The Two Excel Connectivity Tests Most FP&A Buyers Never Run

Bringing your own workbook and refreshing it are the standard demo checks. Two harder tests, structural change and a second login, are the ones that predict whether the connection still holds in month six.

Two demo checks are now standard advice for anyone buying FP&A technology: bring your own messy workbook rather than accepting the vendor’s sample, and run a real refresh to see whether formulas, named ranges, and macros survive it. Both are worth doing. Neither is where these deals go wrong.

Platforms that reach a shortlist tend to pass both, because both test the connection on the day it is made. The failures that surface in month six come from two things a demo is not built to reveal: what happens when a model changes shape, and whether governance holds when a second person logs in.

Neither is an edge case. Both are routine, and both stay hidden in a session the vendor controls, on a workbook the vendor prepared, in a tenant with one user in it.

Why this differs from native Excel tools: Power Query and VBA macros automate tasks only after data lands in a spreadsheet. An Excel-connected platform automates the entire data collection, consolidation, and governance layer upstream, before Excel even opens.

The two checks that most platforms pass

Bring the most complicated model your team runs, not a clean sample, and ask the vendor to connect it during the session. Then refresh it and inspect what surrounds the data: nested formulas, named ranges, conditional formatting, linked workbooks, macros. A platform where Excel is the planning surface handles both. A platform below that level fails visibly, offering a template instead of your file, or handing back a rebuilt sheet. That much is standard advice by now.

It screens out the bottom of the market and then stops telling you anything. Both checks examine the connection on the day it is made, which is the day it is most likely to work. A platform sitting at the second of the four integration levels can pass them with a competent sales engineer, and the tells that usually give that away show up in the sales deck rather than in the demo. Everything left on a shortlist afterwards looks the same.

Test 1: Change the shape of the model, then refresh again

Working models do not hold still. Somebody inserts a row above a total, renames a tab, splits a cost center, or adds an account to the chart of accounts, and that happens several times a quarter rather than once at implementation.

So run the refresh twice. Refresh once, then insert a row in the middle of a mapped table, rename a tab, add an account, and refresh again. Then ask the direct question: what is the mapping anchored to? Cell addresses, named ranges, or account codes?

Cell-address binding is the answer to watch for. It works perfectly on the day it is configured and breaks the first time anyone restructures a sheet, which puts the maintenance burden on whoever in the finance team is least able to refuse it. A platform anchored to account codes or named ranges absorbs the change. A platform anchored to cell positions needs a remap, and the better vendors will say so when asked plainly.

Test 2: Ask for a second login

Governance is the capability most likely to exist only in the demo tenant. Role-based access and audit trails look identical in a sales environment whether or not they hold up under real users, real permissions, and real edge cases, which is why every vendor’s governance slide reads the same.

Ask before the call for a second account with a restricted role, and open the same workbook from it. Two things to check. Can that user reach figures they should not, by opening a hidden tab, refreshing a range outside their scope, or drilling into a total? And does the audit log record the change you made minutes earlier, including who made it, when, and which source the surrounding data came from?

A platform that has this can show the log entry during the call. A platform that does not will describe the roadmap. A vendor that cannot produce the second account before the demo has answered the question already.

Both tests are answered by watching rather than listening, and both take minutes. They are also cumulative in a way the standard checks are not: connecting a workbook once is a different engineering problem from keeping it connected while a team edits it every week, and the second one is what you are buying.

What a connection cannot fix

The nuance most vendor pages skip is that a live connection doesn’t automatically fix data quality. If a spreadsheet was already carrying formula errors or broken links, syncing it faster just distributes bad numbers quicker. Raymond Panko’s review of field audits, the source of most spreadsheet error figures in circulation, found errors in 88% of the 113 real-world spreadsheets examined across seven studies (What We Know About Spreadsheet Errors). Connecting a spreadsheet to live data reduces manual re-entry risk and does nothing to that underlying rate.

Maintenance is the other thing a demo hides. Rollouts that connect to models a team already has tend to complete in weeks, while rebuilds that ask a team to leave Excel behind can run months to over a year. What the two tests above measure is the cost after that: a platform that needs a remap every time a sheet changes shape has moved the work rather than removed it.

Datarails FP&A at a glance

Datarails FP&A carries the role-based permissions and audit logs that test 2 asks for, and they can be checked under a second login rather than taken from a table. Mapping behaviour under structural change is worth putting to any vendor directly, including this one. The table below covers the rest.

Every platform in this category uses the same language, and the levels underneath it are not visible from a homepage, a demo script, or a feature table. The two checks everyone runs screen out the bottom of the market. The two above screen out the rest.

An hour of a demo is the cheapest moment to find out. A pilot is the next cheapest. After that the answer arrives mid-close, which is the worst available time to learn that a workbook has to be remapped.

Two Excel Connectivity Tests FAQs