OI Discussion: Email Attribution
Kath Pay: Welcome to the second of the new format weekly Influencer discussions!
Absolutely chuffed to be leading this discussion and I really hope we all get to share our trials and tribulations when it comes to both accurately measuring and attributing success for email. My thoughts are that by sharing what we've found works or doesn't work, we can all help each other.
To kick off this discussion and to provide a bit of food for thought, I've created a survey, which I'd love for you to complete.
I've also written some thoughts down here on the OI blog (if you've not already read it), identifying a couple of the challenges we face as email marketers. BUT, there are a lot more and I'd love for you to share these with us.
The more we discuss this issue, the more we can learn from each other and hopefully become more successful - both in attribution and in turn, in achieving additional budget.
So, please share your stories! They can be anecdotal horror stories or personal success stories -
Looking forward to it.
Tim Watson: Hello Kath,
This leaves me conflicted.
On one hand I see how much more revenue email delivers than last click attribution. I echo what you say. Just looking at search and I see email campaigns create organic traffic. Send a campaign and people search on the brand name to find the site. It happens all the time.
Running holdouts to understand the uplift of email not tracked by last click is one way to get an empirical correction. Work out the factor to increase email revenue and apply. A reasonable guide is for every $1 of email last click revenue you get another $0.50 not tracked to email.
But leaving alone the whole question of whether last click, first click or weighted attribution is best, here’s the conflict:
I don’t see it as a fight against other channels for a bigger part of budget. A huge amount of email list growth comes from first purchases and capturing prospect email addresses from site visits. Those initial visits come from organic, PPC, TV etc. Sure email nurtures prospects into customers but we need the email address to do our work.
You argue email should get more budget because of mis-attributed revenue. You need to look at the other side of the coin, by the same logic email should pay back the channels that delivered the email address. OK, so I’m not serious such inter-channel games are productive, but you see the point. As an aside I do track email signup sources to understand which sources generate email revenue.
Marketing channels need to work together without channel turf wars. I spend as much time in Google Analytics reports as I do in ESP reports. Looking at email revenue over time and trends of channels relative to each other.
Make the case for email not by debate about the attribution model, because that debate can’t be won, but by making some improvements to show what’s possible, gain credibility to get more budget. CFO’s aren’t excited about debates of which channel should claim revenue; it doesn’t change the size of the pie. Make the case for growing the pie.
Show improvement using an imperfect last click attribution model if that’s what you’ve got. Even on last click there should be strong revenue. Build trust in the channel and the ability of the team to deliver.
Kath Pay: Hi Tim,
Thanks for your thoughts! Sorry if I've not communicated my intentions well. You say:
"Make the case for email not by debate about the attribution model, because that debate can’t be won, but by making some improvements to show what’s possible, gain credibility to get more budget. CFO’s aren’t excited about debates of which channel should claim revenue; it doesn’t change the size of the pie. Make the case for growing the pie"
I totally agree. Business cases need to be built by showing what changes to your digital marketing program can do to increase revenue - regardless of the channel. The problem I've come across is that email is generally under-budgeted to begin with (I'm sure we will agree) - not in every case of course. - but generally speaking. However, it delivers the goods - far in excess of other channels - and yes, I fully understand that all the channels work together - that's why my consultancy is called "Holistic" & we focus on the customer journey & endeavour to break down channel silos :-) I've always said that email is not an island - it NEEDS the other channels to be successful. But, having said that, in order for us to be able to make changes, to build business cases, with the objective of increasing revenue, means that we need to have access to more budget.
The below chart from Forrester clearly explains our challenge.
Then, if we were to look at a chart by Custora, which tracks the orders for 2015/2016, we can see it's up there delivering the goods.
One question I'd love to have answered, as I've posed it many times over the years is this: Does the C-Suite simply struggle to believe that email can deliver more (i.e. increase the pie) as it's already delivering the goods with minimal budget? This of course, is where the business cases come in - to demonstrate that more revenue can be gained by more investment with a positive ROI.
Anecdotally from my clients and students that I work with, as well as from the surveys we're been conducting to deliver some valuable reports to the email community, I see that email as a channel is being held back from increasing the pie, not necessarily from lack of knowledge or expertise by the marketer, but by lack of budget and resources. This limits us to being able to make those changes to demonstrate success and thereby contribute additional revenue to the business (not just the channel).
