Why is it so hard to calculate ROI in Google Analytics? | Hacker News Reader

Why is it so hard to calculate ROI in Google Analytics?

tendinc.com

33 pointsryanevans27 comments

This is disingenuous. You're using a strawman (can't get ROI in GA) backed by a conspiracy theory (it's to make you spend with AdWords!) without disclosing up front that you have a competing product.

Google Analytics is a powerful tool, as you say, and as a result it has a steep learning curve. With that said, calculating the ROI on a marketing effort is not as difficult as you make it seem.

Using goals and ecommerce tracking, you can see the (dollar) return from any source, cohort, or time period. All that's missing is the input (investment), which only you know. Divide your investment by the return and you have your ROI for any segment you choose.

The entire issue with last-click attribution is not really an issue. Not only can you change this setting, but also you can view "Multi-channel funnels" and see every touchpoint (channel) of a user, even if they didn't convert on the first visit.

From a quick glance at Tend, it looks like your value offering is showing the same multi-channel funnel that GA shows, just with a nicer interface. That's fine, a better skin can be a sufficient value prop, but it's dishonest to frame this as though you're providing information that GA does not.

"disclosing up front that you have a competing product"

You're aware this blog post is posted on the blog of said competing product, right?

Not sure how much more disclosure is needed.

I really appreciate the comment. If you think about ROI for each individual customer, it's really tough to calculate that in Google Analytics. Say you have a SaaS product and customer X sticks around for 2 months and customer Y spends sticks around for 2 years. Those are two very different returns and certainly not easy or obvious to track in GA.

As a consultant, I'm sure you know how much even very sophisticated people struggle with this. I'd guess less than 1% of Google Analytics users know there is a multi-channel funnel view, and even fewer can understand it.

That's fair...I should disclose that we have a competing product.

Thanks again for the comment.

> As a consultant, I'm sure you know how much even very sophisticated people struggle with this. I'd guess less than 1% of Google Analytics users know there is a multi-channel funnel view, and even fewer can understand it.

Yes. GA has lots of downsides, and the steep learning curve is one of them. That's what I mean when I say that having a better interface alone might be a sufficient value prop over GA. If that's your advantage over GA, why not focus on that instead? The strawman argument and AdWords conspiracy are off-putting to someone like me, who is on the lookout for services that can complement what GA is doing.

Google Analytics is not meant to track individual customers. Of course you're going to struggle getting a tool to do something it wasn't meant to do! This is a job for CRM, not clicksteam.

I'm reading this as less of a gripe in GA not having a ROI report and more of complaining about marketing attribution. Marketing attribution isn't a super-easy problem to solve.

But complaining that GA doesn't know that a person saw a Facebook ad at some point before ultimately subscribing (and interacting with more ads along the way) is a bit like complaining that your car isn't also a boat and plane.

If you want comprehensive analytics, you need to spend time developing a comprehensive measurement plan and then spend time doing the reporting. Expecting any tool to do precisely what you want, despite the tool neither claiming nor even desiring to do what you expect, isn't a plan.

Marketing attribution is the bane of most marketers' existence, if only because it's never good enough.

No matter which service you're using, you're always going to find funnel use-cases that the service can't track, can't calculate, etc. It's the double edged sword of using an open web.

Your best bet is to sit down, figure out what your channel KPIs are, and then try to do 85% attribution with your different activities. Going for 100%, with GA or any other tool, is just going to drive you mad.

It's not hard to calculate ROI in GA. The author's gripe is that there's no big ROI number on the dashboard. If you ran an ecommerce site with known prices for each item, and your only marketing channels were paid and organic search, I could see how this would be possible. But what if your site is for lead gen. How is GA supposed to know the value of a lead, or your company's closing rate? For many sites, having a dash widget for ROI would be misinformed. And that's why I presume most analytics platforms don't include it by default.

>>How is GA supposed to know the value of a lead

This is something you can assign to any value to a series of events, a single event, or any other combination in GA.

We had a notorious marketing manager who assigned a value of $3,000 every time someone landed on the contact page of our e-commerce page. He did this because you had to sign up and pass a credit score before actually becoming a customer. No matter that you didn't know if you actually got the email or if the customer actually signed up. But man, did it make the ROI numbers look good.

> We had a notorious marketing manager who assigned a value of $3,000 every time someone landed on the contact page of our e-commerce page. He did this because you had to sign up and pass a credit score before actually becoming a customer. No matter that you didn't know if you actually got the email or if the customer actually signed up. But man, did it make the ROI numbers look good.

Heh. When I open up clients' GA I see things like this often. For example, counting visits to a checkout page as a purchase conversion, whether or not the person completed the checkout, assigning dollar values to non-conversion events, etc. To be fair, these were not set up by marketers, and they readily admitted that they were not using GA properly. But to have a marketing manager do that... Ugh.

