests, update the product price manually. For redirect tests, you can remove the experiment.
- Archived - the test variant is discarded. All visitors revert to the control immediately.
For price experiments on non-Plus stores using the auto-setup option, the duplicate product is deleted automatically when you complete or archive.
Troubleshooting
Visitors are not being tracked
- Confirm the app embed is enabled and saved in the Theme Editor.
- Check for browser extensions that block tracking scripts (ad blockers). Test in an incognito window.
- Check the browser console for JavaScript errors from Arktic.
Results are not updating
- Click Refresh results on the Results tab to trigger an immediate rollup.
- Make sure you have at least a few sessions recorded. Results will not appear until there is data.
The embed shows as active but experiments are not running
- Confirm the experiment status is Running (not Draft or Paused).
- Make sure variant weights sum to exactly 100.
- If using a segment, make sure your test visitor matches the segment rules.
