MELLORCATalk to Mellorca
Patterns

Data & Reporting

When Leadership Cannot See Bottlenecks Until Customers Complain

Waiting work makes no noise, so a stuck stage is usually found by the customer long before it appears on a management report.

Mellorca Patterns2026-10-10

A customer emails the managing director directly. Her quote has been "in review" for nine days and nobody has called her back. The director forwards the message to sales, who forward it to the estimator, who admits that a stack of quotes has been waiting for a pricing check since the person who normally does it went on leave. The pile was never hidden. It simply had no way of announcing itself.

That is the condition in its plainest form: the business learns where its slow points are from the people it is serving, not from its own systems.

The condition: the customer is the monitoring system

In a healthy operation, a queue that is getting longer or older triggers a response before anyone outside the business notices. In many small and mid-sized businesses, the first signal is an irritated phone call, a message to a senior person or a bad review. Leadership then reacts to the complaint, fixes that one case, and returns to normal until the next one.

Signs that this is happening include:

  • Complaints reveal delays that internal reports did not show.
  • Managers say "I only found out when the customer called".
  • Fixes are applied to individual customers' cases rather than to the stage that held them up.
  • The same stage, such as credit approval, technical review or dispatch, is behind many separate complaints.
  • Staff know that a pile exists but do not know whom to tell, or whether it is worth telling anyone.

The realisation: bottlenecks have an age, and nobody is looking at it

A bottleneck is not a mysterious thing. It is a point in a process where work arrives faster than it leaves, so items wait. What makes it dangerous is that waiting work looks like nothing. A job sitting in a queue produces no noise, no error and no entry on a revenue report. The only evidence is how long it has been sitting there, and few businesses track that.

The Kanban Guide offers a useful vocabulary here. It describes work in progress, throughput, cycle time and work item age: how long an unfinished item has been underway. Most small businesses track what was completed (throughput, in effect) and almost none track the age of what is still open. That is why a complaint feels like a surprise. The information that would have predicted it, the oldest open item at each stage, was never on anyone's screen.

Identification: a symptom-detection problem

This is an operational-visibility and early-warning problem. The visible symptom is customer complaints. The common assumption is that the team is not doing enough to keep customers informed, or that complaints are an unavoidable part of doing business. The structural cause is that the business does not measure what customers experience, such as waiting time at each stage, so it has no way to notice a deterioration until the customer reports it.

Site reliability engineering, a discipline developed for running large online services, expresses the same idea precisely. Google's SRE book says monitoring should answer what is broken and why, and distinguishes symptoms, which are what users experience, from causes, which are the underlying faults. It advises that alerts should focus on symptoms, because those are what matter to the user, and it warns that alerts which fire on vague suspicions or too often are ignored. You do not need to run a software platform to borrow the lesson. The question to ask of any process is: what would the customer feel first if this went wrong, and are we measuring that?

Diagnosis: why the pile stays invisible

An illustrative example, not a client case. A small manufacturing business takes custom orders. Each order moves from sales to a drawing check, then to purchasing, then to production. Each team keeps its own list. When purchasing falls behind, the drawing team sees its own queue is clear and sales sees orders "in progress". Nobody sees the total time from order to start of production, because that figure crosses three lists.

The mechanisms that produce this are usually structural:

  • Process: there are no agreed service times for the stages, so "late" has no definition.
  • Information: status is recorded as a label, such as "in progress", not as a date or an age.
  • Systems: each department sees its own queue, but the end-to-end journey is not assembled anywhere.
  • People: a busy team that has fallen behind tends to work harder rather than raise a flag, because raising a flag feels like admitting failure.

The commercial value at stake

The value here is customer experience, speed and resilience. Customers who are told early that something will be late are generally more forgiving than those who have to chase. Faster detection also shortens the time during which a stuck stage is quietly holding up revenue, because work that sits in a queue is work that has not yet been delivered or invoiced. Detecting the pile early also protects the senior people whose time is currently consumed by individual escalations.

We make no estimate of what a specific business loses to late discovery; that needs its own data. What can be said is that a complaint is an expensive way to find out, because by that point the customer has already spent their patience. Our related articles on why management finds out too late and when operational problems are discovered in the monthly meeting describe the same delay from the management side.

Possibility: let the process raise its own hand

If each stage had an expected time, and the system could compare the age of open items with it, a stuck queue would announce itself. Reporting tools already support this kind of logic. Microsoft's documentation on Power BI data alerts, for instance, explains that alerts can notify you when data on a dashboard tile crosses a threshold you set, and that they work on gauges, KPIs and cards. It also notes that alerts only work on refreshed data and only on numeric values, which is a helpful reminder that a signal is only as current as the data feeding it.

Alerts of that kind are not the point on their own. The Google SRE guidance is equally clear that alerts should be urgent, actionable and need a human to think, and that less urgent matters belong on a dashboard. A good early-warning design for a business is therefore small: a handful of age-based measures with sensible thresholds, each owned by a person who knows what to do when it trips.

Intervention: start with ages, not with software

  1. Map the journey from the customer's side. List the stages between "customer asks" and "customer is satisfied", and note where the customer is waiting.
  2. Agree an expected time for each stage. Rough numbers are fine at first. The point is to have something to compare against.
  3. Record dates, not just statuses. If every item carries the date it entered its current stage, its age can be calculated by anyone.
  4. Show the oldest open item at each stage. A single list, sorted by age, reviewed every morning, is often more effective than a full dashboard.
  5. Define the response. For each threshold, decide who is told, what they should do and when the customer is contacted.
  6. Review complaints as design feedback. Each complaint tells you which measure was missing.

Our article on what to measure at each stage of a business workflow goes into the measures in more detail.

Practical diagnostic questions

  • What was the last problem a customer told you about before your own team did?
  • For each stage of your main process, can you name the oldest item waiting right now?
  • Does anyone own the end-to-end time from order to delivery?
  • When a stage falls behind, who learns of it, and how?
  • If your three most-used reports disappeared tomorrow, would anyone still notice a delay before the customer did?

The bottom line

When leadership cannot see bottlenecks until customers complain, the business is relying on the people it serves to do its monitoring. The remedy is rarely a bigger report. It is a small number of customer-facing measures, especially the age of work in each stage, that make a slow stage visible while it is still a pile and not yet a complaint.

Sources and further reading

Related Mellorca reading

Article summaryWaiting work makes no noise, so a stuck stage is usually found by the customer long before it appears on a management report.
Explore our solutionsStart a conversation