Prompt for This Post
My good friend Jim Gianoglia from LunaMetrics and I had a lively debate on this topic on Twitter following his post last week on how to track CLV in Google Analytics.
First, let me say that I highly respect Jim’s work. His posts are brilliant, as is his work in Google Tag Manager. I try to read everything he writes, and he has made me a better analyst over the years.
Also, this is not our first friendly debate. And I think that’s a good thing.
And finally, I’m not arrogant enough to assume I’m right on this topic. I will lay out my arguments as thoughtfully as possible and let you, the reader, decide which side of the fence you land on.
If you are a business owner, we found a great service recently where you can use a different address for your business so in a city like Manchester, see https://www.manchestervirtualoffices.co.uk/ for details.
Tracking Customer Lifetime Value in GA Is A Pipe Dream
Let’s get right down to it, wudduyah say?
1. Your business [probably] predates your custom dimension
Let’s say you started company in 2007. You worked hard out of your basement for a couple years, and over the years built up an army of loyal customers who love to purchase from you because they trust you. You are now a trusted business to your customers, you’ve managed to do that because you know how to monitor the online activity of your remote workers, it made you see their activity at work and you’ve improved their productivity greatly.
And now let’s say you read Jim’s post Feb 18th and set up your custom dimension or metric Feb 19th because you have an eager developer who was tugging at the opportunity to do something really cool in Google Tag Manager. (Hey, can you blame him? It’s a playground for developers and marketers alike!)
Your “Customer Lifetime Value” data will start trickling in Feb 19th. That’s less than a week ago. But you’ve been in business nine years. If you want some tips to track that customer information, check with Andy Defrancesco who will be happy to share some tips.
Granted, some of us may have really started living when we could finally post pics of our kids, cats, and dinner. But it’s still not our DOB.
So would you really add a widget to your reporting dashboard titled Customer Age?
I would hope not.
You might title it “Years on Facebook” — and that would be fair game. Well, it’s the same with CLV.
I’m really not trying to be pedantic here. (No, really. It comes naturally.)
But seriously, it’s already a struggle to get decision makers to take analytics data seriously. It’s more difficult to understand than more bottom-line key performance indicators (KPIs). So if I built out a dashboard that promised CLV but I had to explain that, well, Alexander Irvanovich’s CLV wasn’t actually $17 — that’s just the sum of how much he’s spent since last Friday — I’ve just sent a signal to my client (or boss if you have one of those) that the data on this dashboard is unreliable. And you never want to put yourself in that position if you value keeping things like contracts and jobs.
TL;DR: Unless you started your site/business the day you set up CLV in Google Analytics, it’s not CLV. (Spoiler alert: Even if you started tracking CLV in Google Analytics from Day 1 it’s not going to be reliable. Sorry.)
2. People can opt out of tracking
There are a number of browser plugins that allow users to opt out of all tracking. What then? Let’s say I spend $5,000 at Raymour and Flanigan on furniture, but I’ve opted out of tracking. I would still be in their customer database because I had to enter my customer information to make the purchase, but I would be a ghost as far as their analytics data is concerned.
Furthermore, what if I found out I could opt out of all tracking after already spending $3,000. If the analysts for R&F are relying on Google Analytics (or any cookie-based analytics platform) to tell them my story, they might assume I fell victim to a competitor’s charms and wiles. Imagine how off putting it might be for me if suddenly I received an email from R&F telling me that they missed me and asked me to fill out a survey telling them what went wrong.
3. There are much better sources for CLV data
You want to include CLV in your dashboard? Great! I want you too as well! But that’s what databases are for, not web analytics tools.
Most dashboard tools on the market right now connect with databases just as easily as they do API-based tools, like Google Analytics. Hook up that firehose and make it rain CLV data all over your dashboard. Learn how Robert K Bratt DLA Piper and other business experts have managed to use CLV to their advantage.
Google Analytics is a great tool to gather clickstream data, but a database tool it is not. It can’t even reliably track user-based data for most sites — a drum I will beat until there’s a more reliable methodology for tracking across multiple sessions and devices. Expecting it to track users over a period of years is beyond unrealistic.
