What is failure demand?
Failure demand is demand caused by a failure to do something, or to do something right, for the customer. In a contact centre it is every contact that only exists because something went wrong earlier — a wrong bill, a missed delivery, a broken app, a callback nobody made.
It is contrasted with value demand: contacts from customers who want something you are there to provide. A customer ringing to buy, to ask a genuine question, or to get advice that needs a human — that is the work. A customer ringing because their refund never arrived is not work you chose. It is work your organisation created somewhere else and shipped to your contact centre.
The term comes from John Seddon, the British occupational psychologist, whose work on service organisations in the 1990s applied the Toyota Production System’s distinction between value-creating and wasteful activity to service work rather than manufacturing. Seddon’s studies of service operations found failure demand routinely accounting for a substantial share of total contact volume — in some organisations, the majority of it.
The failure is almost never in the contact centre. It just arrives there.
Why it matters more than it sounds like it should
Conventional contact centre improvement makes failure demand cheaper to handle. Shorter calls. Better scripts. Deflection to self-service. Automation. All of it reduces the unit cost of a contact that should never have existed.
Eliminating failure demand means finding the upstream defect and fixing it, so the contact never happens. The distinction is not academic — it changes who has to act. Handling efficiency is a contact centre project. Elimination is a billing project, or a product project, or a logistics project. Which is precisely why it rarely happens: the department that pays the cost is not the department that can fix the cause.
The naming problem. “Failure demand” sounds like an accusation aimed at the contact centre. It isn’t. In most organisations the contact centre is the only department that has been dealing with the problem — and the only one with no evidence to prove where it came from.
What it looks like in practice
You already recognise most of it. It’s the contact that opens with:
- “I’ve already explained this to two people.”
- “I was told someone would call me back.”
- “It says delivered but nothing’s arrived.”
- “You’ve taken the money twice.”
- “I sent that document last week.”
- “Your website said it was in stock.”
Each of those has an owner, and the owner is somewhere else in the business. The first two belong to contact centre operations and routing. The third belongs to logistics or the carrier. The fourth to payments. The fifth to back office. The sixth to inventory or the web team.
The Failure Demand Taxonomy is our public, open catalogue of these recurring defects — what each one is, what it typically sounds like, and which function usually owns the fix.
Three kinds of demand
A taxonomy that only recognises failure demand can’t classify a whole contact stream. In practice you need three categories:
| Type | Definition | What to do about it |
|---|---|---|
| value | The customer wants something you exist to provide. | Nothing. Route it well and resolve it first time. This is the work. |
| failure | The contact exists because something went wrong earlier. | Find the cause, price it, give it to the department that owns it. |
| context | Driven by an external obligation or a life event — regulatory confirmations, a house move, a bereavement. | Can’t be eliminated. Can usually be made much less painful, and often moved channel. |
The third category matters because without it, teams either mislabel unavoidable contacts as failures — and chase eliminations that can’t happen — or quietly widen the definition of “value” until the failure number looks acceptable.
How to measure it
There are three approaches, in ascending order of usefulness:
- Ask agents to tag it. Cheap, immediate, and unreliable — tagging discipline degrades under queue pressure, and the categories are set by whoever designed the disposition list, usually years ago.
- Sample and code manually. Pull a few hundred contacts, read them, classify against a taxonomy. Accurate, genuinely revealing the first time, and too slow to repeat often enough to track whether a fix worked.
- Classify the whole stream automatically. Read every contact’s actual reason text, classify against a taxonomy, cluster into named defects, attach a cost. This is what AURA Source does.
Whichever you use, the output that changes behaviour is not a percentage. It’s a ranked list of named causes with a euro figure against each. A percentage starts an argument about methodology. A ranked, priced list starts a conversation about who fixes what first.
Pricing it
The arithmetic is deliberately simple, because a number people can check is more persuasive than a number they can’t:
avoidable contacts × average handle time × fully loaded cost per agent hour
Handle time should include after-call work. Cost per hour should be fully loaded — salary, employer costs, management overhead, technology, facilities — not base salary. Most organisations understate this by a wide margin.
This is a floor, not a ceiling. It excludes the second-order costs: the complaints a failure generates, the discounts and goodwill payments, the churn, and the opportunity cost of the value demand your agents weren’t handling. Understating deliberately is the right call — the number survives scrutiny, and it’s already large enough to act on.
Try it with your own numbers →
What it is not
| Often confused with | The difference |
|---|---|
| Repeat contacts | Narrower. A repeat contact is the same customer coming back. Failure demand includes the first contact too — if the bill was wrong, that first call was already avoidable. It also captures the cross-customer pattern: two thousand different people contacting about one broken thing. |
| Root cause analysis | Same instinct, different scale. RCA is usually done by hand on individual escalations. This applies it to the entire contact stream at once, with a cost attached so causes can be ranked. |
| Quality management | QM asks how well the contact was handled. Failure demand asks whether it should have existed. |
| Deflection | Close to the opposite. Deflection moves an avoidable contact to a cheaper channel. The defect still exists, the customer is still inconvenienced, and you are still paying — just less per unit. |
| Contact reduction targets | A target says volume must fall. It doesn’t say which contacts, or who owns them. Without a named, priced cause list, contact reduction usually becomes making it harder to reach you. |
Where to start
- Pick one recognisable defect your team is sick of. Don’t start with a programme.
- Count it honestly for a fortnight, including repeat contacts on the same case.
- Price it with the formula above.
- Take it to the department that owns the cause — with the number, not the complaint.
- Measure the volume afterwards. This is the step everyone skips, and it’s the one that makes the second conversation easy.
One priced defect, fixed and measured, will do more for your credibility than a full analysis nobody acts on.