Skip to main content
industry

Data Scientist Invoice Template

A data science engagement rarely bills like ordinary consulting hours. It usually runs in phases: a discovery or feasibility study, a model build that ends with a validation report the client accepts, deployment into the client's systems, and then a monthly monitoring and retraining retainer. Each phase has its own trigger and often its own price, so the invoice should name the phase, the deliverable (report, model version or go-live date) and when it was accepted. Two other lines make or break a data science invoice. Compute: training runs on cloud GPUs cost real money, and the cleanest way to bill them is at cost, from a tagged cost report, with the provider, project tag and period on the line. Confidentiality: the invoice goes through the client's accounts payable, so it describes deliverables by name and version, never by the data they contain or the results they show. This template has a field for each of those, plus the contract clause covering who owns the code. The download is an editable HTML invoice with invoice number, dates, a line-item table and totals; fill it in your browser and print it or save it as PDF, or describe the work and let InvoiceQuickly draft an editable invoice you check before saving. For reporting, dashboards and ad hoc analysis, the data analyst invoice template fits better.

Draft this invoice
Preview this editable document

Download the HTML file to edit its fields, then print or save as PDF. Totals are entered manually.

What This Template Includes

  • Statement of work or contract reference
  • Engagement phase: discovery, model build, deployment or monitoring
  • Deliverable and acceptance: report, model version or go-live date
  • Fixed phase fee, or the share of it billed now
  • Hourly work: task, hours and rate, with any agreed cap
  • Monitoring retainer: month, hours included and used, overage rate
  • Cloud compute at cost: provider, account, project tag and billing period
  • Data, API or licence costs at cost, with receipt reference
  • Ownership of code and models: the contract clause that applies
  • Client purchase order or cost centre

How to Create Your Data Scientist Invoice

  1. 01

    Describe your work

    Type a plain English description of the services you provided, the client, and the amount.

  2. 02

    AI generates your invoice

    InvoiceQuickly's AI drafts the client, line items, dates, and totals for you to review before saving.

  3. 03

    Review, download, and send

    Check every detail, download the PDF, or use Pro to email the PDF directly to your client.

Recommended Payment Terms

Machine learning clients are usually companies paying through accounts payable on net 30, so get the purchase order number and the invoice contact before the first invoice. Tie each phase to a trigger the client controls and you can evidence: discovery on delivery of the feasibility report, model build on written acceptance of the validation report, deployment on go-live. Avoid making payment depend on a model metric reaching a target in production; data drift outside your control can delay it indefinitely, so if the contract has a performance target, define how and when it is measured. Bill the monitoring retainer monthly in advance with overage in arrears. Pass compute through monthly at cost, not at the end of the project, because a few weeks of GPU training can be a large sum to carry on your own card. If compute will be heavy, ask the client to run it in their own cloud account from the start.

Need help writing payment terms? Read our guide to invoice payment terms for best practices and templates.

Tax Information

Most data science work is billed as a professional service, but a few US states tax data processing, and the line between consulting and processing is drawn by each state. Texas is the clearest example: its rule (34 Texas Administrative Code § 3.330) defines data processing as the computerized entry, retrieval, search, compilation, manipulation or storage of data, exempts 20% of the charge, and treats consultation as a nontaxable related service. Model design and advice may therefore be treated differently from ongoing hosting or scoring of the client's data, so check with your state before adding tax; the data analyst invoice template shows the Texas arithmetic. InvoiceQuickly applies one tax percentage to the whole invoice, so put taxable and nontaxable lines on separate invoices. A US business that pays you $2,000 or more for services in 2026 generally reports it on Form 1099-NEC, so expect a Form W-9 request, or Form W-8BEN if you are outside the US. InvoiceQuickly copies saved tax IDs to new invoice PDFs; check the saved IDs and PDF before sending when needed. General information, not tax advice. Last reviewed 28 October 2026.

Industry context (2026)

Data science is a large and growing field: the US Bureau of Labor Statistics' Occupational Outlook Handbook counted about 275,600 data scientist jobs in 2025 and projects employment to grow 35% from 2025 to 2035 (accessed 28 October 2026). Its wage data exclude the self-employed, so they are not a guide to freelance rates, and we do not quote market rates because we found no dated primary source for them. Two practical points shape data science invoices. Compute is metered: AWS, for example, lets account owners tag resources and activate cost allocation tags so that costs are grouped by tag in a cost allocation report, which is the evidence for a pass-through line; AWS also advises against putting sensitive information in tags. Ownership needs paperwork: under US copyright law, a transfer of copyright requires a signed written instrument (17 U.S.C. § 204(a)), and commissioned work by a contractor is a work made for hire only in categories listed in 17 U.S.C. § 101, which do not name software. On tax, a few states tax data processing; Texas defines it in 34 Texas Administrative Code § 3.330. Checked 28 October 2026; general information, not legal or tax advice.

