Travel and Expense Analytics: Why Finance Still Can't Answer 'What Did We Spend on Travel?'

The Question That Takes Three Days

A CFO asks a simple question in a Monday meeting: how much did we spend on travel last quarter, and how does that compare to the quarter before?

In most organisations this triggers a small project. Someone exports claims from the expense system. Someone else pulls the corporate card statements. A third person searches accounts payable for the travel agency invoices. The three extracts use different date logic, different currency treatments and different category names. A day and a half later, a number emerges with a caveat attached.

By Thursday there is an answer. By then the question has usually moved on.

This is not a tooling failure. Most of these teams already have a BI tool, and several have a dashboard nobody trusts. The failure is upstream of the dashboard, in how T&E data is structured — or rather, how it is not.

Why T&E Spend Is Structurally Invisible

Travel and expense is unusual among spend categories because it arrives through channels that have nothing to do with each other.

Channel 1: Employee reimbursement claims. Meals, taxis, incidentals, sometimes flights booked personally. Lives in the expense system. Categorised by the employee, often as free text.

Channel 2: Corporate cards. Hotels, airfare, client entertainment. Arrives as a bank statement feed with merchant codes but no business context — the statement knows a charge was made at a hotel, not which trip it belonged to or which project it should be coded against. We have written separately about reconciling corporate card statements across hundreds of cardholders.

Channel 3: Travel agency and supplier invoices. Booked centrally, invoiced monthly, processed through accounts payable as a vendor bill. From the ledger’s perspective this is a supplier payment, not travel spend, and it usually lands in a different GL account entirely.

Channel 4: Direct and off-platform bookings. A regional manager books a flight on a personal card, a team books accommodation through a marketplace, a client visit is arranged by the client. Some of this surfaces later as a claim. Some never surfaces at all.

Each channel has a different owner, a different system, a different data structure and a different arrival latency. Claims arrive weeks after the trip. Card statements arrive monthly. Agency invoices arrive on the supplier’s billing cycle. The same trip is therefore recorded in three places across three different months.

No one designed this. It accreted.

The consequence is that “travel spend” is not a number that exists anywhere. It is a number that must be constructed, and constructing it requires judgement calls that different people make differently — which is exactly why two reports on the same quarter rarely agree.

Reporting Is Not Analytics

A useful distinction, and one that explains why buying a BI tool usually does not solve this.

Reporting answers questions you already knew to ask, on a schedule. Monthly T&E spend by department. Year-to-date against budget. These reports are usually already running, and they are usually built on one channel — normally the expense system — which means they describe a fraction of actual travel cost.

Analytics answers questions you did not anticipate, on demand. Which cost centre’s travel cost per head rose fastest this year and why? Are we paying more per night at the hotel chain we have a rate agreement with than at the one we do not? Which policy rule generates the most exceptions, and do we approve them anyway?

The second set cannot be answered by a scheduled report because the questions are not known in advance. They require a dataset that is complete, structured and queryable — and that is a data modelling problem, not a visualisation problem.

This distinction matters commercially. The Global Business Travel Association and most travel management practice place supplier negotiation and policy design as the largest levers on T&E cost. Both require analytics, not reporting. You cannot negotiate a hotel rate without knowing your actual room-night volume with that chain, and if a third of your bookings arrive through channels you cannot see, your volume claim is wrong and your negotiating position is weak.

The Four Data Problems That Break T&E Analysis

Before any dashboard is worth building, four specific data defects have to be fixed. Each one independently makes whole categories of analysis impossible.

1. Free-text categories

If employees type their own expense descriptions, or choose from a category list that has grown to forty overlapping options across five entities, nothing can be grouped reliably. “Client lunch”, “Business meal”, “Meals - customer”, and “Entertainment” may all describe the same thing.

The fix has two parts: a controlled category taxonomy that is short enough to use correctly, and automated classification that assigns the category from the receipt and merchant data rather than relying on the claimant. AI classification matters here specifically because it produces consistent categories, and consistency is what makes grouping possible.

2. Missing or wrong cost centres

Cost centre is the dimension almost every T&E question depends on, and it is the field most often wrong — especially on corporate card transactions, which arrive with no cost centre at all and get assigned in bulk during reconciliation.

This is a master data problem as much as a capture problem. Cost centre validity depends on employee records, entity structures and budget hierarchies staying synchronised, which is why we treat master data sync across ERP and HCM as a prerequisite rather than an adjacent concern.

