What Flight 401 teaches about customer early warning systems: six clear signals instead of colourful dashboards

One light bulb, three people in the cockpit, nobody flying

On 29 December 1972, Eastern Air Lines Flight 401 is on its way from New York to Miami. A Lockheed L-1011 TriStar, 176 people on board. On the approach, one indicator light stays dark: the one that shows the nose gear is down and locked. The crew does not know for sure whether they can land.

The aircraft enters a holding pattern at 2,000 feet. The captain looks for the fault in the indicator. The first officer is flying, but he is also busy with the problem. The flight engineer climbs down into the electronics bay below the cockpit to check the landing gear with his own eyes.

Then something happens that nobody notices. The autopilot stops holding altitude and now follows movements of the control column, probably because the captain leaned against it. The aircraft slowly descends. After about 70 seconds and 250 feet of lost altitude, a warning chime sounds. Its speaker is at the flight engineer's station. The station is empty.

Shortly before impact, the cockpit voice recorder captures a question from the captain: "Hey, what's happening here?" The aircraft crashes into the Everglades. 101 people die, 75 survive.

The indicator was dark because the bulb had burned out. The landing gear worked.

The investigation report by the US authority NTSB names as the probable cause that the crew failed to monitor the flight instruments during the last four minutes and did not notice the descent in time. Their preoccupation with the landing gear indicator, the report says, drew their attention away from the instruments.

The aircraft was intact. Three experienced people sat in the cockpit, and all three were dealing with a light bulb. The technology worked. What failed was how attention was distributed.

The pattern in your company: the dashboard as the light bulb

Replace the cockpit with your Monday meeting. On the screen: revenue by region, open quotes, open rates of the last mailing, support cases, satisfaction scores. Plus traffic lights in green, amber and red. The group discusses why one metric is amber and whether the figure is even right. After an hour the meeting is over. Little has been decided.

Meanwhile, a long-standing customer is ordering less often. Their contact person has left the company, and the successor does not call back. None of this shows up on any slide. The customer does not leave with a bang. They go quieter until they are gone.

The signal exists somewhere in the data. There is just nobody sitting where it arrives.

How to recognise the pattern

In mid-sized companies, we often see the same symptoms:

• Several tools, no connection. Customer data sits in the CRM, invoices in accounting, complaints in the ticketing system or in the sales team's inbox. The full picture of a customer only emerges when someone pulls the pieces together by hand.
• No BI team. Reports are built by whoever has time, often someone from sales or finance alongside their day-to-day work. Whatever is easy to export ends up in the dashboard. Whatever is hard to get is missing.
• Metrics without an owner. A traffic light turns red, and the first question is: who is responsible? If that is unclear, the figure gets commented on instead of worked on.
• Amber as the normal state. Some fields have been amber for months. Nobody is alarmed any more.
• The surprise comes with the cancellation. In hindsight, the signs are almost always there: fewer orders, unanswered queries, an open complaint. They just were not where people were looking.

The dashboard is no more to blame for this than the landing gear indicator. What is missing is a decision about which signals count, and a person who looks when they go off.

One question is enough as a test: which customer will be gone in three months, and how would you recognise that today? If your dashboard gives no answer to that, it shows a lot. Just not what you need.

Why more metrics do not bring more certainty

After a surprise, one reaction is obvious: another metric. A customer cancels unexpectedly, so a new field goes into the dashboard. A few years later there are dozens of fields, and the surprises keep coming.

The cause rarely lies in the data. It lies in what happens when a value crosses its threshold. A metric without a defined action leads to a discussion, and no decision follows. Is the drop real or seasonal? Is it the customer or the data maintenance? By the time that is settled, the next week is due and with it the next traffic light. Amber becomes the permanent colour, and red gets negotiated over instead of acted on.

Medicine knows this effect as alarm fatigue. The Joint Commission, which accredits hospitals in the US, estimated in a safety alert from April 2013 that 85 to 99% of alarm signals in hospitals require no clinical intervention. The consequence: staff become desensitised, turn alarms down, switch them off or widen the limits beyond what is safe. More signals bring less attention. A customer dashboard is not an intensive care unit, but the mechanism is the same.

Flight 401 shows the second half of the problem: there was no person whose only job was to fly the aircraft.

Aviation learned from this. Flight 401 is often counted among the accidents that gave rise to Crew Resource Management, a training programme for cooperation and role allocation in the cockpit (it has nothing to do with CRM software). A NASA workshop in 1979 is regarded as the formal starting point. The principle behind it is simple: one person flies, another deals with the fault.

For your company, this means: whoever is responsible for a customer relationship needs a few signals where it is clear what to do. The question of why a metric fluctuates is legitimate work. It just belongs with a different person or in a different meeting.

The test for every metric: what do we do on Monday when it goes off?