Every time I teach a course on email, the same story is told by the majority of my students. They want to be able to get to the next level with their email program as they know it will 'grow the pie', however they're left juggling all their tasks, and end up just focusing on the BAU campaigns (as these are the emails that the C-Suite see) and feel guilty about not being able to improve their programs at all which will increase revenue. They know the potential of email, but they're not able to make it happen because of budgetary limitations and they struggle to make incremental changes and build business cases. They feel like it's a Catch-22 situation.
The first step in being able to let email do its job and be the awesome channel it is, is to ensure that measurement and attribution is correct. I don't look at it as being in competition as such with other channels, but rather, just ensuring measurement and attribution are correct, so email can be given the resources required to increase the size of the pie.
Personally, I too am torn - as whilst this is a challenge for we marketers, once we've made the breakthrough and have some convincing business cases to-hand, then it will be a win-win-win - for the marketer, for the brand and for the consumer.
Dela Quist: Hi Kath/All,
This seems to be a typical attribution lose lose situation for email marketing. Prove it they win – fail to prove it you lose.
The problem is it is not in many peoples interest to fix the attribution problem – I include many here on the list. For to do so would lead you in a direction that flies in the face of best practice. To send more email to large segments of your list. Something I call the halo effect and written about here http://bit.ly/2ySgKi2. I have also written about attribution on the OI blog previously providing everyone who cares to look with a simple way to calculate emails impact on other channels http://bit.ly/1KKUHiL
My advice is to run an analysis of revenue from all channels as a starting point. Then deduct the email specific revenue as you currently attribute it (opens and clicks, first or last touch) and break down this new number into days when emails to the majority of your list were and weren’t sent out. We have done this for a number of clients. The results consistently show the average daily revenue on days in which email was sent to more than 30% of the list was higher for the non-email channels too. We then looked at the source of the lift in revenue, where the lift was most marked based on last touch and first touch. Email had an impact on every other channel search, both natural and paid, affiliate programs, even social.
The other argument I find amusing is the one about cannibalization – unless you can show me incremental I am not interested. These the same folk that think abandoned shopping carts generate incremental revenue (see attached conversation from previous thread). Puhleeeze!!
Kevin Hillstrom had a very interesting take on this a while back his view is that E-Mail ROI Is Overstated Because Of Search http://bit.ly/2OVX8Sm BUT emails PROFITABILTY is underestimated for the same reason. Another good reason to stop using the 38x email ROI number
Tim let’s take your point about email should pay back the channels that delivered the email address – Email does. It’s called cost of acquisition and as far as I am concerned that’s a one off fee right? However once acquired each address keeps delivering these people back to the site and gets NOTHIN. Cost per click is something brands are happy to pay Google FB etc, but won’t attribute similar value to traffic driven by email. Kevin Hillstrom had a very interesting take on this 10 years ago http://bit.ly/2OVX8Sm
Allan Levy: Dela +1
Although rather than look at it as a lose lose i prefer to take the approach it’s a win win. A good email retention program leads to better LTV which in turn allows for higher acquisition payouts, which lead to more buyers which leads to more revenue generated by retention - essentially a virtuous cycle as long as there is continued improvement in the retention program.
Coming from someone who makes his living in retention you may find it interesting to hear me say that acquisition does not always get its fair share of attribution. here is a mini case study:
A client asked us to help with an acquisition program, we set up a landing page to acquire sign ups off of the click to the website from the program so every lead was pushed to the gateway. We established a solid welcome funnel and began evaluating leads by source. We also looked at 6 week conversion to sale of leads regardless of what drove the conversion . Here is what we learned.
Only 30% of the revenue from these new leads came in as last click attribution to the acquisition program
60% of the new customers came in attributed to an email sign up within a day of the sign up (so right there you see that 1/2 of the leads that signed up got attributed by Google to another source)
10% of the new customers (this is incremental to the 60%) came in from people who avoided the sign up process and purchased anyway
10 % of the new customers came in as a direct result of a welcome email or a follow up email in the welcome thread
20% of the new customers came in within 6 weeks of the program attributed to either other emails or other sources entirely
Taking 100% of the revenue into account the program earned within the clients allowable for acquiring new customers. If we take out any of the elements than we end up with too high an acquisition cost to continue the program.
