Travel analytics is the analysis of a corporate travel programme's booking, payment and supplier data to explain why spend and behaviour changed rather than only record what happened, which is why the identity of whoever performed the analysis, and what that party is paid to do, is part of the evidence.
Every travel programme has analytics now. The monthly pack arrives, the quarterly deck gets built, the platform has a tab marked Insights, and somewhere inside all of it is a sentence telling you what your programme did last period and what you ought to do about it.
Almost nobody asks who wrote that sentence.
It is a fair thing not to have asked. The systems were built to record transactions, and analysis arrived later as something those systems also happened to do, usually supplied by whoever already held the data. Nobody sat down and chose the arrangement. It accumulated.
The question matters more than it looks, because an analysis is not neutral in the way a transaction record is. A record states what happened. An analysis states what it means, and meaning is where an interest can sit quite comfortably without anybody lying about anything.
The raw material is scattered to begin with. Research published by GBTA on 19 May 2026, covering 269 travel buyers across North America and Europe, found that within that sample only 12% of global buyers have a consolidated view of their programme from a single data source, and 63% of global travel managers name the lack of consolidated reporting as a top challenge. When the data sits in five places, somebody decides which version of it becomes the answer. That decision is the analysis, and it is made by whoever is holding the pieces.
In this article
- What is travel analytics?
- What is the difference between travel reporting and travel analytics?
- Who produces the analysis of a corporate travel programme?
- Can a travel supplier assess the programme it runs?
- What should you ask about the source of a travel analytics report?
- Why do travel numbers from different systems disagree?
- Does travel analytics need to be real time?
- How often should travel analysis reach leadership?
- Who should own travel analytics inside a company?
- Frequently asked questions
What is travel analytics?
Travel analytics is the practice of joining a corporate travel programme's booking, card, expense and supplier records and analysing them to explain why spend, compliance and traveller behaviour changed. Reporting shows the figures. Analytics reaches a conclusion about them and names what to do next.
The phrase covers three different things, which is worth clearing up if you arrived from a search. It is used for destination and tourism demand analysis. It is used for consumer travel search trends. And it is used for the subject of this article, which is the analysis of a company's own managed programme: its routes, fares, hotel rates, policy, suppliers and the people flying.
Inside a corporate programme the work has three parts. Getting the data into one shape. Finding what changed. Saying what it means.
The first part takes most of the effort and gets none of the attention. One hotel chain appears under a different name in the booking file, the card file and the expense claims, and often under several within each of them. Cost centres move mid-year. A traveller books through the tool in January and direct with the airline in March, and nothing joins those two records to the same person unless somebody makes it join.
The second part is arithmetic. The third is a judgement, and a judgement has an author.
This article is not about whether software can do that work quickly. We have written elsewhere about why travel and expense reporting fails and what an analysis engine does about it. This one is about the prior question: who performed the analysis, and on whose behalf.
What is the difference between travel reporting and travel analytics?
Reporting states what happened. Analytics states why it happened and what to change. The distinction most sources stop at is that one is descriptive and the other explanatory. The more useful distinction is that reporting can be produced without a point of view, and analysis cannot.
The industry version of this difference is right as far as it goes. A report shows air spend down 4% quarter on quarter. An analysis says air spend is down 4% because two teams stopped travelling to one market after a project ended, that the fare paid per sector actually rose, and that the saving will not repeat.
Notice what the second sentence needed. It needed a reason, and a reason is a choice between several that all fit the numbers. Somebody picked one.
There is a second axis the industry version leaves out, and travel managers feel it every week: whether you had to ask. Anything you requested arrives after the decision it was meant to inform, and the request itself is shaped by what you already suspected.
Who produces the analysis of a corporate travel programme?
In most programmes the analysis is produced by whichever party already holds the data: the agency that books the trips, the online booking tool, or the card and expense system. A smaller number of companies analyse it themselves. Very few use a party with no share of the spend.
Each source sees a different slice of the programme, and each is measured on something. Neither fact is a scandal. Both change what the analysis is able to conclude.
The last row is ours, and the honest reading of it is that independence is worth nothing on its own. A neutral party with one feed is a neutral party with one feed. What independence buys is the absence of a reason to prefer one conclusion, which only starts to pay once there is enough data for a conclusion to be contested.
Can a travel supplier assess the programme it runs?
A travel supplier can produce accurate figures about a programme it runs. What it cannot do is go looking for a finding that would cost it the contract. Nobody has to be dishonest for that to be true: it is a property of where the analysis sits, not of the people doing it.
