Report filters, outliers and resetting stats
Report filters narrow a campaign or widget report to a subset of traffic or conversions, such as mobile visitors, visitors from one country, or orders containing a specific product. Statistics are recalculated on the filtered data. The same panel lets you exclude outlier orders, and you can permanently remove outliers or reset a campaign's stats when the data is unusable.
How filters work
- In a report, click the filter button next to the date range, choose a filter, an operator and a value, then apply.
- Applied filters appear as pills above the report. Click a pill to edit it, or remove it. Reset clears them all.
- Every report card on the page is recalculated with the filters. Multiple filters combine with AND. Within one filter, several values combine with OR ("is one of").
- Filters are saved in the page URL, so you can bookmark or share a filtered view.
Each filter acts on one of three things, and that decides what changes in the report:
Filter acts on | Filters | Effect |
|---|---|---|
Exposures | Device, Browser, Operating system, Country, Region, City | Keeps only campaign views that happened on that device or in that location. Reach, views and conversions all change |
Visitors | Visitor type, UTM source/medium/campaign/content/term, Referrer, First visit page | Keeps only visitors with that profile. Reach, views and conversions all change |
Conversions | Goal products, Goal tags, Goal revenue, Z-score threshold | Keeps every visitor but only counts the matching goal hits. Reach stays the same, so conversion rate and revenue per visitor go down |
Available filters
Group | Filter | Operators | Plan |
|---|---|---|---|
Traffic | Visitor type | is: New visitors, Returning visitors | All plans |
Traffic | Device | is one of: Desktop, Mobile, Tablet | All plans |
Traffic | Browser | is one of | Advanced analytics |
Traffic | Operating system | is one of | Advanced analytics |
Location | Country, Region, City | is one of | Advanced analytics |
Goals | Goal products | is one of (products). Stores with a product catalog | Advanced analytics |
Goals | Goal tags | is one of (order tags). Shopify | Advanced analytics |
Goals | Goal revenue | greater than, less than, equals, between | Advanced analytics |
Goals | Z-score threshold | less than | Advanced analytics |
Acquisition | UTM Source, UTM Medium, UTM Campaign, UTM Content, UTM Term | contains, equals, starts with, does not contain, does not equal, is empty | Advanced analytics |
Acquisition | Referrer | same text operators | Advanced analytics |
Acquisition | First visit page | same text operators | Advanced analytics |
The date range is always available. Filters marked Advanced analytics are available on the Premium plan.
Filter details
- Visitor type. New visitors are those who saw the campaign as they arrived on their first visit. Returning visitors are everyone else, including visitors who saw it later in their first visit.
- UTM, Referrer and First visit page use the data saved on the visitor's profile. Referrer and First visit page are first-touch: the first external referrer and the first landing page. Each UTM parameter keeps the latest value seen in a URL the visitor opened, so a visitor who first came from Facebook and later from a Google ad tagged utm_source=google matches UTM Source contains google. See UTM parameters.
- Goal products keeps orders, carts or checkouts that contain at least one of the products. The whole order value still counts, not just those products.
- Goal tags keeps orders that have at least one of the tags, using the order's current tags in Shopify.
- Goal revenue keeps goal hits whose value matches, in your website currency. Use it to exclude test orders (greater than 0) or to study large orders separately.
- Z-score threshold keeps goal hits whose value is below that many standard deviations above your store's average. Goal hits without a z-score are kept. See Outliers.
Outliers
One unusually large order can swing revenue per visitor and AOV. Personizely scores every goal hit that has a value with a z-score: how many standard deviations it is above the average for the same kind of event across your store. Scores are calculated once your store has enough history.
When a campaign has enough conversions, no filters are applied, and some of its goal hits are unusually large (a z-score of 3 or more), the report shows Outliers detected with the number and total value of those goal hits. You then have two options:
Action | What it does | Reversible |
|---|---|---|
Exclude outliers | Applies the Z-score threshold filter (less than 3) to the report | Yes. Remove the filter |
Review → Remove N goal hits | Permanently removes those goal hits from this campaign's stats, then recalculates | No |
Removing outliers only unlinks them from the campaign. The orders stay recorded and still count for other campaigns, widgets and the Dashboard.
Decide your outlier policy before you look at the result, and apply it the same way to control and variations. Removing outliers after seeing which variation they helped biases the test.
Reset stats
Reset stats permanently deletes a campaign's recorded views and conversions so it can start measuring from scratch, for example after fixing a broken variation.
- Go to Campaigns, open the campaign's menu and choose Reset stats.
- Choose All time to remove all stats, or Specific period with a From and To date to remove only that range.
- Click Reset.
Widgets have the same option in the Widgets list. Resetting can't be undone. Visitors keep their assigned variation, so returning visitors are counted again in the same variation from then on.
Behavior details and limitations
- Filters make the verdict approximate. With filters applied, the test progress and winner verdict are hidden, with a note and a switch to show them anyway. Filtered results are a smaller sample with wider intervals. See Statistics methodology.
- The outlier banner hides while filters are active. Clear the filters to see it again.
- Filters aren't retroactive exclusions. They change what you see in the report, not what's recorded. To stop recording certain traffic, use Traffic filtering.
- Location comes from the IP address at the time of the campaign view. Visitors on VPNs appear in the VPN's location.
- Filters combine with breakdowns. Filters are applied first, then the data is split. See Breakdowns.
Examples
- Exclude internal test orders: Goal revenue greater than 1, if your QA orders use a $0 or $1 test product.
- Paid traffic only: UTM Medium equals cpc, to check how a landing-page test performs for the visitors you pay for.
- One market: Country is one of United Kingdom, to read a price test for your UK store.
- Bundle performance: Goal products is one of your bundle SKUs, to see whether a cross-sell widget drives bundle orders.