We did of course continue the program and a few months in looked at sales from those who did not convert at all over the initial 6 week period and found another lift here as well of an incremental 10% - found money !!
The program was a success and lead to a 5% incremental lift in sales for the client counting just sales attributed in the first 6 weeks not the Life Time Sales of the new customers acquired.
- just to be fair once we had this all figured out we did come up with a conversion % on the landing page that allowed us to toggle our compensation to a given source up or down based on a projected downstream value.
Essentially what happened here is we looked at all new sales from a lead source. We built the target price of what we could pay for the lead based on LTV which included a Life Time Markeitng cost attributed to it.
If we had looked at this program in the narrow confines of single or even single + assisted attribution we would never have moved forward with the program.
With the industry focus on CDP’s recently we have been having discussions with our clients about throwing out the dated attribution model all together and looking at customer value and combined costs to get to it. We have not yet refined this so i can’t speak to what is or is not working but i think it’s going to be the wave of the future. We will most probably still look at a simple attribution model for directional information the same way as we looked at landing page conversions in this example or the way we all look at open and click rates as directional information to sales - they will just be a fraction of the total picture
One important PS - leads generated in the above program varied significantly by source and campaign within the source - if we did not monitor the value of the leads down to the micro levels the campaign would not have been successful.
Leading to another discussion point - Attribution needs to be defined to different groups separately - not all leads are created equal therefore not all attribution costs or results should be either
David Baker: ok, I'll bite.
I find this conversation is more a product of a company culture and whether goals are unified .. A myth people keep saying about attribution modeling is.. "it has to be so custom it's not durable over time".. this is just flawed thinking unless you are trying to boil the ocean. For every standard attribution model, there are benefits of using it and gaps. Some models don't take into account the value of Top of funnel advertising, some over weight the values. Some take into account channels based on time, last click, first interaction or linear views of all touch points. To say its futile is just non-sense. it is your goal as a marketer to try as hard as you can to optimize spend to drive a continual flow of prospects, customers, conversions and ongoing engagement. Attribution modeling is not just a budget allocation exercise its a way to find gaps in how you are attempting to understand your customer and their behaviors at a micro and macro level. to understand the impacts of your messaging, the impact of advertising mediums, platforms and ultimately how to lower the cost of touch, which all factor into Lifetime value views of your customers and programs.
For instance, in companies that are silo'd you typically have each channel (search, media, social, email, e-commerce) all using potentially different attribution models to justify their impact. Now, you can argue that they are all correct in their own interpretation, yet when combined with others shows a lack of continuity to unified goals. Email is no different, it gets a small share of the marketing budget and I don't think there is much argument on the value it plays in digital customer experiences. What I think we all lose sight of when looking at the inequities of budget allocation by the channel is the sheer reach of the channels in question. Media is about reaching and reaching new people in new contexts that aren't as direct as other channels. Email is nothing without some form on inbound investment. Social is again, a similar model to media in that it reaches millions is a platform, and has reached to a whole market, not just a subset of your customers that may or may NOT reflect the market's attitudes, beliefs and impressions of your brand. Search is intent based as is site search, so you can presume there has been some other ad impact prior to a search. Email gets a pittance of the marketing budget, primarily due to reach.
a potentially bad metaphor for this is to trying to find attribution for "being healthy" and how to weight a doctor's visit in this linear view? While they may prescribe a medicine or to change eating of health, who gets most credit for the impact? The doctor or the treatment or just you changing how you manage yourself? if surgical, I would think Doctor would get most of the attribution, whereby arthritis might look at this differently? There isn't a winner her cuz you are asking the wrong question, not what instance had the most impact on your health care, rather where should you spend your money to fix, sustain and or prevent illness? allocation and attribution may look far different, but in the end its about health and optimizing how much you spend to achieve it, right? maybe your tredmill gets more attribution than a doctor if you have high blood pressure? get my point? focus on the right questions that impact unified goals. Email is nothing without other channels and if you try to over weight the importance of an email channel, the question is not whether it will work by putting more money into the channel, but rather what "won't you do" to do this? will you sacrifice reaching 15 million people every month with a targeted ad message? or a campaign leveraging Twitter and Facebook for an upcoming event? its all about trade-offs and that in my interpretation is what marketing is all about.
