The Flight Recorder. Any conversion tracking setup either holds this one or it does not, whoever built it.
The Flight Recorder
Your lead pipeline has no undo button. You will find that out on a Saturday.
The propertyNothing is lost to a bad Saturday. Every lead is stored before it's handed on, so an outage is a replay, not a hole.
What a bad Saturday actually costs
When this goes wrong it is quiet. The forms still submit and your team still gets the leads, so nobody raises anything. Meanwhile the handoff to the ad platforms has stopped, and that part has no customer to complain about it.
Then you find out, and you want to fix it by sending the missing conversions on. Whether you can comes down to one thing: did anything write those leads down on the way past. If nothing did, they are sitting in your CRM and the platforms have never heard of them, and there is no going back for them, because there is nothing to go back to.
That costs you twice: the conversions the platforms never counted, and then what they learned from the silence, which keeps quality poor for at least a week after the leads start flowing again. The second half of that bill belongs to Signal Width.
Why I built this
In April 2026 I was talking somebody through an architecture, and the piece I would not trade away was the storage. On the call I said the data has to be always there, and that it’s better that it is delayed than it isn’t there. Late is an inconvenience. Missing is permanent.
I have also made that argument and lost it. On an earlier build the layer that writes every lead down before handing it on was declined, and the leads went from the form straight out to the platform. A Saturday deploy took out the piece doing the sending. The ads ran all day and the leads came in all day. By Monday there was nothing to resend, because nothing had been kept.
Nobody was careless there and the setup was not sloppy. That is simply what a direct wire does. It works exactly as designed until one end of it stops answering, and after that each lead goes nowhere and leaves nothing behind.
What it looks like when it is on
On 26 February 2026, on an account we run, nothing reached the ad platform for 44 hours. The cause was ours. One of our own publishes broke the leg that did the sending, and it broke it silently.
Nothing was lost. Every lead had already been written down somewhere the failure could not reach, so once the fault was fixed we replayed the 63 events that had been missed. All 63 were accepted. Nothing was counted twice.
That is the guarantee, and it is worth being precise about how small it is. Things break, ours included. The promise is only that the conversions a break costs you are still somewhere you can send them on from afterwards.
The honest next question is how those 44 hours got to be 44. Being able to replay a bad Saturday is not the same as knowing on Saturday that it was bad. That one is Positive Control, which is how you find out on the day instead of on the Monday.
Why would tracking break if our stack is reliable?
Usually put to me as“our stack is reliable.”
Uptime is not really the issue. A lead crosses several hops on its way from a form to an ad platform, and each hop belongs to somebody: the site, the CRM, whatever sits between them, the platform's own endpoint, and, on the accounts we run, ours. Any one of them can stop answering while everything around it looks healthy. A bad hour is recoverable only if one of those hops wrote the lead down before handing it on, because that is the only place anyone can go back to afterwards.
Whether your leads are written down before they are handed on is something you can check on your own account without a tool. If you would rather we looked first, send us your busiest landing page on each traffic source you run and we will write up what your platforms cannot see.
Find what your platforms can't see