Right. You can assign a value to events like lead gen in analytics! You just have to tell it what the value of a lead is, and, as demonstrated by your marketing manager, that's not always easy to know precisely. I suspect this is why ROI on GA isn't more automatic.

But yes, if you have a lead value, ROI shouldn't be difficult. So I still don't see what the authors problem is.

Now that this thread is full of advertisement, I will add my:

Microanalytics[1], a web-analytics service you host yourself with a CouchDB only (no backend). It supports custom events (with values) and stores raw data, that you can process easily later to see cool patterns. Like what the Tend homepage shows, you can also see each session, what each user did[3]. There's a command line client[2] that you can use to explore the data, it exposes the data for you in a way that pipes well.

Suggestions and criticisms wanted.

[1]: https://github.com/fiatjaf/microanalytics

[2]: https://github.com/fiatjaf/microanalytics-cli

[3]: For example: zhgj3ytm came from some ad, then visited some page; the other day it came again from Google and registered with the email bananas@email.com, then it left and never came back.

GA is a fantastic tool.

Problem is (not really a problem, but a fact of life) most of today's interactions that are pertinent to your product's success (and the ROI) are taking place outside of your properties that are trackable by GA (e.g. website).

If you look at apps like Meerkat, their meteoric rise can be attributed almost entirely to a handful of influential VC/celebrity figures, who loved the app, started using it, and actively promoting it. Product Hunt may have played a role in the beginning (those referrers would show up in GA of course).

GA is great at tweaking your conversion rates through the funnel, but as far as magic is concerned (what makes something explode), GA is helpless.

Full disclosure: We have a competing product to GA.

"The credit of acquiring that customer, and the cost to acquire that customer, really should be spread among all the different marketing activities."

Wow. I have been banging my head against this problem on the ad-tech-product-side for months. But this puts it really clearly. Other than having omniscient view of all marketing activities, any thoughts on even imprecisely measuring this.

Google Analytics has a built-in "multi-channel funnel" report that shows every touchpoint of a user, from their first visit to their conversion.

It also shows "assisted conversions" for each channel. An assisted conversion means at one point the user arrived at the site through that channel, and they did not convert on that visit but they did convert on a later visit from a different channel. It looks something like this:

Social Media (first visit) > Organic Search > Direct > Direct (conversion)

So what would otherwise be a non-attributable conversion ("direct" could mean several things, none of them useful) can now be attributed partially to Organic Search and Social Media, in this example. How much of the conversion you attribute to those channels is up to you.

If you or anyone else could use some help in evaluating their marketing work using funnel tracking, my contact info is in profile.

I guess I was more interested in channels that aren't visible to google analytics (i.e. anything other than PPC?)

how about banner impressions that don't result in a click and aren't run by Google? What about views on guest posts or partner sites?

If your entire marketing strategy is SEO with clickthrough + PPC + on-site content, then, yeah, seems like GA should solve the problem. But if not? Sample size of me: lots of people do marketing activities outside the Google ecosystem.

For that type of analysis, I believe you'd need a platform that ties all these data sources together.

We just use Excel and Tableau to get it all into 1 view. Don't really have the budget for other tools.

I think Google Analytics is misleading on purpose.

I worked with one small business owner who thought that all Google Analytics website traffic = all Adwords traffic, causing him to overspend on AdWords. Multiply by 1M+ clueless small businesses advertising on Google, and that's Google's profit.

GA and other 3rd party tools work well for small business. As companies grow they should invest in moving clickstream analysis in house. The lack of visibility into bot detection makes GA and other 3rd party tools sub optimal solutions.

As "web stats" - sure. GA is a great alternative to parsing your own server logs, which doesn't make sense to SMB.

But ROI? GA will tell you absolutely nothing about why nobody cares about your product or service, or why your competitor X is succeeding while you are not.

You will get what you pay for, simple click statistics, but don't expect any business insights on ROI.

It's funny because this conversation happens again, and again, and again, and again, with every new business as they grow big enough to start caring. And somehow, multiple attribution (let's not even discuss cross-device attribution, offline marketing, or more advanced ML-based customer identification and fingerprinting) is presented yet again as a freshly discovered Holy Grail. The first time I heard about it was when Rocket Internet told me it was invented at Zalando.

Google Analytics is two things:

- a tracker. That is, people do things on your site and what they do gets sent to a server.

- a data warehouse [1]. This is the part that gets people. Google stores the data, and what you see is the processed stuff that comes out of it after it's been sliced, by Google.

- I said two, but a subset of the data warehouse is the data mart that Google offers on top, which is the slick Google Analytics UI in orange with your gmail login. That thing is so easy and intuitive to use that people take a very long time to switch and have very high expectations when they do so (and usually, deeply entrenched manual processes that they try to replicate with the new solution, causing countless wasted hours).

The tracking part is pretty much sorted out with all competitors in the market.

