Choosing an attribution model: When, not which
Attribution is important, no matter your industry or your buying cycle. As more attribution models are created, it becomes more difficult to understand which one is the right one, and it seems that conversations about attribution models tend to occur with the expectation that each person has chosen a camp and believes in one particular model. The reality, though, is that no single attribution model can be the standalone source of truth — because no single attribution model can paint the full picture.
Part of the reason that attribution models are so difficult to agree on is what they are designed to do: to determine what is adding the most value. Each person, team or agency is likely to be partial to the attribution model that gives the most credit to their work.
Obviously, if you bring your team — or your agencies — into one room to agree on an attribution model, you’ll make them very uncomfortable. So how can you decide which model to use? The truth is that the question should not be about which model to use but when to use each model.
Before we dig into that further, let’s walk through the most common models and their use cases.
- What does it measure? This attribution model is focused on attributing all credit for leads/sales generated to the original source.
- When to use it: Use this model when you want to understand which of your sources are generating new prospects.
- Example: If you were in a position in which you needed to expand your reach with net-new prospects, you could use this model to determine which channels were most likely to produce new leads/sales. This model is especially helpful if you are able to tie lead generation to sales so that you can understand not only which sources are producing the most leads, but more importantly, which sources are driving the most prospects that are most likely to buy.
- What does it measure? This attribution model is focused on attributing all credit for leads/sales generated to the last source that touched the lead.
- When to use it: This model is helpful when you want to understand which of your sources (or even campaigns) are the closers, so to speak.
- Example: If you have a lot of leads in your pipeline already, and you aren’t sure which sources are ultimately driving the final conversion, this would be a good model to leverage. This is also a good model in an instance where you have limited budget, your brand and pipeline are established, and you need to focus your marketing budget only on channels that are most likely to convert.
- What does it measure? This attribution model attributes a higher value to the first and last touch (typically 40 percent each) and divides the remaining value among all touch points in between. Depending on the software or platform you use, you can sometimes customize the value that you choose to assign to first/last touches versus the touches between.
- When to use it: When you want to allocate value to all of the channels that impact your buying cycle, as opposed to giving all credit to first or last, but you acknowledge that the first and last touches have more value than mid-funnel, or nurture, touches.
- Example: If you wanted to look at end-to-end performance, this might be a model that you’d use.
Multi-touch: Evenly weighted
- What does it measure? This attribution model gives equal weight to every marketing touch point, regardless of the role that it played in the sales cycle.
- When to use it: If you want to simply look at all of the channels that are contributing, this could be a good model for you. Before you cut any channels, you might take a look at this model (or another multi-touch model) to make sure that you aren’t taking performance for granted.
- Example: If you typically rely on a first-touch or a last-touch model, you might take a look at this model before cutting a channel — otherwise, you could accidentally undervalue a channel.
- What does it measure? This attribution model is similar to the U-shaped model, with one slight change: it attributes a higher value to the first click, the click that converts the lead, and the last touch (typically 30 percent each) and divides the remaining value among all touch points in between.
- When to use it: This is a good model for businesses with a lead waterfall where there is a standardized process for how leads are handled. Without a standardized process, the middle peak attribution score would be hard to do in an accurate and consistent manner.
- Example: This is another model that is good for holistic attribution. For example, if you had a set process for which leads are nurtured until they are handed over to the sales team, you might use this model to give elevated weight to the lead gen source, the source that triggered the flip to the sales team and then the last touch — while still giving credit to all other channels.
Multi-touch: Time decay
- What does it measure? The time decay model attributes the value across all channels but assumes that interactions closer to the sale are increasingly more valuable, thus giving more weight to those channels.
- When to use it: If you have a long sales cycle, and you’re trying to understand which channels pull leads out of nurture phases, this could be a good model for you.
- Example: If you are bringing in a lot of leads, and the conversion rates are decent, but the sales cycle is drawn out, you might use this model to see which channels have the greatest impact on moving things along. Alternatively, if you have a lot of prospects which you believe to be high-quality traffic, but poor conversion rates from traffic to lead (or marketing prospect to marketing qualified lead), then you might consider looking at this model to see which channels are contributing most to warming up prospects in order to convert them to leads.
Instead of choosing which, choose when
As I mentioned above, different attribution models are built with the purpose of assigning credit based on varying valuations of the impact on different phases of the sales cycle. (This is not unlike the way different campaigns are built to support different pieces of the sales cycle. Do you see where I’m going here?)
Unsurprisingly, most people believe that the part of the sales cycle they work in is the most important and are therefore most likely to leverage a model that benefits their reporting. There are two problems that can arise from this practice:
- If you use only one attribution model, you may not be getting the full picture, which could result in overvaluing or undervaluing sources, channels or campaigns. For example, if one lead is generated by display and closed by paid search, and you use a first-touch model, then the display campaign receives full credit — and you risk undervaluing paid search while overvaluing display.
- If you (or your agencies) use different attribution models that benefit each different campaign, source or channel, and then you aggregate the data from all of these reports, you risk over-reporting by double-counting. To use the same scenario as above, if one lead is generated by display and closed by paid search, and you use a first-touch model to report on display and a last-touch model to report on paid search, then you’ll count that same conversion twice.
By now, you’re probably wondering where I’m going with all of this. Which attribution model is the right one? The truth is that it varies from situation to situation, and here’s the key: it depends on what question you are trying to answer. That’s right — I’m suggesting that you should use multiple, different attribution models to answer different questions.
For example, you likely do want to know which channels are your first-click channels. You might be looking to answer the question, “Which channels generated new prospects?” Look at the first-touch attribution model to answer this question. You likely also want to know which channels are your closers. Look at the last-touch attribution model to answer this. However, do understand that those two reports are giving you different data sets and should not be added together.
At some point, it is likely that you’ll want to look at a higher level snapshot of multi-channel performance, and any time you are using an attribution model to report on overall performance, I suggest looking at a multi-touch attribution model, as it will be more comprehensive. The mutli-touch attribution model you choose really depends on your sales cycle and, again, the questions you are looking to answer.
Leveraging multiple reports to answer different questions allows you to obtain more meaningful and actionable data without bias.
[Read the full article on Search Engine Land.]
Some opinions expressed in this article may be those of a guest author and not necessarily Marketing Land. Staff authors are listed here.