To Thine Own Data Be True
Google Analytics is one of the most powerful tools available to marketers. I’m a stark-raving fan and groan when I have to work with other analytics tools because I truly believe, all in all, Google Analytics provides the most bang for your [imaginary] buck. But it’s very important to understand the limitations of its bells and whistles before you try to use it for the wrong job.
That’s why whether I’m talking about campaign tagging, channel groupings, multi-channel funnels, or dashboards, I’m always going to [at least try to] let you know the caveats and limitations of each of these amazing tools. It’s not to be a naysayer or disrespect, in any way, the work that the amazing team at Google is doing in constantly iterating on an already powerhouse of a tool. It’s so you don’t get caught with your pants down when someone calls you to account for the data you’re presenting. And rightly so. People downstream could be making weighty decisions based on your dashboards. And you don’t want to be party to a bad decision (or multiple bad decisions) because of not understanding the appropriate boundary lines you should operate within when presenting analytics data to senior management.
I’m surprised this friendly debate didn’t enflame our little industry – but it’s not the first time I’m flabbergasted by the lack of real conversation and debate. It seems everyone is chasing the next cool trick or drinking from the wisdom of the few. At Superweek recently I said too much consensus leads to a lack of creativity.
When I read Jim’s post I had a reaction of “oh… wait a minute!” but didn’t botter commenting. But it bogs me for the exact same reasons you mentioned:
– CLV is about “customers”, not about web users, visitors or sessions – so the only proper data source is where your real customer data reside…
– CLV can’t exist only in the vacuum that is the web, it has to include back-office data;
– never use your web analytics platform as a replacement for your CRM or sales back-office system;
– and of course, the LV part stands for “lifetime”… if you can’t embrace your customers’ life, don’t call it CLV… call it “TOV” – Total Online Value at best.
I agree that friendly, respectful debate is GOOD for the industry. I’m also a HUGE fan of capturing a visitors’ customer IDs if they log in to your site. That is huge and an absolute no brainer. That clickstream data provides valuable insights into your customers’ behavior, and you can marry your analytics data with your database data in visualizations that will delight key decision makers. But we can’t ever expect a cookie-based tool to do the jobs databases were intended to accomplish.
Hi Annie – I agree with your point. And would like to suggest that ‘EXPECTED LTV’ is a suitable and much more useable metric. This would need to be determined (by some kind of offline analysis) by the business but would give an indication of the true value at the point of sale/transaction.
What do you mean by the true value? Sound more like profit margin. But Expected LTV sounds more projection focused. So I guess I’m not following.
I meant a reflection of what you expect the true value of the sale to be. In your post you quoted the example of showing LTV to date after 1 week and that this wasn’t a very meaningfull metric. I was suggesting you could, in some circumstances, inject a figure that you expect the sale to be worth long term. But this figure would involve a bit of head scratching and analysis by the business – perhaps using CRM data.
One main reason for using analytics is to assess quality of traffic so you need metrics that help you see this ASAP so you can respond.
I have to agree with you. We just completed a relatively complex database for a client that let us import all of his customer data for the last 10 years from every source he has sold through (there were more than a few). Once we were able ti import and clean all the data we finally started getting some real customer data that you simply can’t get out of anaylitics. Lifetime value, which products produced the most repeat customers, which products produced the most customers for different products, and the surprising one for us, which product seemed to end the repeat customer process the most.
Google Analytics is great, and it is a great tool for getting a general idea on things, but it has some serious limits to it’s usefulness.
Yes! Your case study illustrates beautifully the point I was trying to make. Database data can be stitched together (as you’ve been doing) using primary keys. But it’s mostly solid and not subject to fickle cookies, sampling, or any other challenges we deal with in GA.
Hi Annie!
I thoroughly enjoyed following your debate as it unfolded, as well as the discussion around it.
