Analytics Suite
Knowing what people actually ask, how often, in which language, and where the assistant fails. Without it, every improvement is guesswork.
Knowing what people actually ask, how often, in which language, and where the assistant fails. Without it, every improvement is guesswork.
What it does
The analytics layer records what happened across conversations: volumes, topics, languages, resolution, escalation, and where things went wrong. It is what converts a deployment from something that runs into something that improves.
Its most valuable output is usually the least expected: the list of what people actually contact about. Almost every organisation discovers that the real distribution differs substantially from what it assumed, and that discovery reshapes both the assistant and, frequently, the website.
Where it shows up in a conversation
Contact distribution
What people actually ask, ranked — usually different from what was assumed.
Failure analysis
Where the assistant could not help, which is the improvement backlog.
Language split
How much contact arrives in each language, and whether service quality differs.
Escalation patterns
What triggers handover and whether those escalations were necessary.
Resolution quality
Repeat contact after a resolved conversation, which exposes false resolutions.
Operational signals
Contact spikes that indicate something has gone wrong elsewhere.
Why it is harder in two languages
The measurement most organisations never instrument is service equity between languages. Reporting a single containment or satisfaction figure averages Arabic and English together, which means a materially worse Arabic experience can persist indefinitely without appearing anywhere.
Separating the two is a small change that surfaces a large problem. It also creates accountability: a visible gap between languages is something an organisation can act on, whereas an averaged figure invites the assumption that things are fine.
The same applies to failure analysis. Questions the assistant could not answer should be reviewed by language, because the Arabic failures and the English failures usually have different causes — coverage gaps in one, phrasing gaps in the other.
What to look for when you evaluate it
| Language-separated reporting | Whether Arabic and English are reported separately rather than averaged. |
|---|---|
| Repeat contact | Whether false resolutions are visible, which containment alone hides. |
| Failure visibility | Whether unanswered questions are collected as an improvement backlog. |
| Escalation review | Whether escalations can be assessed for necessity after the fact. |
| Privacy in analytics | Whether analysis runs on redacted text rather than raw personal data. |
| Export | Whether the data can leave for your own reporting, and in what form. |
Where it should not be trusted alone
Analytics should not be used to score individuals. Automated assessment of agent performance from conversation data is unreliable, and in most jurisdictions raises employment and fairness questions that outweigh the operational value.
It should not be reported with false precision. Sentiment, quality and satisfaction measures carry meaningful error, and presenting them to decimal places implies a certainty that will eventually embarrass whoever quoted it.
And it should not run on unredacted personal data where that can be avoided. Analysis rarely needs identity numbers and card numbers to be present, and removing them before analysis narrows exposure without reducing insight.
How this fits the whole
No single capability answers a question on its own. A conversation passes through several in sequence — understanding what was asked, finding the relevant material, composing a reply, checking it before it is sent — and the quality of the result is set by the weakest link rather than the strongest.
That has a practical consequence for evaluation. Vendors tend to demonstrate their strongest component, and a system with excellent retrieval and weak language handling will demo beautifully and disappoint in production, because the demo was conducted in the language and phrasing the system handles best.
It also explains why grounding matters more than raw capability. A more capable model that answers from its own training rather than from your material produces confident, fluent, unsupported answers — which is a worse outcome than a less capable one that retrieves correctly and declines when it cannot. Capability without grounding increases the cost of being wrong rather than reducing it.
The pieces are described separately here because that is how they are evaluated and configured. They are not experienced separately: a customer asking a question encounters one system, and judges it as one.
Where this sits in a conversation
- Something arrivesA typed message, a spoken sentence, an uploaded document. The first job is establishing what was actually received — which language, which script, whether it is legible, whether it is one question or several.
- It is interpretedWorking out what is being asked rather than what words were used. The same question arrives phrased twenty different ways, in two languages, often mixed, frequently misspelt, sometimes as a fragment.
- Material is retrievedThe relevant part of your own approved content is found. This is where answer quality is largely decided: a correct answer assembled from the wrong source is still wrong.
- A reply is composedIn the language of the question, in the configured tone, from the retrieved material and nothing else. Where the material is silent, the reply says so rather than filling the gap.
- It is checkedBefore sending, the reply is examined for claims the retrieved material does not support and for personal data that should not be repeated back. This runs on every reply.
- It is deliveredAs streamed text, as speech, or as a structured response — and the conversation is recorded so that what happened can be reviewed later.
Why this is built rather than bought in
Every capability described here could, in principle, be obtained by calling somebody else's service. That is how most assistants in this market are assembled, and it is why so many of them cannot answer the question that decides regulated procurements: where does the data go.
When a component is a call to an external provider, every question that passes through it leaves the organisation. That may be perfectly acceptable for a retailer answering questions about opening hours. It is usually not acceptable for a bank explaining a transaction, a hospital confirming an appointment, or a public body handling a citizen case — and the constraint is architectural, so it cannot be configured away later.
The second consequence is control over behaviour. A capability obtained as a service behaves the way the provider decides, changes when the provider changes it, and cannot be tuned for the specific way your customers speak or write. A capability that runs inside your own estate can be adjusted for dialect, for your terminology, and for the particular failure modes your material produces.
The third is continuity. External services are deprecated, repriced and withdrawn. A deployment whose core capability depends on somebody else's roadmap inherits that roadmap, including the parts nobody announced in advance.
Measuring this component honestly
| Measured in which language? | Nearly every published benchmark in this field was developed against English. An overall figure that averages across languages hides exactly the gap that matters in this market. |
|---|---|
| Measured on what input? | Clean or realistic. Read speech or spontaneous speech. Typed text or phone-typed text with autocorrect. The difference between these is usually larger than the difference between vendors. |
| Measured by whom? | Self-reported figures with an undescribed method are marketing. Ask for the test set and the procedure. |
| What counts as correct? | A surprising number of published figures count a partially correct answer as correct. Ask what the scoring rule was. |
| How does it fail? | A component that degrades gracefully is worth more than one with a higher headline figure that fails silently and confidently. |
| Does it report uncertainty? | A component that surfaces low confidence lets the system confirm rather than guess. One that hides it produces confident errors downstream. |
What to ask about this specifically
Ask for the Arabic figure
Not the overall figure. Nearly every published benchmark in this field was developed against English, and the gap between the two languages is where deployments disappoint.
Ask how it was measured
Against what test set, by whom, on what kind of input. A number without a described method is marketing rather than measurement.
Ask what it does when uncertain
Whether uncertainty is surfaced or hidden. A component that hides low confidence produces confident errors downstream.
Ask about real conditions
Noisy audio, poor scans, dialect, informal phrasing, mixed languages. Laboratory conditions are not predictive of contact-centre reality.
Ask where processing happens
For any component touching personal data, this is a procurement question rather than a technical curiosity in regulated sectors.
Ask what is retained
What data this component keeps, for how long, and whether it is used to improve anything.
Common questions
The ranked list of what people actually ask. Almost every organisation finds the real distribution differs from what it assumed, and that reshapes both the assistant and the website.
They should be, and this is the measurement most deployments miss. An averaged figure lets a materially worse Arabic experience persist unnoticed.
Yes, and that list is the improvement backlog. It is more useful than any success metric.
Analysis runs on redacted text where possible. Identity and card numbers are rarely needed for insight and their presence only widens exposure.
Aggregate trends are defensible. Scoring individuals from automated conversation analysis is unreliable and raises fairness questions that usually outweigh the value.
Yes, for your own reporting. It is your operational data.