Attribution Revisted
Attribution Retribution
One of the most contentious areas here at Donkey Publishing is attribution. As a publishing company, our budgets, headcounts, and, more importantly, channel champions are tied to what channels bring in the most revenue, readership, and subscribers. It seems as if I cannot go a quarter or sometimes even a month without the email, social, direct mail, search, or web teams asking that we revisit the current attribution model in place, which is last touch. Each department feels as if the last touch (all the credit going towards the last touch/click) doesn’t accurately reflect the channel's efforts, and business review meetings are getting more and more heated around taking credit.
Donkey Publishing has used last-touch attribution for the last seven years due to the simplistic way of calculating who gets credit and management needing to prioritize change. In fact, management is resistant to change since digital is under three separate umbrellas within the company, and direct mail is a part of the sales organization’s budget. I have asked management to consider initiating a project around changing the model, but the response has always been, “Why change something that isn’t broken?” However, it is broken because when each channel presents its numbers to senior management, all done by my team members, there are slight variations and tweaks to show inequity to the other channels inside the SAME company. This sort of thing puts my team in an awkward position to defend our methodology and recommendations on increasing or decreasing areas such as budget or spending in future quarters.
The amount of in-fighting that takes place around quarterly and yearly budgets is hard to watch because of the lack of a modern attribution model. It's clear that each channel treats its area as a separate company in terms of budget, even though each channel is going after the same customer and subscriber. This underscores the urgent need for a more equitable and modern attribution model, which we should implement quickly.
Publishing Attribution Models
If a magic wand were waived and I was able to explore and possibly implement a new attribution model, there are a few that would be in my consideration set:
Time-Decay: This model gives credit to all of the touchpoints that contributed to a conversion (revenue in the form of a subscription or purchase) while considering the time that each touchpoint occurred. It would give the greatest weight to the touchpoint that happened closest to the time of the conversion. There are many pros to a time-decay model, but my concern is the complexity around actually tracking these conversions in the tracking and data systems we have in place. The other concern would be how much weight would be given and how much influence each channel would have in determining that number. For example, the direct mail team would argue that the time it takes to deploy a mail piece is exponentially longer than it would take the email team to send out an email, resulting in the direct mail wanting greater weight when it comes to attribution. However, one area that would benefit the entire organization is the need to direct spending to channels with the lowest costs and higher conversions based on those costs (ROI). Essentially, it would weed out direct mail costs, which we know is expensive, but we don’t have a handle on how that channel fits into the overall conversions and ROI. That said, the coordination/timing of campaigns amongst each channel is non-existent and would require a complete overhaul of the process to which all channels must agree.
Multi-Touch: While challenging to implement, this model emphasizes the need for collaboration. It considers every channel and touchpoint someone has interacted with before the “conversion.” It’s a barometer of the most influential touchpoints and gives some idea of how the channels collaborated to influence this conversion. The biggest problem I see with this is that the channels do not work together regarding campaigns. The email team has different promotions regularly, and the social and web teams run their own individual campaigns, often trying to one-up promotions against each other, almost as if to pit against each other. The DM team is always running 45-60 days behind or ahead of promotions and rarely interacts with the other teams around what they are doing, almost as if it's a secretive channel. This would be the most complex model to implement because it would require everyone to work together, and I suspect this will never happen unless senior management mandates it. However, the potential benefits of such collaboration are immense.
Linear: Linear is a form of multi-touch in that it gives equal credit to all of the channels and touchpoints that a person interacted with before a “conversion.” In my opinion, while as simple to calculate as last touch, it might be a challenge to get the approval of all business units involved as they all feel as if they contribute more to the bottom line than others. In addition, it does not solve the problem of finding out what channel is ultimately most effective, so we can allocate more of the budget and hyper-focus on that channel. This type of attribution might lead to worse things than making things better.
The issue with changing any attribution model is ensuring you have the tools in place to track what you want to track and for it to be sustained. If you don’t, what level of effort would it take internally across all departments to ensure you can bring in the data you need to make the model work?
Donkey Next Steps
Donkey Publishing has a few challenges for us to solve. First, we must convince management that a new model is worth pursuing and would be a significant leap forward in our evolution while also pushing accountability back to the departments. One of the things I keep searching for on the web is a good outline and process of the following steps that a publishing organization needs to adhere to to make this business case. Once we get management's buy-in, the channels will have no choice but to come to the table to discuss what some will have to give up to modernize the way we attribute revenue to each department. While the departments want a change because they feel as if their channel does more and has a more significant ROI than the others, no one will want to flinch for fear that if they speak up, it results in their channel being the ugly duckling, their budget will shrink and their department will cease to exist.
My goal for the next six months is to develop a business case for a new model and see what, if anything, needs to happen internally to make that decision a solid one. The current methodology is no longer sustainable, and my experience is that it will only worsen amongst the groups, leaving the analytics department in a compromising position.
***CORE insight from Andrew Kordek @emailnucleus***
You can have the most intelligent people running your email program or other channels. Still, if departments struggle to work together, no amount of talented people will ever move a program forward. Remember that most organizations in whatever channel are working towards a common goal, and 9/10 are working to get to the same subscriber/customer/prospect. Treating it as a siloed organization that competes internally is a bad situation.
This is an opportunity for you to step up and champion change. This isn’t about who is better but about improving yourself and your customer experience. Suck up your pride and take the lead and stop playing corporate bullshit games. However, come armed with data and a good business case because anyone can complain and take the lead for a change; those who have a solution are looked at as the force for this change. Even if you are wrong, you have at least started the conversation.
Now, here is a word about attribution as it relates to Donkey. They have a bigger problem to solve first before anybody even mentions a new model. Valarie is in a tough spot, wants to help, and needs that champion. However, I have seen many attribution models implemented throughout the years in publishing that work. A time decay is typically the best model, although a custom time decay is suggested. Let’s see how Valarie goes about this project next time she makes a post. I feel bad for her since she seems like she can do the legwork for the groups, but the teams are not playing nice in the sandbox.