I’m too unambitious as an analyst to really take a stance on the interpretation of the KPI, but isn’t the whole concept of CLV kind of paradoxical in today’s measurement space? I mean, if you draw the line of abstraction at GA, saying it’s too inaccurate to shoulder the full weight of CLV (I agree), well what CAN you trust? Extending the example of Facebook vs. DOB, wouldn’t a customer’s journey with the brand start potentially long, long before they become a CRM entry? I’m talking about top-of-the-funnel stuff like exposure to billboard ads, anonymous visits at brick-and-mortar stores, reaction to brand campaigns, etc. things that are extremely difficult to measure and attribute, but things that can have a huge impact on which bucket you’d segment the customer in once they become a CRM entry?
In other words, when does CLV become a reliable metric? If you wanted to be pedantic, while explaining how GA numbers have all these caveats when talking about CLV, you’d need to outline all the caveats for calculations based on your CRM / other integrated systems, since the person you’re tracking / measuring has existed and very probably interacted with your brand even before becoming a customer. I think this is more in response to Stéphane’s entry above, since he talks about “real customer data”, and I think I’m just having a hard time figuring out what this is, since if the customer’s lifetime starts when they’re first introduced to your databases, you’re still cutting out a lot of stuff.
This is all in the spirit of nitpicking, by the way 🙂
Hey Simo!!
Great to see you here, and thanks for weighing in!
I’m very much a pragmatist about CLV and only include it in a dashboard if the company has is tracking revenue. If you start splitting hairs over what makes a customer valuable, you’ll drive yourself nuts and dilute the impact of your dashboard.
You can always use other KPIs in an investigative or forensic capacity — or even in a dashboard. Just don’t call it Customer Lifetime Value. Most executives are going to expect to see total revenue in that column. We just need to make sure it’s not so abstract that we look like we’re grasping at straws to prove our own value.
Hey Annie,
Disclosure: I didn’t read the Twitter thread. But I did read your article and Jim’s.
Jim made an excellent point about the two different ways to tracking CLV, as a dimension or as a metric. Annie, your first point speaks to the challenges with CLV as a metric (something Jim did note as well). But when CLV is dimension, I think it does an excellent job of describing users and is a great for segmentation. To see all sessions of all users who meet a CLV criterion can allow you to see the behavior of those users on your site. For subscription models, this can be quite powerful as well especially coupled with
So, limitations of GA aside (your point #3), there is a lot of usefulness of CLV in GA (especially as a dimension). In the same way that you may want to compare the behavior of users who have purchased 2 or more time, you’d also probably like to see the behavior of users who have spent more than $X,XXX.
WRT point #2, I don’t agree with the concern.. I haven’t seen our colleagues raise the concern of opt-out or javascript blocking as a reason to not trust GA’s story. GA’s story is different from BI tools or database mining, but it serves an important purpose. Furthermore, GA isn’t going to be your marketing automation software. Although it can be a good marketing platform as it relates to remarketing via custom dimensions. So while you aren’t going to send emails to people based upon GA data, you probably do want to create audience lists that leverage CLV data (again, as a dimension).
I’m not so enthralled with using CLV as a custom metric in GA; I don’t think that it has too much utility there. But as a custom dimension, I think that it has its place.
Yehoshua
Hey, Yehoshua!
Always a pleasure to see you here! The issue I have is recycling a term that’s well established and recognized, replete with scientific merit. If Jim or now Google (with their new CLV feature) called it something else that communicates it’s an aggregation of revenue data across multiple sessions, I wouldn’t have any issue with it. But it’s just not lifetime value. I’m really not trying to be dogmatic.
I’m really not trying to be pedantic on this one. I think this is where our industry would really benefit from some kind of peer-review process. We’re too loose with our handling of data and expect clients to understand our highly nuanced definitions that we come up with on the fly.
Another reason for not doing CLV calculation in Google Analytics is that Customer Returns can’t be captured in Google Analytics.
Thinking of an online shop like zappos: In case someone returns a shoe parcel because they don’t fit this should also be taken into account when analysing the CLV.
In a general sense, returns can be imported into GA, but I don’t think it would help with this issue. But, yeah, tracking CLV in GA is a bad idea on so many levels.