Attribution explained

The question attribution answers

A customer sees a YouTube ad on Monday, clicks a Google search ad on Thursday, opens your email on Saturday, and buys on Sunday. Which channel gets the credit?

Attribution is the set of rules you use to answer that. There is no single right answer, which is why Orkie gives you several models side by side rather than one number to argue about.

The models

Each model divides credit for a sale across the touches that came before it.

1. First touch. All credit to the first visit. Best for understanding what introduces new people to your brand.
2. Last touch. All credit to the final visit before purchase. Best for understanding what closes.
3. Linear. Equal credit to every touch. A fair default when you want to see every channel that participated.
4. Time-decay. More credit to touches closer to the sale, less to older ones. Useful for longer consideration cycles where recent nudges matter most.
5. Position-based. Heavier credit to the first and last touch, with the remainder shared among the middle. Rewards both discovery and closing.

Orkie also shows an "influenced" view. It answers a simpler question: did this channel appear anywhere in the journey? That is useful for channels like video and social that rarely close but often start the story.

Lookback windows

A lookback window is how far back Orkie looks for touches before a sale. A seven-day window ignores a visit from three weeks ago. A ninety-day window includes it.

Choosing a window is a business decision:

• Short windows (7 to 14 days) fit impulse purchases and low prices.
• Medium windows (30 days) fit most stores.
• Long windows (60 to 90 days) fit considered purchases, B2B, and high prices.

You can change the window and see how credit shifts. Big swings usually tell you something about how long your customers take to decide.

Each sale is counted once

This is the part most tools get wrong, and it is why platform reports never add up.

Orkie holds a single record of each order. That record is confirmed against your store, so a test order, a duplicate event, or a browser refresh does not become a second sale. Refunds are applied back to the original order. Every attribution model then divides that one confirmed order across its touches.

Orkie calls this a canonical order. Whatever model you look at, the total revenue is the same. Only the split changes. If you sum every channel under any single model, you get your real revenue, not 180 percent of it.

Why platforms report more than you sold

Each ad platform counts conversions its own way, and each one only sees its own touches. Some common reasons the numbers disagree:

• Overlap. Google and Meta both claim the same customer because both were in the journey.
• View-through credit. A platform counts a sale because someone saw an ad, even without clicking.
• Different windows. One platform uses 7 days, another uses 28.
• Modeled conversions. Platforms estimate sales they could not observe directly.
• Refunds. Platforms rarely subtract them.

None of this makes platform reports useless. They are good at optimizing within their own walls. They are not good at telling you the whole story.

Reconciling with what platforms report

Orkie keeps the platform's numbers next to its own so you can see the gap and understand it.

A practical way to read the comparison:

1. Start from confirmed revenue in your workspace. That is the anchor.
2. Look at each platform's claimed conversions beside its attributed share under your chosen model.
3. Expect platforms to claim more. The size of the gap, and whether it is growing, is the signal.
4. Use the "influenced" view to give credit to channels that start journeys but rarely finish them.
5. Feed consented conversion data back to platforms so their optimization is based on real sales.

Your operator reviews this with you and will recommend a primary model for decision making, usually position-based or time-decay for stores and first touch for lead generation.

Where attribution shows up

Attribution powers more than one report. Product pages show which sources sell which items. Customer records show the journey that produced each order. Audiences can be built from acquisition source. And your AI assistant can answer "which channel drove our best customers last quarter" without you exporting anything.

Sales that arrive with no signal

Sometimes an order has no traceable touch: a phone order, a walk-in, a marketplace sale. Orkie does not force these into a channel. They are reported honestly as unknown, with an explanation of why, and a mapping tool lets you classify recurring patterns so they land in the right place going forward.

Want to see your own channels under all five models? Talk to us or contact us.

---

Last reviewed 2026-09-20. Something unclear? Tell us and we will fix the article.