Attribution settings sit on top of the attribution model and answer two different questions:
Which touchpoints are eligible for credit at all? This is controlled by Is Paid Only and the lookback window.
What day does the credit land on? This is controlled by Cash vs. Accrual mode.
Because they are independent of the model, you can hold the model fixed and change your numbers materially just by adjusting a setting. The sections below cover each one, then explain when the settings are active and why some orders show up as Undefined.
Is Paid Only
When Is Paid Only is enabled, only paid touchpoints are eligible to receive credit. Organic and direct touchpoints are removed from the path before the model runs, so they receive nothing regardless of where they sat in the journey.This is a toggle you can apply to any model.
What "paid" means here
The setting relies on a flag Polar attaches to every touchpoint, IS_PAID. The flag does not mean "this came from media spend." It means:
This touchpoint resolved to an identifiable marketing campaign, or to an ad-platform channel.
Counts as paid | Does not count as paid |
Ad-platform clicks: Google, Meta, TikTok, Bing, Snapchat, Pinterest, Amazon, Criteo | Direct (typed URL or no referrer) |
Any touchpoint whose ad ID resolves to a campaign, adset, or ad | Organic search |
Any UTM-tagged link that resolves a campaign name, even from a non-ad source | Organic social referral with no click ID and no UTM |
Klaviyo flows and campaigns that resolve a name (email / SMS) | A plain referral (blog, news, partner) with no |
Ometria (email / CRM) | Anything whose source, medium, or campaign contains |
View-through vendors (Tatari, TV Scientific, Cohere One, Pebble Post, LS Direct, Podscribe) | Email / SMS with no identifiable flow or campaign |
Edge cases worth knowing
A few behaviors surprise people the first time they look closely:
Email and SMS are often paid. Klaviyo flows and campaigns resolve a campaign name, so they pass the rule, and Ometria is whitelisted outright. If your mental model of "paid" is media spend, seeing owned email show up as paid is the surprising part, but it is working as designed.
UTM-tagging anything makes it paid. Put a
utm_campaignon an owned or organic link and the campaign resolves, which flips the flag to paid, whether or not any money changed hands. This is the most common reason a touchpoint looks paid when you did not expect it to.The
organicguard is the safety net. Any touchpoint whose source, medium, or campaign containsorganicis excluded from paid attribution, even when other signals are present.
When Is Paid Only is on and an order has no paid touchpoint in its path, that order is attributed to Undefined. See Undefined orders below.
Lookback Window
The lookback window sets how far back, before a conversion, a touchpoint can sit and still be eligible for credit. It is measured backward from the conversion timestamp. A 30-day window means any interaction in the 30 days before the conversion is in scope; anything older is dropped.
The window does not change how the model splits credit. It changes which touchpoints make it into the path in the first place.
You can set the window to any number of days, or choose an unlimited window, in which case every touchpoint attached to the conversion is eligible.
The length you pick biases the result in a predictable direction. A longer window keeps more upper-funnel touchpoints alive in the path, so awareness channels accumulate more assist credit. A shorter window collapses the path toward the most recent interactions and pushes any model toward last-click behavior in the limit.
Exception: the Full Impact model does not accept a lookback window. All touchpoints are eligible by default.
When the only touchpoints attached to an order fall outside the chosen window, that order is attributed to Undefined. See Undefined orders below.
Cash vs. Accrual
This setting controls the date that conversion and revenue credit lands on. The model still splits the credit the same way; Cash and Accrual only change which day each piece of credit is reported against. The names mirror cash-basis and accrual accounting in finance.
Cash assigns all credit to the day the order was placed (the day of the transaction).
Accrual assigns credit to the day each contributing touchpoint occurred (the day of the interaction).
A worked example
A shopper interacts with three channels over six days, then buys. We will use the Linear model, so each of the three touchpoints earns an equal share: one third of the order, and €40 of the €120 revenue.
May 4 — clicks a Meta ad
May 6 — clicks a Google ad
May 9 — clicks a Klaviyo email and places a €120 order
Cash mode dates everything to May 9, because that is when the order was placed:
| May 4 | May 6 | May 9 |
Meta | €0 | €0 | €40 / 0.33 orders |
€0 | €0 | €40 / 0.33 orders | |
Klaviyo | €0 | €0 | €40 / 0.33 orders |
Accrual mode dates each share to the day its touchpoint happened:
| May 4 | May 6 | May 9 |
Meta | €40 / 0.33 orders | €0 | €0 |
€0 | €40 / 0.33 orders | €0 | |
Klaviyo | €0 | €0 | €40 / 0.33 orders |
Which one to use
The two modes answer different questions, and neither is more "correct" than the other.
Use Cash for reporting and cash flow. It tells you what money came in on a given day and which channels touched those orders. It matches how most ad platforms and finance teams report, and it is the natural fit for business-level metrics like blended efficiency, where you care about money in the door on the day it arrived.
Use Accrual for optimization. It moves each channel's credit back to the day its touchpoint happened, so daily performance reflects when the marketing actually did its work rather than when the sale closed. This is the mode to use when scaling paid media or comparing channel performance over a recent window.
The gap between the two grows with your conversion cycle. If customers typically take a few days to buy, the difference is small. If they take weeks, a strong revenue day in Cash mode may be the delayed result of marketing you ran much earlier, which Accrual would have credited to those earlier days.
When are attribution settings active?
The attribution settings appear whenever an eligible metric is brought into a report, for example Polar Pixel conversions. Metrics that do not run through Polar's attribution, such as GA4 sessions, do not show the settings at all.
There is also a case where the settings are present but not doing anything. If a report contains Polar Pixel conversions with no marketing breakdown (no channel, campaign, or similar dimension), there is nothing to allocate credit across, so the model selection is effectively inactive. Total conversions under any model equal total last-click conversions, because there is no distribution to perform.
When a setting is actively changing the result, Polar shows it in purple so you can tell at a glance that attribution is shaping the numbers.
Undefined orders
An order is attributed to Undefined when it has no eligible touchpoint to receive credit. There are two reasons that happens:
No touchpoints were tracked for the order, so there is nothing to attribute.
No touchpoint is eligible under the current settings. The touchpoints exist, but the settings filter them all out.
The second case is the one to watch, because it is a direct consequence of your settings:
If Is Paid Only is on and every touchpoint on the order is non-paid, the order is Undefined in that view.
If a 7-day window is selected and every touchpoint sits more than 7 days before the conversion, the order is Undefined in that view.
In both cases the order is not lost. It moves out of Undefined as soon as you loosen the setting that excluded its touchpoints.