A spreadsheet service is easier to explain when the result is concrete: one tidy inventory file, a dependable contact import or an expense sheet with consistent categories. “I can help with Excel” is not a scope. Your first offer should name the file, the problem and a test that proves the work is finished. You do not need to claim to be a financial analyst or promise that a spreadsheet will transform a business.
Pick one narrow cleanup
Choose a problem you can reliably solve: duplicate rows, inconsistent dates, extra spaces, missing category labels or a broken import format. Do not bundle bookkeeping decisions, forecasting and automation into your first cleanup. If a customer asks you to decide which transactions are deductible or which accounts a payment belongs in, that is a different service requiring different competence.
A useful offer is: “I will standardize one supplied contact file to your provided column template, flag duplicates for your decision, and return an exception list.” Put a row limit and one revision in the scope. Never automatically delete suspected duplicates where two people might share a name. Flagging uncertainty is a deliverable, not a failure.
Build a sample without borrowing anyone’s data
Create a synthetic workbook with deliberately messy dates, duplicate identifiers and inconsistent capitalization. Keep a read-only original. Produce a clean copy, a short change log and a page explaining which ambiguous entries you left untouched. A prospective customer should understand the result without opening a file with macros.
Your startup-cost worksheet should include software you are licensed to use commercially, secure file storage if needed and time to learn the specific import requirements. Enter actual quotes; do not buy a subscription just to make your service look established. An existing spreadsheet application may be enough to build the sample. Check compatibility with the customer's version before promising formulas or formatting.
Find the first customer with a problem-led offer
Look for a small organization you already know that manually retypes the same information. Ask permission to show your synthetic example. A useful question is: “Which file do you have to fix every month before you can use it?” Do not ask them to send the real file immediately. Start with a screenshot with private details removed or a small redacted extract.
Offer a paid pilot with a clear maximum file size and delivery date. If someone cannot describe the desired output, charge for a separate diagnosis only if they understand what it includes. Do not commit to a fixed-price rescue based on a vague screenshot. Use the pricing guide before quoting.
Agree on acceptance tests and a data boundary
Before you begin, write down these checks:
- Input row count, output row count and the reason for any difference.
- Totals or identifiers that must remain unchanged.
- Expected date formats, column names and allowed categories.
- How the client will review exceptions and approve the output.
- Where files will be stored, who can access them and when copies will be removed.
Ask the customer to make a backup and work on a copy. Do not upload their file into an online converter or AI service without explicit permission and a suitable data arrangement. Do not ask for account passwords. If you need access, ask for a limited invitation through the client's approved system. Avoid live production imports on a first engagement; supply the file for their administrator to test.
Deliver a handover, not just an attachment
Return the clean file, the change log and the unresolved-questions list. Explain which transformations were automated and which were manual. Ask the customer to test the file in its destination before accepting it. Agree in advance whether fixing an error you introduced differs from adding new requirements.
Record all the time: discovery, messages, backups, cleaning, checking and the handover. Use the true hourly calculator. A cleanup that looks quick can be unprofitable when the input changes halfway through. Keep a new version number for every delivery so a client cannot accidentally import an earlier draft.
Know when to stop or specialize
Decline work involving unlicensed financial advice, payroll corrections beyond your competence, hidden password removal or data obtained without permission. If every job has a different problem, narrow your offer further. A repeatable import cleanup can be a better experiment than becoming a general emergency spreadsheet desk.
Check business requirements with your local authorities; the SBA explains that permits depend on activity and location. After your first paid pilot, ask what will happen to the file next month. Only offer recurring work if the client has an actual recurring task and you can describe the maintenance boundary. Compare a recurring service with one-off work in the platform-versus-direct guide.
Your working checklist
- Build a synthetic before-and-after workbook.
- List exactly which changes are included and excluded.
- Agree on backup, acceptance tests and deletion before access.
A common question
Should I accept a spreadsheet containing sensitive customer data?
Not as a casual first project. Ask for a redacted or synthetic sample and agree on approved storage, access and deletion. Decline regulated or highly sensitive data unless you have appropriate expertise, safeguards and contractual arrangements.