Look at what the arrangement asks of a supplier. It holds the booking data because it made the bookings, and it gives the reporting away because that is cheaper than charging for it. The relationship comes up for renewal every few years, and the reporting is one of the things a buyer weighs.
None of that requires anybody to falsify a number. It requires only the ordinary tendency to pursue the questions you would like the answer to and leave the others until next quarter. It is why companies appoint external auditors for the accounts rather than asking the finance team to certify its own work, and nobody reads that arrangement as an accusation against the finance team.
The pattern is the norm in this category rather than the exception, and it can be counted. When we measured the organic results for the search phrase business travel reporting on 24 August 2026, using OpenSEO in the US market (location code 2840), eight of the seventeen organic results were travel management companies, across seven distinct domains. Every one was publishing guidance on how to report on a travel programme, and every one sells services to programmes whose performance those reports examine. That is a PredictX measurement of a public search result and you can reproduce it in an afternoon.
Travel buyers describe the consequence without ever using the word independence. Mark Ziegler, a corporate travel buyer quoted by Business Travel Executive in May 2025, said of hotel pricing: "We can't measure or audit dynamic rates. We are left to trust the GDS and the hotels."
Scott Davies, chief executive of the Institute of Travel Management, told The Business Travel Magazine in August 2024 that "Travel data analytics is a complicated area, with no ideal solution yet to help buyers resolve the challenges of how to consolidate travel and expense data not only via their TMC, but from multiple other sources."
The test applies to us, and it should. PredictX sells software rather than travel and takes no share of the fare, which removes one interest but not all of them: we have an obvious reason to want you to conclude that your current analysis is compromised. So run the five questions below over this page too.
What should you ask about the source of a travel analytics report?
Ask five things of any travel analysis before you act on it: who performed it, whether it could have concluded something expensive for them, what it could not see, whether you can trace a figure back to the records behind it, and whether it would have arrived if nobody had asked.
We call this the Provenance Test. The Provenance Test is a five-question check on where a piece of travel analysis came from, run before the finding is acted on rather than after it is disputed. It takes about ten minutes and needs no software.
The fourth question is the one most people skip and the one that separates a finding from a claim. If the analysis says three teams are driving a rise in last-minute bookings, you should be able to ask which three, which bookings and on which dates, and get rows back rather than the original sentence rephrased.
Why do travel numbers from different systems disagree?
Travel numbers disagree because each system counts a different event at a different moment. The booking tool counts the reservation, the agency counts the ticket, the card counts the charge and finance counts the accrual. None of them is wrong. They are answering four different questions that share a name.
Then there is the mess underneath. A ticket exchanged twice shows as three transactions in one system and one in another, and a refund lands in a later period than the booking it reverses. Currency converts at the booking date in one feed and the settlement date in another.
Normalising all of that is the hard part of travel data analytics and it is invisible in every demonstration you will ever be shown. It is where most in-house projects stall, because it never finishes: a schema change at a supplier breaks a connector, and a connector nobody maintains produces confident numbers that are quietly wrong.
Here is why it bears on provenance. Whoever does the reconciling decides which of the four numbers becomes the answer. That decision is rarely written down and almost never surfaced to the person reading the report.
Does travel analytics need to be real time?
Travel analytics rarely needs to be real time, and most programmes are better served by a figure that is stable than one that is current. A monthly snapshot lets the words and the numbers agree. That choice has a cost: an intra-month question gets last month's answer.
The genuine real-time cases in corporate travel are disruption and duty of care, and both belong to systems built for them. Spend analysis is not one of them. A fare trend that is meaningful on Tuesday is still meaningful on Friday, and a figure that changes while two people are discussing it is worse than a figure that does not.
This is the design choice behind Overture, and it is a limit rather than a feature. Detection runs deterministically at the warehouse, so the same data produces the same findings, and the narration is generated from those findings and cached. That keeps the sentences and the figures in agreement, and it means re-running the pipeline mid-month still returns the previous month's snapshot. Anyone offering both stability and currency is offering more than we are.
There is a longer treatment of what happens when the cadence is quarterly instead, and of who is holding the pen at that meeting, in our paper on the quarterly review.
How often should travel analysis reach leadership?
Travel analysis should reach leadership monthly, and it should arrive without anyone requesting it. A quarterly cycle means a problem found in month one is raised in month four, by which time the negotiation it affected has closed. Monthly is the shortest cadence at which travel data is stable enough to narrate.