Many customer scorecards work like the speaker at the empty station. The figures are collected, change colour and end up in the monthly report. What remains open is who does what at which value. Why every activity figure needs an outcome figure beside it is explained in activity and outcome as a pair of figures.

That is why we check every metric with one question: What do we do on Monday when it goes off? If there is no concrete answer, the metric belongs in reporting and not in the early warning system. Reporting is useful. It has no place on the scorecard you use to keep customers.

Three questions to check for yourself

Go through your scorecard line by line. A metric only stays if you can answer all three questions with yes.

1. Is there a defined threshold? "Dropped significantly" is not enough. You need a figure or a clear condition, such as "no order for X weeks". A starting value from your own experience is better than none. You can refine it later.
2. Is there a named owner? One person or a clearly filled role. "The team" or "sales" does not count. Responsibility that hangs on an empty chair is no responsibility at all.
3. Is there a concrete action with a deadline? "Call the customer within three working days" is an action. "We'll keep an eye on it" is not.

If a metric fails one question, add what is missing, or the metric moves to the monthly report. Either is fine. A warning signal without a response is not.

Six early warning signals with a clear action behind them

The following six signals are a suggestion, and the thresholds are placeholders for your own values. More important than the exact figure: each signal has a person, a deadline and a first step.

1. Usage or order frequency drops

What you observe: a customer orders less often, in smaller quantities, or uses your product less. This rarely happens abruptly. The intervals gradually get longer, and nobody notices, because the customer is still ordering.

Threshold (example): the gap between two orders is longer than [your factor] times this customer's usual rhythm. Or the volume over the last [period] is [your percentage] below the same period last year.

Who, when, what: the responsible account manager calls within a set deadline, for example within one working week. This is an open question rather than a sales pitch: has anything changed on your side? The result goes into the CRM afterwards. How to work out the usual rhythm for each customer is shown in churn early warning with order rhythm and contact clock.

2. The contact person changes or disappears

What you observe: your contact leaves the company, moves to another department or no longer appears in communication. With them often goes the knowledge of why you were chosen in the first place.

Threshold (example): any change of the main contact person. Or the main contact has not taken part in any exchange for [period].

Who, when, what: within about two weeks, account management finds the new contact person and arranges an introductory meeting. The aim: to make the case again for what the collaboration is worth to this person. For important customers, management comes along to the first meeting.

3. A build-up of unresolved support cases

What you observe: a customer reports several problems in a short time, and some remain open. A single ticket is normal. Several open cases show that the customer is losing patience, even if they don't say so yet.

Threshold (example): more than [number] open cases at the same time. Or one case unresolved for longer than [period].

Who, when, what: the head of support takes over the case personally, for example within two working days, and gets back to the customer with a plan for a solution and a date. Account management is informed so it does not walk into the next conversation unaware.

4. Payment behaviour changes

What you observe: a customer who used to pay on time now pays later, short-pays invoices or questions individual items. This can be down to liquidity. It can also mean that someone internally is checking whether they still need you.

Threshold (example): payment received more than [number of days] after the due date from a customer who has paid on time so far. Or [number] invoice queries in [period].

Who, when, what: accounting reports the anomaly to account management as well as to dunning. Account management clarifies in conversation whether it is a payment issue or a satisfaction issue. The reminder letter alone does not answer that question.

5. The customer no longer responds to meetings or queries

What you observe: emails go unanswered, meetings are postponed or cancelled. Silence is one of the clearest signals, because it is easily mistaken for being busy.

Threshold (example): [number] contact attempts in a row without a reply. Or [number] cancelled or postponed meetings in [period].

Who, when, what: account management switches channel, from email to phone, and tries a second person at the customer. If that is still unsuccessful after about two weeks, the customer goes on the agenda of the weekly meeting. There, it is decided whether your management approaches the customer's management directly.

6. The contract end approaches without a conversation about renewal

What you observe: the contract is about to expire, and nobody has spoken to the customer about continuing. If you only call shortly before the notice period, you often end up negotiating over a switch that was decided long ago.

Threshold (example): [number of months] before the notice period ends, no renewal conversation is scheduled.

Who, when, what: account management arranges a conversation about the coming year. Content: what has the customer achieved, what is missing for them, what is changing on their side? The renewal should come out of this conversation instead of being its topic.

What this selection achieves

For none of these signals does the work end with spotting it. Each one leads to a call, a meeting or a handover to a person.

The selection is a working hypothesis. A vendor of customer success software similarly advises focusing on four to six signals that are linked to success or risk. Test the hypothesis on your own customers: which of these signals were visible beforehand with the customers you lost most recently?

From signal to routine: who looks, and when?

The usual course of events: early warning signals are defined in a workshop and built into a dashboard. After that, nobody looks at them regularly any more. An early warning system is never finished; it has to keep running. That takes three things.

