When working in online advertising, if you try and evaluate individual channels as effective or non effective you are in danger of losing efficiency and money.
It is much more important to look at “the big picture” and to analyse how and when consumers use individual channels in the different phases of the purchasing process.
Both in advertising research as well as in practice it is increasingly recognized that individual advertising channels should not be examined separately. Users interact with brands and products via multiple channels – driven by growing multimedia consumer behaviour – creating numerous contact points.
Longer running display campaigns, for example, can be used to draw attention to new products, whereas specific search queries often come at the end of a purchasing process. Under the last cookie wins model, search queries are attributed more conversions than they are actually responsible for. This illustrates that this method does not adequately evaluate the effectiveness and the efficiency of marketing measures.
Individual attribution models are more effective because they are more precise. By placing a different emphasis on the touch points, other channels will inevitably be rewarded as well, thereby changing their share and contribution to the overall sales. By using individual attribution models, advertisers can find out which channels initiated the purchase, which ones prepared the purchase and which ones profited from the purchase.
First Cookie Wins
Obviously, if we’re looking for the opposite of the “last cookie wins” we have to look at the “first cookie wins” model where the sale will be attributed to the first ad a user clicked on, but what is the difference?
The following charts represent a hypothetical example of different conversion attribution models and their effects on conversions and costs, starting with first cookie wins vs. last cookie wins.
The table below compares two attribution methods, last cookie wins and first cookie wins. Under the last cookie wins model the conversion is attributed to the last “touchpoint” with the ad creative and vice versa. The table shows the difference between two hypothetical scenarios and shows which conversions (per channel) and costs (per channel) arise under the different attribution scenarios.
While under the last cookie wins model retargeting is considered one of the two best channels, under the first cookie wins model retargeting is less efficient. The same applies for brand search. Display advertising and Search generic gain sales and efficiency under the first cookie wins model, which underlines their awareness generating effect.
This is all interesting reading, but can be very confusing for the client – But which should an advertiser choose? Even if you have data about one’s advertising activities available, conducting the most difficult calculations, and having large information about the consumer behaviour, you will never find a clear answer to this question due to the fact that even consumers cannot clearly say why they bought a product from a specific provider.
If every product buy could be attributed to the first click or the first recommendation, only the first cookie wins model would be applicable. But consumers change their mind, they meet new providers and products while deciding which product to buy, and they are also strongly influenced by their direct environment. A consumer does not determine his or her opinion after one impression. On the other hand, you can also argue against the last cookie wins model. How would a channel, such as retargeting, receive 100% conversion, if the user didn’t even have the chance to see a retargeting banner without upstream channels (i.e. affiliate or SEA generic)? Also SEA brand clicks are always the result of brand awareness, created by other advertising channels and PR.
Therefore, it is recommended that advertisers use the uniform distribution model, where every click on an ad during a customer journey receives the same conversion share. This method allows you not to limit yourself to one individual channel, but instead look at the interaction of all channels. This should be the basis for a more efficient budget allocation. Other models, such as the uniform distribution or completely individual distributions should only be used in exceptional cases.
The second table, shown below, compares the last cookies wins model with the equal attribution of conversions. The latter attaches equal importance to each channel within the customer journey. The table shows how the different models affect conversions (per channel) and costs (per channel).
The uniformly distributed allocation model is the best compromise between the first cookie wins and the last cookie wins model, since it also takes those channels into account that appear in the middle of the purchase decision process. Initiating and spreading channels, such as display advertising and Search generic, benefit from the uniformly distributed evaluation compared to the last cookie wins model, while the efficiency of absorbing channels, such as retargeting and SEA brand, is downgraded.
This knowledge helps advertisers determine the effectiveness and efficiency of individual channels, and to allocate online marketing budgets based on these results.
The matrix below shows how efficient and effective each channel is, the cost per order of the different cookie scenarios are shown in relation to their share of the overall sales percentage. Therefore it is possible to evaluate each channel according to its efficiency (cost per order) and effectiveness (percentage in relation to overall sales) within the different cookie scenarios.
You can see that in the uniformly distributed attribution model display advertising is responsible for many sales, but at relatively high costs compared to affiliate. Shifting budget from display to affiliate would allow generating conversions way more cost efficiently, increasing also the share of affiliate sales.
At the same time you could relocate a significant share of the SEA generic budget to display and affiliate. Since all three channels address the user at similar points, adverse effects are hardly expected.
However, you need to be careful when allocating to absorbing channels, i.e. SEA brand and retargeting. It may seem logical to invest more into these channels, since their CPOs are considerably low compared to other advertising instruments. But traffic in these channels is often limited; it therefore happens frequently that the budget cannot be completely allocated. Also, it would lead to cuts in awareness generating channels, causing a drop in image and brand awareness.
To sum up, a multi-channel analysis together with the uniformly distributed attribution model has the big advantage of taking into account every channel in the customer journey by simply placing equal importance on each channel. This allows advertisers to decide how to allocate your budget more efficiently. It would be advisable to record CPO modifications, at least on a monthly basis, to reach the budget optimum.