| Course | MKT 462 Digital Marketing and Advertising |
|---|---|
| Topic | Topic 3 |
| Paper type | Attribution model comparison |
| Length | About 950 words, 2 pages |
| Format | APA 7 student paper |
| School | Grand Canyon University (GCU) |
| Program | BS in Marketing |
| Updated | October 2026 |
Free sample paper for MKT 462 Topic 3
Five Models, Five Answers: What Each Attribution Model Assumes About How an Eyewear Purchase Happens
[Author Name]
Colangelo College of Business, Grand Canyon University
MKT-462: Digital Marketing and Advertising
Topic 3 Attribution Comparison
[Instructor Name]
[Date]
The organization, setting and figures below are a composite written as a model document. No real employer, client, colleague or patient is described.
Five Models, Five Answers: What Each Attribution Model Assumes About How an Eyewear Purchase Happens
The Question
A composite online eyewear retailer ran one quarter's 4,000 orders through five attribution models and got five different answers about which channel earns its budget. The marketing team wants to know which model is right. This comparison argues that the question is better put another way: what does each model assume about how a purchase happens? The chart a model produces follows from that assumption, so choosing a model means choosing an assumption. All figures are illustrative.
The Five Results Side by Side
Table 1 shows each channel's share of credit for the same 4,000 orders under each model.
Table 1
Share of Credit for 4,000 Orders by Model, Illustrative
| Channel | Last click | First click | Linear | Time decay | Position-based |
|---|---|---|---|---|---|
| Branded search | 42% | 27% | 33% | 38% | 35% |
| Non-brand search | 12% | 16% | 14% | 13% | 14% |
| Paid social | 8% | 38% | 22% | 14% | 23% |
| Display retargeting | 14% | 5% | 11% | 13% | 9% |
| 14% | 6% | 11% | 13% | 10% | |
| Direct and other | 10% | 8% | 9% | 9% | 9% |
Last Click
Last click gives all credit to the final touch before the order. Branded search takes 42 percent and paid social 8 percent. Hidden premise: the last touch did the persuading, and everything before it was irrelevant. For a shopper who searched the brand name after seeing an ad three days earlier, last click credits the search and ignores the ad.
First Click
First click gives all credit to the first recorded touch. The order nearly inverts: paid social takes 38 percent and branded search 27 percent. Display retargeting, which by design reaches only people who have already visited the site and therefore can never be a first touch for most paths, falls from 14 percent to 5. Hidden premise: whatever introduced the shopper caused the purchase, and everything after merely followed.
Linear
Linear divides credit equally among every touch in the path. Paid social and branded search land between their first-click and last-click shares. Hidden premise: every touch mattered equally, so a retargeting impression on the way to checkout counts the same as the video that introduced the brand.
Time Decay
Time decay gives more credit to touches closer to the order, often halving the weight for every seven days earlier. Its results sit close to last click. Hidden premise: influence fades with time, so recent touches persuade more than early ones.
Position-Based
Position-based models give 40 percent to the first touch, 40 percent to the last and divide the remaining 20 percent among the middle. Paid social and branded search share the extremes. Hidden premise: introducing and closing matter most, and the middle of the path matters least.
What the Swings Reveal
The size of each channel's swing across models is itself informative. Non-brand search and direct traffic barely move, between 12 and 16 percent and 8 and 10 percent, which suggests they appear in similar positions in most paths. Paid social swings from 8 to 38 percent because it almost always appears early: it introduces shoppers who then return through other channels. Retargeting and email swing the other way because they almost always appear late, reaching people who have already shown interest. A large swing does not mean a channel is good or bad. It means the channel's credit depends almost entirely on an assumption that the data cannot check.
Paths the Models Never See
All five models work only on tracked paths. Shoppers who saw a social video on a phone and later bought on a laptop without logging in appear as a laptop path with no social touch at all. Shoppers who declined tracking cookies may appear with no path. Because early, phone-based touches are the ones most often lost, every model probably understates channels like paid social, though by an amount no model can estimate from the data it has.
What All Five Share
Every model divides credit among touches that happened. None asks which orders would have occurred without a given touch. That is the question a budget decision actually needs answered: if the retailer cut retargeting spending, how many orders would it lose? A shopper who was going to buy anyway may still see a retargeting ad on the way, and every one of the five models will give that ad some credit for an order it did not cause. Li and Kannan (2014) built a more sophisticated model that estimates each channel's contribution from paths, and validated it with a field experiment, showing that rules of thumb such as last click misstate channel value. But even the best model built from observed paths assigns credit among touches; only an experiment that withholds a channel from some shoppers measures what the channel adds.
Why Experiments Are Needed and Hard
Lewis and Rao (2015) showed that measuring the return to advertising precisely is difficult even with experiments, because individual purchases vary so much that very large samples are needed to detect realistic effects. Johnson et al. (2017) developed ghost-ad methods that make such experiments cheaper by identifying, within the control group, the people who would have seen an ad. For the eyewear retailer, the practical lesson is to use attribution models to describe paths, and to reserve budget decisions for channels whose effect has been tested: for example, by pausing retargeting for a random half of past visitors for a month and comparing their orders with the other half.
Recommendation
The retailer should stop asking which model is right and report all five side by side, as in Table 1, so the team sees how much each channel's credit depends on the assumption. Channels whose credit swings most between models, such as paid social and retargeting, are the ones most in need of an experiment before their budgets change.