A fixed, short meeting. Once a week, with a clear time limit, around half an hour. Only one question: which signals went off this week, and what happens now? A general discussion of metrics has no place there. If nothing went off, the meeting is over after five minutes.

A fixed owner for each signal. Sales might own the change of contact person, customer service the open support cases, accounting the payment behaviour. The exact assignment is secondary. It just has to be settled in advance so that nobody has to ask when it matters.

An escalation path. Define when a case drops out of the routine, for example when two signals go off for one customer at the same time or a deadline passes without a response. Then the case goes to management. This path is written down before the first case occurs.

You usually already have the data

Many decision-makers hesitate at this point because they fear a big data project. The sources have been in the company for a long time:

• CRM: contact history, contact persons, meetings, contract terms
• Accounting or ERP: order frequency, payment behaviour, reminders
• Ticketing system or service inbox: open cases, repeat cases, handling time

To start with, a saved report in the CRM, an export from accounting and a simple spreadsheet that puts the two side by side once a week are enough. It isn't elegant. But you quickly see which signals hold up and where the data quality falls short.

Start with the signals whose data is already clean today, name the owners and schedule the first weekly meeting. After a few weeks you will see which thresholds trigger too early or too late, and where a manual step is so tedious that automation pays off. Only then is it time for the conversation about integration and tools.

What you can leave out

The Monday test has an uncomfortable consequence. Some metrics that have been on your scorecard for years fail it. That does not make them wrong. It is just that nobody does anything differently on Monday when they move. These values belong in the monthly report.

Typical candidates for the move:

• Average satisfaction across all customers. The average shows the mood, but not which customer is about to slip away.
• Net Promoter Score (NPS). Useful as a trend over several quarters. As a weekly traffic light, it mainly produces debates about interpretation.
• Totals across all customers. Total tickets, newsletter open rates, portal logins across all accounts. Totals hide exactly the individual customer that matters.
• Total revenue and customer count. Indispensable for running the company, too late as an early warning. By the time these figures drop, the customers are already gone.
• Metrics that only one person understands. If the calculation has to be explained in the meeting, the figure does not belong on the scorecard.

It depends on how a figure is used. Falling logins at a single customer can be signal 1. The total of all logins is reporting. A few meaningful metrics reveal more than many put together at random.

Why is leaving things out so hard? Every metric was introduced at some point by someone with good reasons. Removing it feels like a risk. But the crew of Flight 401 was watching plenty. All three were watching the wrong thing at the same time.

Whoever decides what the team looks at every week also decides what it does not look at. That is a leadership decision. Write it down: which metric moves to the monthly report, who reads it there and when you will next review the selection. Then it is clear that the figure was moved on purpose.

Conclusion: fly the plane

Warning signals are rarely in short supply. What is missing is someone who is responsible when one goes off. Three steps:

1. Check your scorecard with the Monday test. Whatever fails it moves to the monthly report.
2. Reduce to a few signals. One signal with a fixed action helps more than ten that people only discuss.
3. Set owners and a weekly meeting. Each signal gets a person, a deadline and a deputy.

This is exactly where UNIT ONE works. We treat a health score as an operating system rather than another dashboard: signals, thresholds, ownership and routines that run in your existing systems and trigger an action on Monday. If you would like to go through your scorecard with us, the Reactivation Audit is the place to start.

Who in your company is flying the plane right now?

Frequently asked questions

What is a customer health score, and does a mid-sized company need one? A customer health score summarises whether the relationship with a customer is stable or whether something points to churn. You don't need a large customer success team for it. You do need a fixed action that follows from the score.

What is the maximum number of metrics a customer scorecard should contain? We start with six signals. What matters is that every metric passes the Monday test.

Which early warning signals point to churn? As a starting point, we suggest six: falling usage or order frequency, a change of contact person, a build-up of unresolved support cases, changed payment behaviour, a lack of response, and an approaching contract end without a renewal conversation. None of them alone proves that a customer is leaving.

How do I set thresholds if I have no comparison data? With your own data: look at customers who have left, check how their behaviour changed beforehand, and derive a first threshold from that. After a few months, adjust it depending on whether it went off too often or missed a departure.

Do I need new software or a BI team for an early warning system? Not to start with. Most signals are in the CRM, accounting and the ticketing system, and automation is worth discussing once the routine is in place.

Who should be responsible for responding to warning signals? Each signal needs exactly one named person with a deadline and an escalation path. Management is responsible for the early warning system as a whole.

Sources

• Wikipedia: Eastern Air Lines Flight 401
• NTSB: Aircraft Accident Report AAR-73-14 (Eastern Air Lines Flight 401)
• The Joint Commission: Sentinel Event Alert on medical device alarm safety (April 2013)
• Wikipedia: Crew Resource Management
• NASA Technical Reports Server: workshop report Resource Management on the Flightdeck (1979)
• Gainsight: Customer Health Scores