> There's always been this thought that I've had, that advertising and "paid" attention is really in no one's best interest. You're likely to get users who really aren't going to get any value out of your software, so your churn increases, and you've also just paid for that user that's churned. These things get harder to tease out since it's almost impossible to ask "show me all the churned users this period acquired from advertising channels." Maybe that's possible, but you'd have to do a lot wrangling together to get it all working.
Correct me if I’m wrong, but isn’t precisely that kind of analytics simply table stakes for any modern crm/marketing/customer intelligence suite in 2019? It seems like that is absolutely a solved problem.
Yeah, it's a pretty contrived example on my part. My sentiment here is that there's so many inputs to modeling behavior, and trying to find signal in the noise, that at this scale your time is likely spent better elsewhere. Unless you have the revenue stream to do it and do it well, then the effort can be a time sink
How many variables do you need to track, though? $X spend, Y signups at $Z each, AA% churn, $BB retained MRR for an estimated $CC CLV based on an $X/Y*(1-AA%) CAC - why does it need to be much more complicated than that when you don’t have millions of users?
(Seriously though, I’m asking, not poking fun. You probably know more about this stuff than I do, having actually done it. It seems really simple to figure, to me. What am I missing?)
It's more a matter of being messy rather than complicated. You're trying to track users who often visit your site multiple times before purchasing, coming from different sources, on different devices, and somehow tie that attribution to the purchase. None of the data inputs are consistent nor tied together. Analytics tells you differently than your ad platform, your payment processor tells you yet another thing, and your email marketing tool, and your CRM. (And if you want the serious tools for data monitoring and reporting have moved to focusing on enterprises...) You have to somehow factor in refunds, free trials, prorated billing, early cancellations. You have multiple ad campaigns running ad variations. Don't forget a/b testing.
Finally you see some data point that hints something might be working, but you know you have to account for all the other factors involved. Did I make any website edits that day? Did the ad network change their algorithm slightly? Was there a holiday affecting traffic? When did I insert that new ad again? Wait, I know I changed my ad bid at some point... Did I get an influx of traffic from another source? Was it just a fluke?
If you want good, real data, it's messy. And far from a solved problem.
Stripe honestly does a lot of this for you. You can hack all that together with what you've got, the problem is that there's some seasonality to business as well as other market effects that make it tough to determine _why_ something is trending one way or the other.
For instance: we had a week last year where we had a flood of cancellations at once and there was nothing I could attribute it to. Looking back it was just a coincidence, however it consumed a lot of my time (writing emails and looking at analytics) to figure out why instead of just pushing ahead.
I was likely over-reacting, but I have noticed there's a lot of people spending a lot of time doing analytics and doing research on trends in stead of just executing. And, especially early on, you should just be executing and not thinking too much about trends.
Correct me if I’m wrong, but isn’t precisely that kind of analytics simply table stakes for any modern crm/marketing/customer intelligence suite in 2019? It seems like that is absolutely a solved problem.