3. Currency without conversion basis

Multi-country T&E generates claims in a dozen currencies. If the system stores only the converted base-currency amount, you have lost the ability to analyse the original spend — and you have no way to explain whether a rise in Indonesian travel cost is real or an FX effect.

Store three things on every line: original currency and amount, conversion rate, and rate date. Analysis then becomes possible in both constant and reported currency, and FX variance stops being an unexplained residual — a discipline the ACCA treats as basic hygiene for any multi-currency cost reporting.

4. No trip identifier

This is the one most organisations have never considered, and it is the reason cost-per-trip is so rarely reported.

A single business trip generates a flight on an agency invoice, a hotel on a corporate card, four taxi claims and two meal claims — six records across three systems with nothing linking them. Without a trip identifier propagated from the travel request through every downstream record, cost per trip and cost per traveller cannot be computed at all, and the most useful unit of analysis in the entire category is unavailable.

Attaching this identifier at the travel request stage is far easier than reconstructing it later, which is the core argument for integrated travel and expense management and for starting the process with a structured request rather than a form. We have covered the related problem of travel requests approved on estimates and booked at actuals.

What a T&E Analytics Layer Should Actually Show

Assuming the data problems are addressed, six views carry most of the decision value.

Total T&E by entity, cost centre and category. The baseline. Must reconcile to the ledger, which means it must include agency invoices booked through AP, not just employee claims.

Cost per trip and cost per traveller. The normalising metrics. Total spend rising 20% means nothing without knowing whether trip volume rose 25%. This is the view that turns a cost conversation into a productivity conversation.

Policy exception rate and disposition. Not just how many claims broke a rule, but which rules, in which cost centres, and how many were approved anyway. A rule with a 90% override rate is not a control; it is a note. Our guide to flagging policy violations at submission covers the enforcement side of this.

Claim cycle time. Submission to approval to payment, segmented by approver and cost centre. This is the employee-experience metric and the one most likely to surface a single approver who is holding up an entire department.

Unsubmitted and unliquidated backlog. Expenses incurred but not claimed, plus cash advances issued but not liquidated. This is an unrecorded liability and it distorts every period comparison. It is also directly actionable — see automating cash advance liquidation follow-up.

Supplier concentration. Room nights by hotel chain, sectors by airline, spend by agency. This is the input to every rate negotiation and the view that is impossible without channel unification, because most supplier spend arrives outside the expense system.

Billable versus absorbed cost. For organisations that recharge travel to clients, the split between recoverable and absorbed spend — and what is recoverable but not yet invoiced — belongs in the same pack, because it is the difference between cost and margin.

The Accrual Problem Hiding Underneath

There is a reporting consequence of poor T&E visibility that rarely gets named: period misstatement.

An expense incurred in March and claimed in May belongs, under accrual accounting, to March. In practice most organisations recognise it in May, because that is when it entered the system. With a small and stable backlog the distortion washes out. With a large or volatile backlog — a conference season, a quarter-end sales push, a submission deadline that everyone ignores — it does not.

The month-end symptom is familiar: departmental costs that swing without an operational explanation, and a controller who cannot say whether a cost centre is genuinely over or simply had a late-submission month. Guidance from bodies including the Institute of Management Accountants and IFAC is unambiguous on the principle; the obstacle is almost always that the backlog is not visible enough to accrue against.

Measuring the unsubmitted backlog is what makes the accrual possible. It is also, usefully, the metric that motivates employees to submit on time, because a backlog that is reported by department gets managed by department. For teams working on close cycle time more broadly, this connects directly to orchestrating month-end close.

How Peakflo Helps

Peakflo’s travel and expense management module is built so the analytics layer is a by-product of processing rather than a separate reconstruction effort. Claims, imported corporate card statements and travel supplier invoices processed through accounts payable land in one structured dataset, with AI classification assigning consistent expense categories and GL coding from receipt and merchant data instead of relying on claimant free text.

Cost centre, entity, project and trip references are validated against live master data at capture, and original currency, rate and rate date are retained alongside the base-currency amount so FX effects can be isolated rather than assumed. Because the travel request, the booking, the card charge and the downstream claims carry a common reference, cost per trip and cost per traveller are computable without reconstruction — as are policy exception rates, cycle time by approver, unsubmitted backlog by department and supplier concentration across airlines, hotels and agencies.

The result is that a CFO question about last quarter’s airfare is a filter, not a project. To see what your own T&E data would look like unified across claims, cards and supplier invoices, request a demo.

Our Verdict: Where Does This Pay Off?