The cadence question turns out to be a question about who prepares it. A quarterly pack exists because a meeting exists, and somebody spends a week building it. That person has a view, a deadline and an audience, and the pack carries all three.
A briefing produced on a schedule by a process nobody had to trigger has a different property. No person chose the story, because nobody was there to choose it. Fixed rules decide what counts as a finding and the model picks the most salient of them to lead on, so what appears is drawn from whatever detection found that month, welcome or not.
We publish claims of our own about how much more attention a narrative briefing gets than a dashboard, and we cannot show you the workings behind them. By the standard of the Provenance Test above, that makes them claims rather than findings, and you should weight them accordingly.
What we can set out is the mechanism. Overture's monthly executive reporting runs in a fixed order: deterministic detection at the warehouse, then a language model that narrates what detection already found without looking beyond the findings it is handed, then a composed briefing that reaches the reader on a schedule rather than on request. The model is a renderer, not a detector. Overture was announced recently and carries no deployment history you can check, which is worth knowing before you weigh anything said on this page.
Who should own travel analytics inside a company?
Travel analytics should be owned by whoever is accountable for the travel budget, usually procurement or the travel programme lead, with finance owning the reconciliation to the ledger. The supplier that books the travel should supply data into it. It should not be the party that decides what the data means.
That split is easy to write and awkward to implement, because the supplier is often the only party with anyone who knows how to run the query. The practical route is not to take the work away on day one. It is to separate the two jobs, supplying the data and interpreting it, so they can be held by different parties later.
Three things make the separation possible. Get the booking, card and expense feeds delivered somewhere you control, in raw form, as a standing arrangement rather than an export somebody runs. Write down the definition of every figure in a leadership report, including which system it comes from and when it is counted. And ask for the rows behind any finding you are asked to act on, until asking stops being unusual.
None of those three needs new software. They need the arrangement written down once, which is the part that has usually not happened, because the arrangement was never designed. It accumulated.
Frequently asked questions
What is the difference between analysing a travel programme and reporting on it?
Reporting assembles figures and presents them. Analysing decides which figures matter, why they moved and what should change, which means it makes choices a report does not. Both are useful. Only one of them carries a point of view, and only one needs its author disclosed.
What data does travel analytics need?
Four feeds joined to the same records: bookings from every channel, the fares and rates available at the time, card and expense transactions, and supplier contract terms. The available-fare feed is the one most often missing, and it is the one that turns a payment into a saving you can evidence.
What KPIs does a travel analytics programme track?
Most programmes track air and hotel spend against budget, fare and rate paid against a reference, policy compliance, advance booking window, channel adoption and supplier share against contract. The specific list matters less than whether each figure has a written definition and a single system of record behind it.
What is the difference between travel analytics and travel data analytics?
In practice the two phrases describe the same work, with travel data analytics putting slightly more weight on the joining and cleaning stage and travel analytics on the conclusions drawn from it. Neither term is defined by a standards body, so ask what a supplier means by it.
Can a travel manager verify a travel analytics figure independently?
Yes, if the underlying rows are available. Ask for the transaction-level records behind any figure in a leadership report, then re-run the count yourself in a spreadsheet. A figure that cannot be reproduced from records you hold should be treated as an opinion with a number attached.
Does travel analytics require an in-house analyst?
No, though most programmes that do it well have had one at some point. The analyst's largest contribution is the joining and the definitions rather than the queries. Once those are written down and maintained, the analysis can be produced by software or by a party outside the company.
How much travel data do you need before analytics is worth doing?
Twelve months of joined history is the practical floor, because nearly every useful finding is a comparison against the same period last year. On a small programme the sample may be too thin for patterns to separate from noise, and a spreadsheet with clean definitions will serve better than a platform.
Who should sign off the analysis of a travel programme?
Whoever is accountable for the budget it describes, which in most companies is procurement or the travel programme lead rather than finance. The supplier that books the travel should be asked to comment on the findings and should not be the party that approves them for circulation.
Before you accept the next analysis
Ask for the rows. That one habit does most of the work of the other four questions, it costs an email, and the quality of the answer will tell you more about your current arrangement than any assessment of the software would.
If you want to see what the alternative looks like, the Overture page sets out what arrives each month, how detection and narration are kept apart, and what the briefing will not do.
Related Posts
What an executive dashboard is for, and the month it stops being opened
Why leadership does not fund your travel programme, and what executive buy-in turns on
Natural language generation: how to tell whether a report a machine wrote can be trusted