References
Johnson, G. A., Lewis, R. A., & Nubbemeyer, E. I. (2017). Ghost ads: Improving the economics of measuring online ad effectiveness. Journal of Marketing Research, 54(6), 867-884. https://doi.org/10.1509/jmr.15.0297
Lewis, R. A., & Rao, J. M. (2015). The unfavorable economics of measuring the returns to advertising. Quarterly Journal of Economics, 130(4), 1941-1973. https://doi.org/10.1093/qje/qjv023
Li, H., & Kannan, P. K. (2014). Attributing conversions in a multichannel online marketing environment: An empirical model and a field experiment. Journal of Marketing Research, 51(1), 40-56. https://doi.org/10.1509/jmr.13.0050
What the MKT 462 Topic 3 instructions ask for
Topic 3 of MKT-462 asks students to compare attribution models. Strong comparisons apply several models to the same orders, set the results side by side, and state the hidden premise of each about how a purchase happens. Students should explain what the swings between models reveal about each channel's position in the path, note paths the models never see, show what all rule-based models share, and explain with research why only experiments measure what a channel adds and why such experiments are difficult.
Inside the MKT 462 Topic 3 example
The same 4,000 orders from one quarter appear five times in the finished comparison, each figure labeled illustrative. Under last click, branded search takes 42 percent of the credit and paid social 8. Under first click the order nearly inverts: paid social takes 38 percent and branded search 27, while display retargeting, which by design reaches only past visitors, falls from 14 percent to 5. Linear, time-decay and position-based models land between those poles. A table places the five results side by side, and a paragraph beneath each model states its hidden premise, for instance that the last touch did the persuading or that every touch mattered equally. The comparison then makes the point all five share: each divides credit among touches that happened, and none asks which orders would have occurred without them.
Reading the MKT 462 Topic 3 rubric
Faculty reward a comparison that judges models by their assumptions. An Excellent comparison tables five models for 4,000 orders, with each column summing to 100 percent. Credit goes to stating each premise, such as the last touch doing the persuading. The rubric values explaining why paid social swings from 8 to 38 percent and retargeting from 14 to 5. Strong comparisons cite Li and Kannan and note untracked phone touches. Top marks go to Lewis and Rao and Johnson and colleagues on experiments and a retargeting holdout proposal.
MKT 462 Topic 3 help before you submit
Comparing models by their diagrams is where papers on this topic lose the most, since describing the shape of position-based credit says nothing about what it does to this retailer's channels. Running the models on invented data that fails to add up is the next weakness, and markers often check that each model's shares total one hundred percent. Drafts frequently crown one model as correct, usually data-driven, without noticing that every model on the list allocates among recorded touches. Leaving out the assumption behind each rule turns the comparison into a menu. Retargeting's swing between positions is commonly missed, although it is the clearest sign that credit reflects the rule rather than the channel. A further loss comes from recommending a budget shift on the favored model's numbers, which the testing topics later in the course exist to challenge.
Get an MKT-462 Topic 3 example written to your instructions
Send the MKT-462 Topic 3 instructions and the rubric attached in your classroom, with the campaign data or case your section provided. We write a custom example to them, with one set of conversions credited under each model, the shares reconciled, every model's assumption stated and causal limits named, in 24 to 48 hours. The first one is free.
More MKT 462 papers and BS in Marketing samples
- MKT 462 Topic 1: Tracking Coverage Audit
- MKT 462 Topic 2: Cross-device Path Analysis
- MKT 462 Topic 4: Channel Economics Comparison
- MKT 462 Topic 5: Incrementality Dq Post
- MKT 462 Topic 6: Geo Lift Test Design
- MKT 462 Topic 7: Budget Reallocation Memo
- MKT 462 Topic 8: Digital Measurement Plan
- MKT 315 Topic 3: Segmentation Variable Analysis
- MKT 445 Topic 3: Secondary Data Audit
- MKT 345 Topic 3: Motive Gap Analysis
- MKT 443 Topic 3: Identity Resolution Analysis
MKT 462 Topic 3 questions, answered
What is position-based attribution?
A rule that gives fixed shares of credit to the first and last touches in a path, commonly 40 percent each, and divides the remaining 20 percent evenly among the touches in between. It is sometimes called U-shaped. The premise is that opening and closing a path matter most. Like every rule-based model, it sets those weights by assumption, and the example shows how its results differ from linear credit on the same orders.
Is data-driven attribution more accurate?
It fits the retailer's own data rather than a fixed rule, comparing paths that converted with paths that did not and assigning weights accordingly, which is an improvement in one sense. It still works only with recorded touches, so device breaks, showroom visits and people who would have bought anyway remain outside it. Its credit describes association within observed paths, so the example still recommends testing.
Which attribution model should a company use?
The example declines to name one as correct, because each rests on a premise about how buyers decide and none measures what would have happened without the spend. Companies often keep one model for routine reporting, with its bias understood, and use experiments to check the channels where the models disagree most. The comparison ends by naming those channels as candidates for the tests that come later.
Where can I find a free MKT 462 Topic 3 sample paper?
Every part of the attribution comparison is on this page: the question, five results side by side, each model's premise, what the swings reveal, paths never seen, what all five share, why experiments are needed and hard, and the recommendation. Send your conversion data and we will compare models free.
What do all five models miss in MKT 462 Topic 3?
Each divides credit among touches that happened, and none asks which orders would have occurred without a given touch. Only an experiment that withholds a channel measures what it adds.