Worked example

Worked example (illustrative rates, not market data): a freelance data scientist builds a demand forecasting model for a regional grocery distributor under SOW-03. The SOW sets discovery at a fixed $4,000 (invoiced and paid in September), the model build at a fixed $18,000 (50% at kickoff, already paid, and 50% on acceptance of the validation report), deployment at a fixed $6,000 on go-live, and monitoring at $1,500 a month including 8 hours, with overage at $150 an hour. Because the client's cloud account was not ready, training runs in the consultant's own AWS account, with every resource tagged project=grocer-forecast. Invoice INV-0072, 31 October 2026, net 30, PO 88-412. Line 1: 'Model build, stage 2 of 2, 50% of $18,000, validation report v2 accepted 24 October', 1 × $9,000.00. Line 2: 'Deployment of forecast model v1.2 to client production, go-live 29 October', 1 × $6,000.00. Line 3: 'Cloud compute at cost, AWS, tag project=grocer-forecast, October cost allocation report', 1 × $612.48. Subtotal $15,612.48, tax 0% (consulting and model development are not taxed in the consultant's state; this varies), total $15,612.48. Notes: 'Code and model assigned to client on payment of this invoice, per SOW-03 clause 9.' The invoice never mentions forecast accuracy, store names or sales data. Invoice INV-0079, 30 November: monitoring retainer, November, 1 × $1,500.00; overage, retraining after a supplier change, 11 hours used against 8 included, 3 × $150.00 = $450.00; cloud compute at cost, November, 1 × $84.20. Total $2,034.20. Running compute in the client's own account from the start would have removed $696.68 of pass-through and the cash-flow gap that came with it.

Common mistakes to avoid

Putting results on the invoice

'Churn model, 91% recall on customer table' reveals findings and data to everyone who handles the invoice. Name the deliverable and version instead.

Carrying compute costs to the end

Billing three months of GPU training in one final invoice strains your cash and surprises the client. Pass compute through monthly, at cost, with the cost report.

Payment tied to a production metric

If payment waits for accuracy to hold in production, data drift can delay it indefinitely. Tie phase payments to accepted deliverables you can evidence.

No ownership clause

Without a signed assignment, the client may not own the code it paid for, which surfaces during audits or due diligence. Reference the clause on the final invoice.

Retainer hours without a count

A monitoring retainer that never shows hours used makes the first overage line look arbitrary. Show included and used hours every month.

Invoicing Tips

  • 1Name the deliverable by version and acceptance date ('Demand forecast model v1.2, validation report accepted 24 October'), not by its results. Accuracy figures and data descriptions belong in the report, not on a document that passes through accounts payable.
  • 2Bill compute at cost from a tagged cost report, and put the provider, project tag and month on the line. Attach the report summary, not raw usage logs.
  • 3Agree before the build starts who runs the cloud account. If it is yours, invoice compute monthly; if it is the client's, there is nothing to pass through.
  • 4Give the monitoring retainer its own line with hours included and used. Retraining after drift is the most likely overage, so state the overage rate in the contract.
  • 5Reference the contract clause on ownership of code and models in the notes. Software written by an independent contractor generally needs a signed written assignment to transfer to the client.
  • 6Keep exploratory work you did not agree to bill off the invoice, or agree it as a change before you do it. A surprise line for 'additional feature engineering' is the one that gets queried.

Related Resources

Frequently Asked Questions

How do data scientists bill clients?

Usually by phase: a fixed fee for discovery or feasibility, a fixed or capped fee for the model build paid on acceptance of the validation report, a fixed fee for deployment, and a monthly retainer for monitoring and retraining. Some engagements are hourly throughout. Compute and data costs are passed through at cost.

What should a data science invoice include?

Your details and the client's, an invoice number and date, the contract reference, the phase and deliverable with its acceptance date, the fee or hours times rate, any retainer hours included and used, compute and data costs at cost with their references, tax only where it applies, the total, due date, PO number and payment details.

How do I bill cloud compute costs to a client?

Tag the cloud resources for the project, take the month's cost from the provider's cost report, and invoice it at cost on its own line with the provider, tag and period. Mark it up only if the contract says so. If compute will be heavy, have the client run it in their own account instead.

What is the difference between a data scientist and a data analyst invoice?

A data analyst invoice usually bills recurring reporting, dashboards and ad hoc analysis, often on a monthly retainer. A data scientist invoice usually bills a modelling project in phases (discovery, build, deployment) followed by monitoring, with compute costs passed through. Many freelancers do both and use whichever structure matches the contract.

Should results or client data appear on the invoice?

No. Describe deliverables by name, version and acceptance date. Accounts payable, auditors and email systems see invoices, so keep personal data, confidential findings and model metrics in the report delivered to the project team.

Who owns the model and code a freelance data scientist builds?

It depends on the contract. In the United States, software written by an independent contractor is generally not a work made for hire, so ownership passes to the client only through a signed written assignment. Many contracts assign the code on final payment, which is worth noting on the last invoice.

Can I create a data science invoice in InvoiceQuickly?

Yes. Describe the phase, deliverable, hours and compute costs, and InvoiceQuickly drafts an editable invoice for you to check before saving. Free covers 5 invoices, 5 clients and 3 AI drafts in total. Pro, at $9 a month, adds unlimited invoices, email delivery, recurring invoices for a monitoring retainer, and payment links through your own Stripe or PayPal account.