“How’s the dish?”, I asked inquisitively to my father with a spark in my eyes. I assumed that I was about to be credited for making one of the most spectacular dishes of the evening.
“I was the one who chopped the vegetables!” my younger brother voiced out to ensure that he gets his share of credit.
“Hello lockdown chefs, don’t forget who guided you throughout”, my mother exclaimed with a pinch of sarcasm. Confused left, right and center, my father uttered “Great teamwork guys”.
I wish that answer would have sufficed in digital marketing (DM) as well but sadly it does not work that way. Every marketer wants to increase ROI (return on investment) of his media spend. As a marketer runs a campaign on TV and varied media channels, a detailed percentage break up of which channel contributed to conversion indeed becomes pivotal in increasing ROI.
From the above table, it becomes clear that a marketer needs to push its budget more on digital assets such as Google Ads, Facebook Ads etc. as compared to TV as they are giving a higher ROI. Dissecting digital assets, a marketer needs to find out which asset is contributing the most. In marketing terms, it is termed as attribution. It helps in knowing the measurable contribution of each media intervention in making the sale happen. Remember we brought out an article on marketing attribution this earlier? So get set as this would be a long but an interesting read.
Let’s say Nidhi is looking for a birthday gift for her friend. As a consumer, she will have a journey. She would not buy something at the first instance when she is in the awareness and consideration phase.
- She will type “Best birthday gifts to buy online” as a query on Google. An ad pops up from the platform Gourmet Basket.
She clicks on the ad and it suggests a make-up kit. However, she realizes that giving a set of lipsticks would be a better idea.
2. A while later, she searches for “Best lipstick set under 2k”. Another ad pops up from Gourmet Basket. She gets interested in a Lakme gift set offered on it.
3. A few days later, she searches for “Best price for Lakme lipstick set” and Gourmet does a perfect job of putting a beautiful ad with an offer in front of her. With a priceless laughter and a blemish smile, she clicks on the ad and buys it.
The long and short of the story is “Which ad will get the credit of making the conversion?” Which ad actually made the purchase happened? Is it ad 1 or ad 2 or ad 3?
A similar conundrum can occur across media channels as well. Let’s say OnePlus is advertising on different social media platforms and Google. As a digital marketing manager of OnePlus, you see that 100 people saw the Google ad, clicked on it, and bought the handset. Is it right to give all the credit to Google ads for the conversion?
What about those people who have seen that ad on Facebook already and then saw the Google ad and got converted? Won’t it be unfair to give all the credit to Google? Well, different attribution models exist around giving credit to different media interventions. These attribution models provide a framework to decide which media touch point should receive credit and how much. Let’s deep dive into them one by one:
- Last Click attribution model
As the name suggests, all the credit of conversion is given to the last interaction of a user before getting converted. For example: a user might have discovered you on Facebook, saw you again via a YouTube ad, and finally visited your website and converted. In this model, 100 % credit will be given to website organic traffic for getting a convert. Figure 1 above is a classic case of a last click attribution model. It is worth to note that it is the default attribution model on a majority of analytics platform.
Although simple to implement and evaluate, the model does not give a very clear picture. It ignores all the other interactions that might have played critical role in making the conversion happen.
- First click attribution model
As opposed to last click model, this model gives all the credit to the first interaction with the user and ignores all the subsequent interactions. A user might have read about you on a blog, received an email and then got converted. The model will just consider blog to be a rock star as it will be given 100% credit for the convert.
Though the model is good for companies having a short buying cycle, it discounts the importance of any media touch point after the first interaction.
- Time decay attribution model
A smart folk once said that time and tides wait for none. Well, time waits to decay in this model. The model gives more credit to the touch point that was nearest to the conversion and will keep decreasing it as the touch point is farther away from the conversion. For example: A user got an email, read about it on a blog, found it on Pinterest and finally got converted. The time decay model will give maximum attribute to Pinterest followed by blog and mail.
This model might be useful when relationship building might be an important parameter in sales (more common in B2B). However, this model discounts the contribution of top of the funnel activities (ToFA) concentrated around awareness & consideration stage.
- Linear attribution model
If Mahatma Gandhi would have been alive, this would have been his favourite model as it works on the principle of equality. The model distributes the credit equally among all the touch points that a user had before getting converted.
Linear distribution indeed gives a balanced approach to look at importance of the different channels. However, some channels are always more effective in making conversion than the other channels. This model fails to capture that relative importance.
The problem with all these attribution models is that they fail to capture the “true attribution” of a channel. However generally speaking, time decay/linear model are better than first/last click attribution model. So, the next prudent question is which model to follow? The model name is Data driven model (DDM) as it backs up all the contribution percentage by data.
Data driven model
A company might have a lot of data from the past regarding different paths followed by the user before getting converted. Some users might have converted directly after reading an email where others might have seen a search ad followed by a display ad before getting converted.
All the unique paths are captured and classified into different clusters (teams). Each cluster presents a unique path followed by the user. This approach ensures that all the unique journeys are considered.
Now, find the probability of all the teams. For ex: probability of conversion for team 1 is 2/5= 0.4. Similarly, the probability for team 2 and team 3 is 0.33 and 0.25 respectively. In this case, team 1 consisting of search ad followed by display ad & an email has a higher probability of converting.
Additionally, we can find the contribution of individual channel by looking at teams from micro lens.
The presence of display ad in the centre increased the probability of conversion by 0.07 (0.40 – 0.33). So, the counterfactual gain due to display ad is 0.07. Similarly, we will find probabilities of all the channels by using different clusters (assume it came 0.15 for search ad and 0.18 for email).
The picture seems much clear now as it gives percentage distribution of different channels in making the conversion. The good thing about the model is it does not assign fix weight to a channel and the projections are made based on data of our own campaign. Additionally, it factors in all the paths that did not result in the conversion.
One pertinent thing to note is all these models do not take offline media into picture. A user might have seen a billboard and searched for a product and bought it. In this case, the billboard is not attributed.
This article has been authored by Sumit Gattani
Sumit Gattani is currently pursuing Marketing from SIBM Bengaluru. An ex-Merchant Navy officer & a curious soul, he has a deep interest in human psychology and his Google search history is usually filled with statements starting with “Why..”. Occasionally, he also writes about things that catch his fancy.