Why it can ‘take a few days’ to be unsubscribed from an email list (2019)

I managed a multi-million member mailing list in the late 90s. It would take 24+ hours to send a single mailing, despite us writing our own multi-threaded SMTP senders, per-destination optimized number of concurrent connections (too many and you get blocked, too few and it's too slow), highly managed hard/soft bounces to keep the list clean, etc

Aside from the "day to deliver", it was still pretty common for mail providers to have deep SMTP queues that were very far away from real time. They would very commonly hold mail in batches and deliver them a couple of times a day.

So, we would say "a couple of days to unsubscribe" to account for all that. I noticed other large email lists using the same terminology.

Then, the smaller lists would sort of "cargo cult" this messaging because the lower expectations made their life easier.

My guess is that everyone held onto that as email delivery tech improved to give themselves more leeway, or sneak in a few extra "last" emails, etc.

I feel like the only solution is a law that simply requires businesses to make it equally easy to do something as it is to undo it (especially for things like being able to “click here” to enroll but “on hold for 45 minutes” to cancel...give me a break). Some states are getting there.

A bit of a non sequitur, but this part reminds me of a place I worked where there was an old computer terminal set up in the middle of the receptionist’s office, which was tiny, and you had to kind of scootch around it to get in. It was like that for an unknown number of years.

One day before I started working there a couple of smart fellers asked what it was for, and nobody knew, so they investigated and found it wasn’t plugged into anything except the electrical outlet in the wall. Even so, it took them a while to gather the courage to unplug it, because who knows what crazy systems the people before you have set up.

Epilogue: They moved it to a stairwell where it sat for another couple of years looking like the saddest computer in the world before it was finally hauled away. [^1]

[^1]: https://raw.githubusercontent.com/aboxwithrocksinit/test-buc...

CAN-SPAM gives a 10-day window to honor the unsubscribe and if a marketing team can shovel more marketing your way for another ten days you'd better believe they will.

Here are the most common troubleshooting things that come up:

- Your email service may strip out the unique URL that unsubscribes you. In which case you may go to a blank unsubscribe page. If it's not well designed, you may think you are unsubscribing but really you submitted an empty form request.

- (We can't always do one-click unsubscribe anymore because some modern SPAM filters work by clicking every link in an email for you.)

- Similarly, the unsubscribe page can often be just a web form. Some blocking software like Ghostery may think it's a marketing form and straight up disable it.

- Some other team or department in the company set up their own email service for customers or leads without telling the marketing department and doesn't do anything to coordinate the opt-out process (This happens remarkably often.)

- The MOST COMMON reason unsubscribes don't work is some set of forwarding rules. Someone signed up on the website with the address for a mailing list, or a junk email that forwards it to their actual email address. We can't magically guess the right address to unsubscribe from (we had a case where 8 different people on a 15 person product team tried to unsubscribe the list, and one person on the team kept resubbing them every time).

Our particular team always setup multiple alarms to catch system outages but you never know how quickly the service provider will resolve their bugs.

Any way to lay down the hammer as a dual US / EEA (basically lots of EU legislation apply) citizen?

Or bullshit jobs.

well you had me until here. a perfect parable of tech debt and unwieldy bureaucratic quagmire... except it was racism the whole time. what a waste.

For context, in the late 90s, I started at "the first iteration of" Frontier Communications, not long after it transformed from Rochester Telephone and Telegraph[0], before its acquisition by both Global Crossing and Citizens (I went with the Global Crossing side).

We were a ~20,000, no 5,000, no 15,000, no 10,000, well, somewhere in there depending on what part of any given year over the 2 decades that I worked for the variety of companies that started, for me, there.

On one occasion, we had a process for collecting tariffs. That term isn't exactly right, but it describes it best. Basically, when a phone call is placed in the US, if it does not terminate with your carrier, your carrier is paying a fee to another carrier, or to a long-distance provider[1]. For call termination purposes, Global Crossing was almost always the long-distance provider, not the local provider[2] and we were paid an amount by one of the larger players that doesn't go unnoticed when it's missing from the balance sheet.

The finer details are in a previous comment of mine, but the end result was my discovery that the process for getting paid was tied to a decade old desktop PC sitting on a shelf in the main data-center with a USR modem attached to it which was recently moved from Green Bay, WI to Detroit, MI (and none of the original staff came along with it). It was sitting at a boot screen asking for hard disk parameters. I reported to a VP at the time who came to know me as the guy to go to for the strange stuff, and having spent my teens building PCs and inputting those parameters I recognized label with the odd acronyms/numbers on it as the parameters it was complaining about -- clearly the last guy noticed the dead battery and the label maker, not a replacement battery, was available.

Up until a few years before I left, the "trouble-ticketing" system used for all customer and IT support was purchased, source code and all from a bankrupt ISV[3]. It used a Java fat client that talked SOAP, I think, to the worlds most normalized, awful performing database I've ever encountered. I saw queries coming from this app where the UI indicated date ranges like "1/1/2003 - 1/1/2005" and the app would spit out a query with "BETWEEN" statements for each day, with nested parenthesis. It was clearly the worlds ugliest ORM, and trying to make sense out of the over 2,000 tables that made up the application was, barely possible.

To resolve this grief, our internal development team undertook writing a web UI for it. At a certain point, I want to say we had twice as many physical hosts operating our internal ticketing system than we did hosting everything we exposed externally, and not just "more boxes" but the most powerful hardware in the DC.

There seems to be a lot of sentiment, here, that if this were a "sign-up" process, it'd be far faster/better.

Anecdotally, my experience leads me to think this is more "laziness/enterprise inertia/large-company edge-cases" than it is "it's not bringing in money, so we'll do it the dumbest way imaginable". At least, I know for a period of time large pieces of the on-boarding had the appearance to the customer of being highly automated, but at the end of the day, it was a Clarify ticket passed between several hands who made phone calls/sent faxes. They'd certainly allocate larger staff to something that brings in revenue, but anything regulatory-related, at least where I was (and probably similar in all Telecom -- almost certainly more obnoxious in anything banking-related) was taken very seriously.

The fact that they have contract staff in India handling the process manually says two things to me: (1) somewhere, there's a fine/other clear risk if they don't and (2) someone decided this was the cheapest way to do it. I'd challenge whether that last part is actually true, but I've seen so many strange things where the reason turned out to be "there's some lawyer/some rule somewhere that says we have to also do this odd thing when we unsubscribe a user and writing a tool to access the system that does that one thing will cost us $150,000". Usually that "odd thing" is some form of tracking for auditing purposes and the system that would be used for tracking whatever that odd thing is ends up being either ancient/impossible to interface with or comes with some insane Enterprise-y license model that results in costs doubling whenever you want to do anything useful.

[0] Name escapes me.

[1] I have been out of Telecom since just before Level 3 became CenturyLink (now Lumen). This was the case when I left, it is probably still the case, today, but someone might have fixed that by now, too.

[2] This was GC days, still operating a few actual termination points, most of the business was IP or traditional long-distance; Technically, though, GC was a CLEC.

[3] My memory is foggy, I may have details incorrect here but the product was called Clarify. To make matters worse, there was at least three different vendors out of ... eight or so? ... that called their product "Clarify" or used that in its name. I'm not even sure if ours was just "Clarify" because by the time we were done with it, its UI was entirely internal and it had an (ugly) internal logo.

4-hour workweek and such books have taught Westerners to de-humanize Indians and give them the most menial jobs imaginable at next to nothing salaries