About five years ago someone connected with me on LinkedIn, started a sales pitch a couple of days later, was told no, acknowledged that, and then sent me another sales message five days later as though the previous conversation hadn’t happened.

The first response had been perfectly reasonable:

“No problems Andrew. Completely understand. Great to connect anyway and be part of your network.”

Five days later:

“Hi Andrew – thanks for the connection.”

And off we went again.

I wrote about it at the time because it seemed such a simple example of bad marketing. Someone had either forgotten the previous conversation, ignored it, or was working from a process that didn’t recognise what had already happened.

Five years later, I’m still seeing versions of the same thing. The difference is that the technology is considerably better at knowing who it is contacting and remembering what happened next.

In theory, that should mean fewer irrelevant approaches and fewer follow-ups after someone has already said no. I’m not convinced that’s what happens

I get loads of messages from people who say they’ve been looking at my work, followed by an offer which suggests they probably haven’t. Others pick up the name of a podcast, something from my profile or a phrase from a recent post, work it into the opening line, and then move straight into what is clearly a fairly generic sales message.

That isn’t really personalisation. It’s simply better insertion of personal data into the same process.

The follow-up sequences haven’t gone away either. If anything, automation has made it easier to continue them because there is less friction in sending the next message.

And this is where I think the problem has changed.

Five years ago, I could reasonably assume that somebody had failed to remember a conversation we’d had a few days earlier. In 2026, the systems are quite capable of remembering it.

They can retain previous interactions. They can use information about the person being contacted. They can adapt what they send next. So if somebody has already said no and the process carries on regardless, that isn’t simply a failure of memory. It tells us something about how the process has been designed and what it is being optimised to do.

The technology isn’t making those choices on its own. Somebody decides what counts as a useful signal, what happens after a rejection, how persistent the sequence should be and when it should stop.

I ended the original post with “Be better.”

Five years later, I think the question is slightly different.

We’ve built much better systems for knowing who we’re talking to. Have we actually built better marketing?