Budget for a pilot and measure what changes

What makes up the budget?

Separate one-time preparation from ongoing usage. Configuration, catalogue mapping and integration have different costs from analyses and support. Retail adds devices, installation and onsite operations. The quote should define a cap and approval of extra work.

A scope worksheet instead of a universal price

Planning example, not a quote: one market, one category, 50 variants, a limit of 500 analyses and four weeks of observation. Fill in setup, catalogue review, usage and support costs. Total budget is the sum of those items plus agreed hardware costs when needed.

Measure impact, not just clicks

Shoppers who choose an analysis may already be more likely to buy. Comparing them with everyone else does not establish an effect. Where traffic allows, randomly offer the entry point to some eligible visits and compare with a control group.

• Define the primary metric and purchase window before launch.
• Report starts, completions, errors, product visits and purchases.
• Keep incomplete analyses in the appropriate denominator.
• Treat small samples as directional evidence, without claiming a certain uplift.

Decide whether to expand

Agree continuation criteria before the pilot: suggestion quality, device behavior, support cost and shopping outcomes. Insufficient data may call for a longer observation period. Expand the catalogue or markets after reviewing the evidence.