Share article
Try working out, in your own company, how long it takes for a customer’s problem to reach someone who can actually do something about it. Not someone who logs it in the system. Someone with the authority to pick up the phone or change a process.
I’ve asked plenty of people in the industry this question, and almost none of them could answer straight away. When they did stop to think about it, the number usually caught them off guard, because it was higher than they expected. It’s not that they aren’t measuring anything. They are, plenty: NPS (Net Promoter Score, how willing a customer is to recommend the company), CSAT, operational metrics, you name it. But there’s a gap between what gets measured and what gets done about it. And that’s exactly where customers quietly slip away, before we’ve even had the chance to notice.
Reporting, or managing?
This is a distinction that gets blurred in practice far too often, and yet it matters enormously: reporting and managing experience in real time aren’t the same discipline running at different speeds. They’re two different things, with two different purposes.
A report answers the question: how are we doing? It looks backwards, it’s aggregated, it’s statistically sound, and it’s useful for comparing quarters or defending a budget. Real-time management asks something else entirely: is something happening right now that needs action? It looks forwards, even if only by a few minutes, and its value lies not in the precision of the numbers but in whether the information reaches someone who can still do something about it.
Qualtrics, through its research arm the XM Institute, estimates that poor customer experiences cost companies worldwide roughly three trillion dollars a year. That figure isn’t the result of companies failing to measure things. It’s the result of the gap between collecting data and acting on it, a gap a monthly report will never close.
That gap also carries a cost in smaller, more concrete numbers, and one really struck me when I read it last year. A 2022 analysis by CustomerGauge found that companies which respond to a customer’s rating within 48 hours, telling them what was done about their feedback, have 12% higher retention than companies without such a process. In the next round of NPS measurement, these companies also had three times as many promoters as those that never closed the loop. Eaton, the electrical equipment manufacturer, used this approach to cut its feedback-to-response time from four months down to 48 hours.
Four months down to 48 hours. That’s not a process running a few percent faster, that’s an entirely different way of dealing with customers.
A similar pattern shows up outside the world of satisfaction too, for instance with sales enquiries. A well-known 2011 Harvard Business Review study (Oldroyd, McElheran, Elkington) looked at the response times of 2,241 American companies to online enquiries. The average response time came out at 42 hours, and 23% of companies didn’t respond at all. Companies that contacted the customer within an hour were nearly seven times more likely to have a meaningful conversation than those who waited just one hour longer, and over sixty times more likely than companies that took 24 hours or more. That study is fifteen years old, but the underlying mechanism hasn’t changed. If anything, the opposite: as we’ve grown used to instant replies from online shops and chatbots, our patience has only shrunk further.
And here’s the bitter irony we shouldn’t overlook: according to 2025 research from Gartner, only 16% of customers actually believe their feedback leads to real change. We’re collecting more and more data, yet trust that anything is actually being done with it keeps falling. That’s not a tools problem. That’s a problem of speed and visibility of response.
How to build alerts
When I first started setting up triggers, the rules meant to automatically flag a critical situation, I thought a simple threshold would be enough: if CSAT drops below 3, send an email to the manager. It didn’t take long before the manager made it very clear that this was a brilliant way to turn him against me within the first week.
A working alert never rests on a single metric. It rests on a combination of signals that make sense together. It’s a bit like a smoke detector: the simplest kind goes off at every burnt piece of toast, while a smarter one combines smoke, temperature and airflow before deciding it’s genuinely a fire. In CX, that means linking low satisfaction with a high effort score (CES, Customer Effort Score, how hard it was for the customer to get something resolved), together with customer value and history, in other words, whether this is their first complaint or their third this month.
In practice, it helps to distinguish three levels of urgency. Immediate escalation, which you respond to within minutes: typically a combination of a highly loyal or high-value customer showing a churn signal, such as mentioning a competitor or cancelling a service. A priority queue, dealt with within hours: a negative experience with no immediate risk of churn, but with the potential to spread further, such as a public complaint on social media. And a systemic signal, which belongs not to one person but to the whole team: a sudden spike in complaints about one specific feature after a website update, something the product team should step in on, not a frontline agent.
This distinction is probably the single most important point in this whole piece. Without it, everything lands in one queue with the same priority, and that’s exactly where the next problem begins.
Who needs which data
This is where a lot of customer experience programmes get needlessly tangled up. Not everyone in an organisation needs to see everything in real time. In fact, it would probably do them more harm than good.
A frontline agent needs data about the person they’re talking to right now, within seconds: their interaction history, order status, most recent rating. They don’t need a quarterly trend, they need context for this one conversation.
A team leader or CX manager needs an aggregated view within hours, by the next day at the latest: how many escalations have come in, where a problem is building up, which agent needs support. This is about managing a team, not a single contact.
A product or operations team needs signals about systemic problems within a matter of days, quickly enough to act before a small issue grows, but not in real time, because fixing a product isn’t a matter of minutes.
And leadership needs trends and strategic context over weeks and months. Here, interestingly, the traditional report comes back into its own, because for investment decisions, the stability of a signal matters more than its immediacy.
The basic rule: the speed of data should match the speed of the decision it supports. Giving a director a real-time dashboard full of individual complaints will just drown them in noise they have no use for. Giving a frontline agent a monthly report is equally useless, it tells them nothing at the moment they’ve got an angry customer on the line.
The risk of signal overload
And now to the part that fascinates me most as an analyst, because it shows that speed alone doesn’t win anything.
There’s a field that has been dealing with this exact problem for decades, and has excellent, well-documented data on it: healthcare. It’s called alarm fatigue. Systematic reviews published in academic journals (such as Biomedical Instrumentation & Technology or peer-reviewed studies on PubMed Central) repeatedly show that 80 to 99% of hospital monitor alarms are false or clinically insignificant. Nurses in intensive care units can face hundreds, even thousands, of alarms in a single shift, and when there are too many, the brain naturally defends itself by starting to ignore them, mute them, or respond to them more slowly. The literature calls this the “cry wolf” effect, after the fable of the boy who shouted “wolf”: after enough false alarms, we stop believing even the real one.
The same thing happens in customer experience, just with less dramatic consequences, but with exactly the same logic. Set your alerts too sensitively, and you flood the team with notifications, most of which are trivial or outdated. People get used to it. They stop responding immediately, because they’ve learned it’s probably nothing again. And then comes the one alert that genuinely signals a key customer about to leave, and it gets lost among dozens of previous false alarms.
What can be done about it? Less is more, fewer alerts with higher precision beat a flood of imprecise ones. Thresholds need regular review, setting them once and leaving them to run for a year is a sure way to make them stop working. It’s worth separating information from action, not every signal needs an instant notification, plenty of them can simply be logged on a dashboard someone checks when they have time. And finally, measure how well your alerts actually perform, not just how many you send out. How many led to real action, and how many arrived too late or turned out to be pointless?
The takeaway
Real-time CX isn’t about a beautiful dashboard blinking away in real time. It’s about recognising which signal deserves a minute of your attention, which deserves an hour, and which can happily wait a month, and having the courage to let most other things wait, without feeling guilty about it.
It took me quite a while to understand this myself. I was convinced that the more we see in real time, the better we manage the customer experience. But the data says something different: we manage it better when we know which minute deserves our attention and which one can simply pass us by. Speed without discernment is just another form of chaos, only a faster one.