Prioritise T&E analytics if:

• Your travel spend runs across more than one entity or currency, where manual consolidation is both hardest and most error-prone
• You have rate agreements with airlines, hotels or agencies and cannot verify your own volume against them
• Departmental cost comparisons are routinely disputed and the disputes are about data, not performance
• A significant share of travel cost arrives as agency invoices through AP, which means your current expense-system reporting is structurally incomplete
• Month-end travel accruals are estimated rather than calculated

Lower priority if:

• Nearly all T&E flows through a single channel already and your expense system reporting is genuinely complete
• Total T&E is small enough that even a 15% saving would not fund the work — check this honestly before assuming it, because the invisible channels are exactly the ones that make T&E look smaller than it is
• Your immediate problem is control rather than insight, in which case pre-payment expense audit screening is the higher-value first step

The most common mistake is sequencing. Teams buy the dashboard, discover the data will not support the questions, and conclude that T&E analytics does not work. The unification and structuring work has to come first; the visualisation is the easy part.

Conclusion

The reason finance cannot answer travel spend questions quickly is not that the question is hard. It is that the answer has to be manufactured each time from sources that were never designed to be combined.

Every hour spent reconciling claims against card statements against agency invoices is an hour spent rebuilding a dataset that should have existed from the moment each transaction was captured. And because it is rebuilt by hand each time, it is rebuilt slightly differently each time — which is why nobody quite trusts the number.

The fix is unglamorous: unify the channels, enforce structured categories and cost centres, keep the currency basis, and attach a trip identifier early. Do that, and the dashboard takes an afternoon. Skip it, and no amount of visualisation tooling will produce a number anyone is willing to defend in a board meeting.

Frequently Asked Questions

What is travel and expense analytics?

Travel and expense analytics is the practice of consolidating T&E spend from every channel — employee claims, corporate cards, agency invoices and direct bookings — into one structured dataset so finance can answer questions about cost, policy compliance and supplier performance without rebuilding a spreadsheet each time.

Why can’t finance teams see total travel spend?

Because T&E arrives through at least four disconnected channels, each with its own data structure. Employee claims sit in one system, corporate card statements in another, travel agency invoices in accounts payable, and direct bookings on personal cards nowhere at all. No single system holds the full picture, so totals are assembled manually and only on request.

What metrics should a T&E dashboard track?

Six metrics cover most decisions: total T&E spend by entity and cost centre, cost per trip and per traveller, policy exception rate by category, claim cycle time from submission to reimbursement, unclaimed or unsubmitted expense backlog, and supplier concentration across airlines, hotels and agencies.

How is spend visibility different from budget tracking?

Budget tracking is forward-looking and asks whether a planned commitment fits within an approved allocation. Spend visibility is backward-looking and asks what was actually consumed, by whom, with which suppliers, and against which policies. Most organisations have some budget tracking and very little spend visibility.

How long should it take to answer a travel spend question?

Seconds, not days. If a CFO asks what the group spent on airfare in Q3 across three entities, a working analytics layer answers immediately. If the answer requires exporting from three systems and reconciling in Excel, the organisation has reporting, not analytics, and the answer will be stale before it is delivered.

What data quality problems break T&E analytics?

Four recur: free-text expense categories that cannot be grouped, missing or wrong cost centre codes, currency amounts stored without the rate or date used for conversion, and no trip identifier linking related expenses. Any one of these makes cost-per-trip and per-category analysis impossible regardless of the dashboard tool.

How do you measure travel policy compliance?

Measure the exception rate — the percentage of claims that trigger a policy rule — segmented by category, cost centre and grade, plus the value of those exceptions and how many were approved anyway. A high approval rate on exceptions means the policy is aspirational rather than enforced.

What is an unsubmitted expense backlog and why does it matter?

It is the pool of business costs already incurred by employees but not yet claimed. It matters because it is an unrecorded liability: the expense belongs to the period it was incurred, not the period it was submitted. A large backlog distorts month-end accruals and makes departmental cost comparisons unreliable.

Can AI improve travel expense analysis?

Yes, in two specific ways. AI classifies free-text and merchant data into consistent categories, which fixes the grouping problem that breaks most dashboards. It also detects pattern shifts — a cost centre drifting upward, a traveller consistently near limits — that threshold-based reports cannot surface.

Who should own travel and expense analytics?

Finance should own the data model and definitions, because the numbers must reconcile to the ledger. Procurement or a travel manager should own supplier and negotiation insights drawn from it. Splitting ownership the other way produces dashboards that look useful and do not tie to reported cost.