Email simply won't help you capture eye balls and absent eye balls, you simply have a newsletter list. Yet in the day and age of Digital Disruption and Digital Transformation i think companies are looking at tech investments far differently now and attribution models impact in some ways (short-term spending), but the CIO is less enamored with channel specific attribution models for tech investments that are typically looked at on a 3-5 year plane vs. quarterly impact. Yet if you find a unified goal of lower customer attrition? that is something email has a huge impact on that advertising channels cant' touch? and might be worth a budget allocation decision to put more weight on fixing a leaky boat vs. a more powerful engine to propel it.
I think attribution models are vital and even silo'd ones have value in companies. any attempt to try to understand where to apply more funds in advertising and high impact marketing is a good thing and models help inform this. While NOT truely cause-effect. the constructs of running a last click attribution on google adwords or Facebook ads over Twitter Ads this is helpful. Yet the company culture is the catalyst to models becoming meaningful budget discussion points vs. models that drive every budget decision.
I'd encourage anyone wanting to put some time and energy into this for "email", that you focus on first learning how your channel brothers and sisters in your company use their models to answer their own department budget questions and separate the answers you are seeking into what you really want to accomplish? Do you want to get more budget for an upcoming campaign cuz you think it will convert better than $50-$100K will in another channel? or do you want to make a technology ask with this information to upgrade what you are doing? or do you simply want to use it so the email program is central to the discussion vs a downstream execution channel (which is what most people feel). As the old saying goes, "there's an app for that".. well, there is an "attribution model for that".
Since no one seems to be addressing education in this post vs. opinions, here is something you should read and do some research on if you don't understand these standard 11 attribution models in some form. these are great discussions to have with your department leaders and drivers.
- First Touch Attribution: The First Touch model gives 100% of the credit to the marketing effort that drove a visitor to your website for the first time.
- Lead Creation Touch The Lead Conversion touch model can often be confused with the ‘First Touch’ model. That’s because in a marketing analytics system built around lead generation, like a marketing automation platform, the website session where the lead was created is the first session where data is tracked and measured. In systems like this, if the true first touch was anonymous (read: the visitor didn’t fill out a lead form), it’s like it didn’t exist.
The benefit of the Lead Creation Touch model is that it helps you understand what marketing channels drive lead conversions.
- Last Touch The primary appeal of the Last Touch model is that it is the simplest model for attribution systems to measure.
By measuring and crediting the entire creation of a sales opportunity (the end of most B2B marketing funnels) based on the last touch, the analytics technology has the smallest time window for an error to occur. In long B2B buyer journeys, the time period from last touch to conversion is much shorter, than say, the first touch or the lead creation touch. This matters because many tracking cookies have a 30 or 90 day expiration. If the creation doesn’t happen within that window, the marketing channel data will be lost. By attributing 100% of the credit to the last touch, this expiration window becomes superfluous because no time lapses between the last touch and the conversion.
- Last Non-direct This is slightly more useful than the Last-Touch model, only because it eliminates the limitations of ‘Direct’ data.
When it comes to web and marketing analytics, ‘Direct’ data is a huge pain. Traffic attributed to Direct is typically defined by marketing analytics as any time a visitor manually enters your URL. But in reality, just about every marketing analytics product considers any visitor who doesn’t have a referral source as Direct. A common behavior that gets classified as Direct is traffic from untagged (or improperly tagged) social posts, social ads, or untagged emails. Rather than having its own filter, Direct becomes the catch-all bucket for traffic that doesn’t qualify for any of the other filters.
In short, Direct data is often misleading. The major upside of the Last Non-Direct Touch, then, is that you avoid the troubles of Direct channel data.
- Last <channel> Touch This is the catch-all bucket for channel-specific attribution models. Search marketers will want to use the Last AdWords Touch model. Social marketers who want to show their value will use Facebook’s default Last Facebook Touch model or Twitter’s default Last Twitter Touch model.
Note that by ‘last touch’ it means the last touch before whatever conversion you configured the analytics to measure. This could be the lead conversion, the opportunity conversion or whatever you set it up to be.
The upside is that these models usually come standard with their channel -- Facebook Insights uses a Last Facebook Touch model, AdWords Analytics uses a Last AdWords Touch Model and so on.
