Built with purposeThe most common hesitation I hear before someone sends over an export is not about the diagnostic itself, it is about the export. People worry, reasonably, that handing over a spreadsheet means handing over everything: every client name, every price, every employee's salary. It does not need to, and honestly it should not, because a diagnostic never needed all of that in the first place.
What a diagnostic is actually checking
The point of an initial review is narrow: to see the real shape of your data before anything gets quoted. That means checking how many source systems are involved, whether dates and identifiers are formatted consistently, whether the same customer or product appears under two slightly different names in two different exports, and roughly how many rows and columns we are dealing with. None of that requires knowing which specific client paid which specific amount. It requires knowing that a "client" column exists, what type of data lives in it, and whether it is clean.
I can tell almost everything I need to know about a POS export, an accounting ledger or a payroll file from its structure and a representative sample of rows, not from every row for every month you have ever operated.
What to strip out before sending anything
Three things are worth removing or masking before an export leaves your building, in order of how often people forget them. Names, whether client, employee or vendor, can be replaced with a generic label like Client 1, Client 2, keeping the label consistent across rows so relationships in the data still make sense. Anything genuinely sensitive on its own, like a national ID number, bank account number or salary figure tied to a named individual, should come out entirely rather than masked, because a diagnostic never needs it. Exact prices can be scaled by a consistent factor if pricing itself is the sensitive part, which preserves every pattern in the data (margins, seasonality, mix) while hiding the real numbers.
None of this takes long. Most spreadsheet tools can do a find-and-replace on a name column in under a minute, and scaling a price column by a fixed multiplier is one formula copied down the sheet.
Why a sample beats a full dump
A common instinct is to send everything, on the theory that more data means a more accurate quote. It usually does the opposite. A twelve-month export with every transaction takes longer to review, is more likely to contain something you forgot was sensitive, and does not actually tell me more about the shape of the data than three representative months would. What I am checking for, structural consistency and data quality, is visible in a sample just as clearly as in the full history, usually more clearly, because a smaller file is easier to actually look at line by line instead of skimming.
If seasonality matters to your business, a sample spanning a slow month and a busy month tells me more than twelve identical-looking months from the middle of the year, even though it is a fraction of the size.
What actually needs to travel intact
Strip the names and the truly sensitive fields, but keep the structure exactly as it is: column headers, date formats, how numbers are entered, any codes or categories the system generates on its own. That structure is the actual object of the diagnostic. A file that has been reformatted to look tidier than what your system actually produces will get me a diagnostic based on data you do not really have, which helps nobody once the real build starts pulling from the real, messier export.
If you would rather not send a file at all
Sending nothing and just describing the company, what software you run and what you wish you could see, works too. It produces a rougher estimate than an actual sample, because I am working from a description instead of the real shape of the data, but it is a legitimate way to start if privacy concerns outweigh the value of a tighter number at this stage. Either way, the contact form is the same one, and the choice of how much to send is entirely yours.