For the small sum of $150,000 + around $10,000 for BigQuery per annum, you can access a few layers of your "raw" data with a few days lag. This includes the famed multiple attribution, because of course GA free pulls out only the last channel (plus or minus a few rules, such as ignoring direct traffic). The API will drive you nuts and the data is not in a relational format. I would say the sampling limit is the biggest drawback of using GA Free, particularly if you track things like product impressions on a search (which might add up to 100x product views). Once you have the raw data [2] in a decent relational database, it's very easy to get multiple attribution models of your choice, and to link them up to specific discount and customer rules, and do whatever else you want.

And of course, if you didn't bother thinking through your tracking codes, you won't get much out of Google. Google encourages you to think of the UTM parameters as "tags" as opposed to giving structure to your marketing tracking, and that results in some pretty messy schemas and workarounds afterwards (not their fault, but when you're the guy who has to clean up afterwards, you grind your teeth).

As far as I'm concerned, a much saner approach is to host the "data warehouse" part yourself - as part of your production backend, fed by the tracker directly (let your DBA figure out the specifics). Then, you get the data live, and you don't pay the Google tax. More importantly, you don't separate your customer, product and order information - which sits in your backend - from your web analytics information - which would sit in Google's DWH - requiring extraordinary ETL efforts. For the Google tax, you can hire a decent DBA and have enough spare change left for a very decent Postgres box on AWS RDS.

I'm sorry for the rant on your post, I know first hand how hard early stage sales are in SaaS (currently between contracts actually), but needed to get the last few years experience off my chest in the hope that it will save some pain to other people.

I also think you should make it clearer how you differ from GA. I'd love to know a. whether your tracker is any different b. whether your data warehouse, API, and so on are different c. whether your front end is different/more limited/more intuitive and of course d. whether your biggest USP is price.

[1] you can read the wiki page, but really, on a model, not implementation, level, it means "the single source of truth that contains all the data my company will ever have". You cannot have multiple DWHs running in parallel. As another HN poster once said, businesses are resilient to application errors, so correctness isn't strictly necessary; but it is a massive help and an enabler of profit generation and as such a very desirable goal. [2] https://support.google.com/analytics/answer/3437719?hl=en

Thank you kindly for for the detailed and thoughtful comment. We definitely need to do a lot better job explaining our USP. The biggest differentiator is that that you can start tracking conversions (and attribution) in 2 minutes in an intuitive way. We also allow you to tie individual customers to referrers, pages and specific ads.

Look, again, I don't want to impede your sales effort, but I basically ask "how is your Porsche different from my Ferrari" and you answered "well, it takes 2 minutes to intuitively start the engine, and also, it has wheels, seats, a gear stick and air conditioning".

I don't think you can set up tracking in two actual minutes with any of the competitors, nor that there is a significant difference with any of them - you still need to write the tracking pixel into your codebase, get it through QA, and so on. How is your implementation different? The tracking pixel is still code provided by Google/you/Webtrekk/Piwik/Heap/KissMetrics/etc. and it will still require putting into the codebase. An old team member recently admitted to using GA on his personal page because "it's so easy to roll out".

All competitors except IBM and Google Free let you tie requests, visitors, visits and transactions via their respective IDs which easily gives you all of the above if you know how to write basic SQL. All track the (single-r [1]) referer (ok, just nitpicking here), just as your server logs do, all track ads via the equivalent to UTM parameters or GCID, all track metadata such as the browser and device fingerprint which you can use for further identification, all track logins, add to mailing list requests, and whatever else you can dream of because at the end of the day it's just a request to a server and you can put whatever you want in it.

Attribution wise, it's a matter of having the code at hand. Having done this many times now, I can just rewrite the queries from memory once I put the data in the right format. Admittedly, it will take me half an hour rather than 2 minutes. But most clients are happy with that; it means you can run 40 parallel attribution models if you like, and include backend information such as discounts.

So once again: what is your USP?

[1] http://en.wikipedia.org/wiki/HTTP_referer#Etymology

I've been using Tend for about 4 months now.

In all seriousness, Google analytics, and my inability to use it, was one of the most frustrating things I dealt with as the founder of a SaaS product.

Google analytics regularly makes me feel like an idiot, and my frustration level, even with better products like heap and kiss metrics, had gotten me to the point where I was ready to give up on analytics entirely.

By using tend I learned that I was literally tossing away thousands of dollars every month and it's clearly shown me where we get our customers.

It's hands down the best analytics tool I've ever used.

This reads like an advertisement. What's unique about Tend?

Well there's two primary things: 1. It shows you ALL referrals that happened prior to a conversion. For example, if someone clicks on a retargeting ad, then a google ad, then our blog, then our docs, then a sign-up. I see that all together.

2. They have a conversion report that shows the weighted conversions, which allows me to see what impact each individual referrer has.

Reply on news.ycombinator.com