A customer discovers you through organic search. Two weeks later, they click a social post. Then they return through a paid search ad, open an email, and finally type your address into the browser before buying.
One journey. One $100 sale. Five channels touched it.
Now ask a simple question: which channel generated the revenue?
There is no neutral answer hiding in the data. There is only a rule for assigning credit. Change the rule, and the same sale tells a different story.
That is the part attribution reports often make too easy to forget.
One journey, five touchpoints
For a concrete example, use this path:
- Organic Search, 28 days before the purchase
- Organic Social, 14 days before the purchase
- Paid Search, 7 days before the purchase
- Email, 2 days before the purchase
- Direct, at the moment of conversion
The order and timing stay fixed. The purchase value stays $100. Only the attribution model changes.
Zenovay currently supports five models in the revenue view: last touch, first touch, linear, position based, and time decay. We ran this example through the same allocation function used by the product.
Here is what each model reports.
| Model | Organic Search | Organic Social | Paid Search | Direct | |
|---|---|---|---|---|---|
| Last touch | $0 | $0 | $0 | $0 | $100 |
| First touch | $100 | $0 | $0 | $0 | $0 |
| Linear | $20 | $20 | $20 | $20 | $20 |
| Position based | $40 | $6.67 | $6.67 | $6.66 | $40 |
| Time decay | $2.37 | $9.50 | $18.99 | $31.16 | $37.98 |
None of these rows changes the sale. Each row answers a different question about it.
Last touch asks who closed
Last touch gives all credit to the final known channel. In this journey, Direct receives the full $100.
That can be useful when the question is operational: what brought the customer back at the moment they bought? It is also simple to explain and works when only the current visit is available.
But it can make discovery disappear. Organic Search introduced the product. Social, paid search, and email kept the journey moving. The report still gives them zero.
There is another wrinkle with Direct. A direct visit is often real, but it does not necessarily mean the customer arrived without prior influence. They may have remembered the brand after seeing the earlier touches. Last touch cannot express that history when it assigns everything to the final visit.
First touch asks who introduced
First touch reverses the story. Organic Search receives the full $100 because it was the first known channel in the journey.
This view is useful for discovery. It helps answer which channels bring new people into the measurable path. It can be particularly helpful when evaluating content, communities, or campaigns whose main job is creating awareness.
Its blind spot is the close. An early visit may have introduced the customer, but a later email or paid campaign may have done the work that converted interest into action. First touch cannot share that credit.
Linear refuses to choose
Linear attribution divides the sale evenly across the five distinct channels. Each receives $20.
This is the calmest model in the set. It acknowledges every measured touch and avoids claiming that one position is always more important than another.
That fairness is also its assumption. A quick social click receives the same credit as the email opened two days before purchase. Equal credit is not the absence of a judgment. It is a judgment that all measured touches deserve the same weight.
Position based rewards the beginning and the end
Zenovay’s position based model gives 40 percent to the first channel and 40 percent to the last. The remaining 20 percent is split across the channels in the middle.
In our example, Organic Search and Direct receive $40 each. Organic Social, Paid Search, and Email share the remaining $20.
This model fits a common marketing intuition: discovery and conversion matter most, while the middle assists. It preserves both ends of the journey without ignoring everything between them.
But 40, 20, and 40 are still chosen weights. They are not facts observed in the visitor’s mind.
Time decay rewards recency
Time decay gives more credit to touches closer to the conversion. Zenovay uses a seven day half life. A touch seven days before the sale receives half the raw weight of a touch at conversion. A touch 14 days before receives one quarter. A touch 28 days before receives one sixteenth.
After normalizing those weights across this journey, Direct receives $37.98, Email $31.16, Paid Search $18.99, Organic Social $9.50, and Organic Search $2.37.
This is useful when recent touches are more likely to influence a decision. It also avoids the hard cliff of last touch, where every earlier interaction instantly becomes worth zero.
The model still embeds a belief: recent interactions deserve more credit, and seven days is the right pace of decay. That may fit one buying cycle and misrepresent another.
The model should follow the question
Teams often ask which model is correct. A better question is what decision the report needs to support.
Use first touch when you are studying discovery. Use last touch when you are studying the final return or close. Use linear when you want a neutral view of all measured participants. Use position based when the first and last touches are strategically important. Use time decay when recency should matter.
The dangerous move is changing models until a preferred channel looks good. If a paid campaign disappoints under last touch, switching to linear may make its revenue rise without changing a single purchase. The new number can be valid under the new rule and still be a poor basis for comparison with the old one.
Pick the question first. Keep the model visible. Do not compare periods that use different rules as if the measurement were unchanged.
History limits matter before model choice
A multi touch model can only allocate credit across touches the system can connect.
In Zenovay, first party cookie tracking can preserve a journey across days. Cookieless mode has less history, so its multi touch result may be limited to the visits that can be linked within the available window. The revenue view marks this as limited history when appropriate.
That does not make the calculation wrong. It makes the boundary of the evidence important. A precise allocation over an incomplete path is still incomplete.
Last touch is less sensitive to missing earlier history because it needs only the final known touch. First touch changes meaning when earlier visits are unavailable. Linear, position based, and time decay can redistribute credit only among the touches they can see.
What we compare in practice
We do not use one model as a machine for discovering truth. We compare models to find where a conclusion is fragile.
If Organic Search performs well under first touch but nearly disappears under time decay, it is probably strong at discovery and distant from conversion. If Email rises sharply under time decay, it may be helping late in the journey. If a channel remains meaningful across all five models, confidence in its role is stronger.
This comparison is more useful than arguing about one universal winner. The disagreement between models is information.
Before acting on an attribution report, we now ask four things:
- Which business question are we asking?
- Which allocation rule matches that question?
- How much visitor history was actually available?
- Would the decision change under another reasonable model?
If the answer to the fourth question is yes, the conclusion needs more caution, not more decimal places.
Attribution is a lens, not a receipt
The $100 sale happened once. The five reports are not five competing transaction records. They are five lenses placed over the same journey.
Good attribution makes the lens explicit. It preserves the underlying conversion, documents the rule, and lets the reader see how much the answer depends on that rule.
The goal is not to find a model that ends the discussion. It is to choose a model that makes the discussion honest.
