AI SaaS pricing research: test packaging before changing prices

Start with the pricing decision, not a price recommendation

A founder asks an AI system what their software should cost. The answer looks precise, but it hides several different decisions: what to charge for, which features belong together, how to meter usage and who should buy each package. Separate those questions before asking for a number.

For this illustrative example, imagine a reporting tool used by small marketing agencies. A price per seat may discourage inviting clients. A price per workspace may fit the agency’s work better, but make occasional users worry about paying for inactive projects. These are packaging hypotheses, not evidence that either model will sell.

Build a pricing evidence sheet

Collect dated competitor price pages, anonymized purchase objections, current plan usage and examples of the work customers complete. Record billing intervals and usage limits alongside advertised prices. Comparing an annual equivalent with a monthly charge can produce a misleading market comparison.

Keep separate columns for observation and interpretation. “Three prospects asked whether client viewers need paid seats” is an observation in a small sample. “All agencies reject seat pricing” is an unsupported generalization. Missing information should stay visible.

• Segment and buying situation: who approves the budget and when?
• Value delivered: which task becomes easier, faster or more dependable?
• Alternative: another tool, manual work or delaying the purchase.
• Commercial terms: currency, billing period, limits and cancellation conditions.
• Evidence quality: source, date, sample limitations and unanswered questions.

Compare packages under the same scenario

Write two package cards with equivalent detail. Keep the buyer, core outcome and billing period constant; change the packaging variable you want to investigate. Ask for possible confusion, adoption barriers and reasons a buyer might choose neither option.

In Mirror, frame this as a scenario exploration using a concise, non-sensitive brief. A simulated agency owner, operator and client can expose different concerns, but they do not become three independent customer interviews. Use the disagreements to improve your research questions.

A prompt for a pricing research brief

The requested output should be a research plan, not a demand curve. Replace the bracketed fields and include only information you are authorized to use.

Validate the assumption with real purchasing behavior

Interview qualified buyers about a recent purchase and the alternatives they considered. Ask what happened, who approved it and which trade-offs mattered before showing your proposed packages. Hypothetical enthusiasm is useful context, but it is weaker than an actual decision involving money and implementation effort.

If you run an offer test, make its terms clear and decide in advance what evidence would change your mind. Track qualified opportunities, accepted offers and objections by segment. Do not infer causation from a tiny before-and-after comparison where the audience and offer both changed.

Leave with a decision record you can revisit

Your final brief should name the preferred experiment, the assumption it tests, the owner and the review date. Preserve rejected alternatives and the reason for rejecting them. This prevents the next pricing meeting from starting with another unsupported number.

To explore your own packaging decision, open Mirror with one segment, two package cards and a short evidence summary. Use the report to prepare the next customer conversation. Review available plans if you need more simulation capacity; the software supports exploration, while buyer research establishes whether the offer works.

Common questions

Further reading