Why your frontline is inconsistent — and why training and scripts haven’t fixed it

If you run more than one location, you know the feeling. Same brand, same training, same playbook — and two of them produce completely different results. So it’s easy to conclude that some people are simply better with customers.

I ran that world for five years, across dozens of service counters. When I finally went looking for the mechanism underneath, I started where most people start: that the answer was perception — help people notice what they miss, and they will correct themselves. Eighteen months took that apart. This is what replaced it.

A conversation isn’t one thing. It’s a chain, and it runs across two people, not one.

Four links. Every one of them necessary. Not one of them sufficient.

That’s the part worth sitting with, because it’s what makes a conversation unlike most things you manage. A chain doesn’t average out. It doesn’t matter how strong three links are if the fourth gives way — the encounter is only ever as good as whichever link was weakest that day. Which is how a floor of good people, well trained, still produces results that scatter.

And then the harder half: it isn’t the same link twice. It changes between locations, between shifts, between two customers handled by the same person an hour apart. That wasn’t just my observation — it kept showing up wherever I looked. There is no one link that is the problem.

Which finally explained something that never made sense to me. Most interventions a business can buy for this are built around an assumption about where the problem lies. A script assumes a better move is needed. Training assumes a capability gap. Targets push willingness, but can’t supply a move he doesn’t have. Hiring bets on capability once, then leaves it alone. And whatever reaches you afterwards — the survey, the complaint, the mystery shopper — arrives when the conversation is over and the customer has gone.

So you buy something identical for every location and aim it at a problem you haven’t actually located, in a chain where the weak link keeps moving. Sometimes it lands on the link that broke. It works, and you roll it out. In the next location it lands on a link that was fine, and dies — and you conclude the problem is the people there.

Now a warning, because I walked into this one myself.

I assumed the beginning and the end of the conversation were the most fragile links. It was a reasonable hypothesis. It was also wrong.

The problem was that I was looking for a general weak link — something I could identify once and then design around. But the evidence kept pointing somewhere else: the links that break are not equally reachable from outside the conversation.

Whether your person knows the product, you can find out. Ask him. Whether a customer decided in the first thirty seconds not to say the real thing, or whether your person knew exactly what to do and didn’t: nothing about either reaches you from outside. Those failures leave little behind. So your attention goes to the links that answer back, and the break carries on wherever it likes.

I won’t hand you a fix in a paragraph. But there is a better question than “are my people any good?” — which has no useful answer at all.

It is: where is it breaking, and why?

That one can be answered. Not in a report, and not with one answer for the whole company — but inside the conversation, one encounter at a time. That is where the real work starts, and it changes where you’d spend the next rupee.