As of 13 August 2026, AI can only partly analyse your customers' purchase patterns.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Export the relevant order and customer data from your sales, ecommerce or CRM system into a spreadsheet, including an order ID, customer ID, order date, product or category, quantity, order value, refund or cancellation status, and any customer or channel fields you are entitled to use.
   2. Remove unnecessary names, email addresses, telephone numbers and other direct identifiers, then save a separate copy of the untouched export for comparison.
   3. Open the cleaned spreadsheet and record the date range, currency, VAT treatment, meaning of each column, treatment of refunds and cancellations, and the business question you want the analysis to answer.
   4. Upload the cleaned file to Julius AI or another approved AI data-analysis tool, paste the prompt, and replace each bracketed slot with the recorded details before asking for the analysis.
   5. Ask the tool to repeat any important totals and segment counts using a different grouping or calculation, then compare those results with pivot tables or formulas in the original spreadsheet.
   6. Check a sample of individual customers, orders, refunds and product totals against the source system, and remove any finding that depends on a data issue the tool identified.
   7. Send the final report to a colleague who understands your sales process, asking them to challenge the segment definitions, alternative explanations and proposed actions before you change pricing, stock or marketing.

   Prompt

   Analyse the attached customer purchase data for the period [START DATE] to [END DATE]. The columns are [DESCRIBE EACH COLUMN], and monetary values are in [CURRENCY] excluding or including VAT as follows: [STATE VAT TREATMENT].
   
   First, check the data for missing values, duplicate orders, inconsistent customer identifiers, cancelled or refunded orders, impossible dates, and unusual values. List every issue and do not silently correct anything.
   
   Then produce:
   1. total orders, customers and revenue, with the calculation method;
   2. repeat-purchase rate and customer purchase frequency, defining each measure clearly;
   3. customer segments based on recency, frequency and monetary value, with the rules used;
   4. product, category and basket-combination patterns;
   5. changes over time and any meaningful differences between segments;
   6. three practical actions for [BUSINESS GOAL], each linked to a specific finding;
   7. limitations, alternative explanations and analyses that the data cannot support.
   
   Separate observed results from interpretation. Do not claim that one factor caused another. Do not infer sensitive personal characteristics. Use only the supplied data, show the rows or calculations needed to reproduce important findings, and state when the data is insufficient. Return tables suitable for checking in a spreadsheet and a short plain-English report.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.