What are you running for tracking that isn't a monthly bill?

My hosted tracker just went from $89 to $129 a month and I'm paying it out of margin that is honestly not there right now.

What I need it to do: click tracking with sub IDs, a redirect that doesn't add half a second, postbacks to and from the networks, and split testing on landers. That's genuinely it. I don't use 80% of what I'm paying for and the campaign flow builder thing I have never once opened.

What's everyone using that isn't a monthly bill? Willing to put a weekend into setting something up, I'm not scared of a VPS. Just don't want to be the guy who saves $130 a month and spends every Sunday fixing it!

Self host. Is one time license or free depend which one you take, and put on 2gb vps for six dollar.

I run mine three year like this. Anyway your redirect get faster also because the box is yours and you choose the location — hosted tracker is in one datacenter and your traffic is not.

How many clicks a month? You've listed features and no volume, and volume is the only thing that decides this.

about 900k clicks last month across everything, peak day was maybe 60k

That's small. 60k a day is 0.7 requests a second average and maybe 15/s at peak, which a $6 box will do without noticing as long as you don't let the database eat itself.

The two things that actually break self-hosted trackers at your size, in order:

1. The click table grows forever and nobody sets up rotation. At 900k/mo you'll be at 30 million rows inside three years and every report starts timing out. Partition by month from day one and archive anything older than 90 days to cold storage.
2. Disk fills at 3am on a Saturday and the redirect starts 500ing, which means the traffic you're paying for goes nowhere and you find out from your spend, not from your tracker.

Second one is the whole ballgame and it's why people go back to hosted. Minimum viable version of not-being-that-guy:

# /etc/cron.d/tracker-watch — every 5 min
*/5 * * * * root /usr/local/bin/check-redirect.sh

# check-redirect.sh
code=$(curl -s -o /dev/null -w '%{http_code}' -m 4 https://t.example.com/health)
if [ "$code" != "200" ]; then
  echo "tracker $code $(date -u)" | mail -s 'TRACKER DOWN' [email protected]
fi
df -h / | awk 'NR==2 && int($5) > 85 { print "disk " $5 }' | grep . && \
  echo 'disk high' | mail -s 'TRACKER DISK' [email protected]

If you're not willing to write that much, pay the $129. Genuinely, no judgement — the fee is the price of somebody else caring at 3am.

I self-hosted for four years and I moved back last year, so let me argue the unpopular side.

The comparison everyone makes is $129 against $6 and that's not the comparison. It's $129 against $6 plus your time plus the risk. My honest accounting from the four years:

• Two incidents where I lost click data entirely. One was a disk, one was me upgrading PHP on a Friday because I am an idiot. Combined, maybe $900 of spend I couldn't attribute and two campaigns I had to restart.
• Roughly one evening a month on maintenance, averaged. Some months zero, one month four days after a botched migration.
• One incident where the box got flagged by a network because the IP had previous, which cost me a week of arguing.

Against that, $123 a month saved is about $1,476 a year. It was worth it in the years I was scaling and it stopped being worth it the year I had two clients depending on the numbers.

The honest rule I'd offer: self-host while your own money is the only money on the line. The moment somebody else is reading your reports, the failure has a second cost and the maths flips.

I have not tested any of the newer cheap hosted options in about eighteen months, so this may be dated.

four years and two incidents. that's an argument for self hosting mate, not against it

It's an argument about who the incidents happen to. Both of mine landed in a week where I was busy, which is not a coincidence — I upgraded PHP on a Friday because I was trying to clear the deck.

Ran both in parallel for two months in 2023 specifically to settle this for myself. Same traffic mirrored into each.

	Hosted	Self-hosted
Monthly cost	$99	$11 (VPS + backups)
Redirect p50	118ms	41ms
Redirect p99	890ms	260ms
Clicks logged	1.94M	1.94M
Conversions recorded	8,410	8,447

The latency difference is real and larger than I expected. The 37 conversion gap is the interesting row — self-hosted caught 37 the hosted one didn't, all from one network whose postback the hosted tracker was rejecting on a malformed parameter it wouldn't let me configure around.

At $89/mo and 900k clicks the money isn't the argument. The argument is that you own the postback handling, and postback handling is where the money actually goes missing.

37 conversions is a month of profit for me

Caveat: I've never run more than about 3M clicks a month, so this is not a scale answer.

I do self-host, and the thing that made it survivable wasn't the monitoring, it was 1) putting the redirect behind a CDN so a dead origin serves a cached 302 instead of nothing, and 2) accepting that reporting is allowed to be down. Those are two different services and people treat them as one. Reporting can be broken for six hours and cost me nothing. The redirect being broken for six minutes costs me real money.

Split them in your head and the maintenance anxiety mostly goes away. You are not running a tracker, you are running a redirect that also has a database attached.

Best post in the thread and it's a footnote.

Did it! $7 box, took me most of Saturday and about two hours on Sunday morning to get the postbacks matching. Redirect is noticeably faster, I can see it in the lander load numbers.

Set up the cron thing from post 5 basically verbatim. Cancelled the hosted one this morning.

Update nobody wants but here it is. Box ran out of disk on the 9th, at like 2am, and the health check emailed me at an address I'd stopped reading because it was getting hammered by the disk warnings for two days before that.

Lost three days of click data. Traffic still converted because the redirect was actually fine, it was the logging that died, so I didn't lose money directly - I lost the ability to know which sub IDs were working during a test I'd been running for a week. Had to restart the test. Call it $400 of wasted spend.

Still self-hosted. Not going back. But hexforge was right and I was the guy who thought monitoring means setting it up, not reading it.

The alert going to a channel you've muted is the most common outage in the industry and nobody writes it up. Thanks for posting it.

Send the disk one somewhere it interrupts you and leave the health one on email.