The negative is that each of these models are extremely biased to their own channel and overvalue their respective impact. If you use each of these attribution models separately, and then aggregate them into a single report, you’ll likely double-count or triple-count conversions. For example, if a visitor clicks on a Facebook ad on Monday and then an AdWords ad on Tuesday and then converts, both Facebook’s Last Facebook Touch model and AdWords’ Last AdWords Touch model will claim 100% of the conversion credit.
- Linear Attribution Linear is the simplest of the multi-touch attribution models. It distributes credit by evenly applying credit to every single touch in the buyer journey.
The positive is that it is a multi-touch model, so it gives credit to marketing channels throughout the multiple stages of the funnel.
The negative is that it doesn’t take into account the potential for varying impact of marketing touches. For example, if a prospect spends two days at one of your user conferences and then goes home and visits your site 19 times via Direct and then converts, your user conference will get 5% of the credit, even though it likely did most of the heavy lifting. Direct, on the other hand, will get 95% of the credit.
- Time Decay Attribution The Time Decay model is a multi-touch model that gives more credit to the touchpoints closest to the conversion. It makes the assumption that the closer to the conversion, the more influence it had on the conversion.
The problem with this assumption is that it will never give a fair amount of credit to top-of-the-funnel marketing efforts because that will always be the farthest away from the conversion.
- U shaped Attribution The U-Shaped model, which Google calls Position-Based, is a great multi-touch attribution model for marketing teams that focus on lead generation. It’s a multi-touch model that tracks every single touchpoint, but rather than give equal credit to all touchpoints like the Linear model, it emphasizes the importance of two key touchpoints: the anonymous first touch that got the visitor in the door and the lead conversion touch. These two touches get 40% credit each and the remaining touchpoints equally split the remaining 20%.
The downside to this model is that it doesn’t consider marketing efforts beyond lead conversion. This makes it an ideal model for lead reports or for marketing organizations that don’t do marketing targeted to prospects beyond the lead stage.
- W Shaped Attribution The W-Shaped model takes the concepts of the U-Shaped model to the opportunity stage, which for many organizations is the end of the marketing funnel. In addition to giving extra emphasis to the anonymous first touch and the lead conversion touch, the W-Shaped model also gives emphasis to the opportunity creation touch. These three key touchpoints receive 30% credit each, and the last 10% is split equally among the remaining touchpoints.
In spreading the credit with this distribution, the W-Shaped model highlights the three key funnel transitions that marketing impacts in the customer journey.
- Full Path (Z) Attribution Taking it even one step further, the Full-Path model (or Z-Shaped model) also accounts for marketing beyond the opportunity stage. Rather than two key stages represented in the U-Shaped model or the three key stages in the W-Shaped model, the Full-Path model adds a fourth key stage: the customer close.
In this model, each of the touchpoints at the four key stages receives 22.5% of the credit and the last 10% is split equally among the remaining touchpoints.
While it may seem like more key touchpoints is a better and more accurate representation of the customer journey, this model is really only appropriate for marketing organizations that do marketing to existing sales opportunities. Most organizations, unless there is extreme alignment between the sales and marketing teams, tend to let their AEs control the conversation and messaging when they are trying to close deals. Before you try to adopt this attribution model, be sure to sync with your sales team.
- Custom or Algorithmic Attribution : The last attribution model is to have a data scientist build a custom or algorithmic attribution model that best matches the customer journey specific to your buying process. By analyzing your existing customer data, you can see which marketing channels have an outsized impact or if there is a particular step in the customer journey that is important.
This is clearly the most difficult and time-consuming model to build, maintain, and use, but it will also give you the most accurate representation of your marketing’s impact on the customer journey
Credits for this breakdown go to Google bizable for a great general write up: https://www.bizible.com/blog/marketing-attribution-models-complete-list
My words of advice to anyone wanting to tackle this... focus on the questions you want to answer, ensure your answers help achieve unified goals and then socialize your thinking with others challenged with the same thing. This is easier to answer if you are a smaller team, as the larger you get the more politics involved with the answers. Some may think of this as a bad thing, but it also keeps you from being a lazy marketer and settling on something vs. continually picking at it.
Focus on what you want to do first, then play with different models, understand your gaps in data to accurately fill the model and while I hate to say do a customer model, in the end, I believe everyone will somewhat customize their own nuances to their models, but you have to start somewhere.