THE USEFUL ANSWER
A revenue number becomes useful only when you know what was counted, over which period, for which accounts and against what comparison. A later sale alone does not establish that a chatbot caused it.
- Separate gross sales, attributed sales and contribution after changing costs.
- Look for changes in audience, pricing and workload alongside the tool change.
- Treat invented examples and vendor case studies differently from controlled evidence.
Ask what the headline actually measures
“Revenue increased” may refer to total account receipts, a subset of transactions attributed to a tool, gross sales before deductions or another defined metric. Those numbers answer different questions.
Write the metric, denominator, period and account scope next to the claim. If any of those are missing, label the claim incomplete rather than filling in a favourable interpretation.
The comparison worksheet separates claims from evidence. Keep that separation when the provider is familiar or when the number supports the option you already prefer.
Reconstruct the comparison period
| Question | Why it matters |
|---|---|
| Were the same accounts compared? | A different account mix can change the total |
| Were the periods equally long? | More days can produce more receipts without better performance |
| Did the audience change? | More eligible subscribers can explain more sales |
| Did pricing or offers change? | A higher transaction value may have another cause |
| Did staffing or coverage change? | Additional labour may contribute to the outcome |
| Were unusual campaigns included? | A temporary event may not represent steady operation |
A before-and-after example can be useful descriptive evidence. It does not automatically isolate the effect of the chatbot from all other changes.
Ask whether unsuccessful accounts or periods are included in the report. A collection of selected success stories cannot establish the typical result for every buyer.
Separate sales from contribution
Consider two entirely hypothetical scenarios. A produces $10,000 of sales with $2,000 of changing service costs, leaving $8,000 before other excluded costs. B produces $11,000 with $3,500 of changing costs, leaving $7,500 on the same limited basis.
B has 10% more sales, but $500 less contribution in this illustration. Neither figure is full business profit because other costs are excluded. The example shows why a sales uplift and an economic improvement are not interchangeable claims.
Use consistent cost definitions. Include the labour, variable fees and additional services that change between the setups. Keep shared costs and excluded costs visible rather than silently moving them between columns.
Examine attribution without assuming causation
A tool may assign sales to interactions using its own attribution rule. That can be useful for billing or operational reporting, but the rule does not necessarily establish what would have happened without the interaction.
Ask how the attribution window works, how overlapping human and automated work is treated, and whether the reported amount uses the same base as the invoice. A sale after a message may be associated with the message without being fully caused by it.
Do not accuse a provider of misreporting simply because its definition differs from yours. First establish the definition and whether the claim is presented with enough context to interpret it.
Turn a claim into a trial question
Instead of inserting a vendor’s uplift percentage directly into a forecast, ask what a bounded evaluation could observe in your own workflow. Define the scope, period, operational metrics and cost record before starting.
Keep the configuration and major contextual changes in the report. Where a controlled comparison is not practical, describe the result as observational and retain the uncertainty. Avoid presenting a small or selectively chosen trial as a universal success rate.
The comparison tool can organize requirements, but it does not convert a claim into verified performance evidence.
Use a decision range rather than a promise
Build a cost scenario that does not require unproven revenue growth to remain viable, then show what different outcomes would mean. State the threshold at which the choice should be reviewed.
Our Onlytool and Substy comparison and Infloww and Supercreator comparison apply this approach to public fee structures. Onlytool operates this publication; no product receives a demonstrated performance advantage merely because it is mentioned here.
A strong buying record explains the evidence, the assumptions and the next check. It leaves unsupported uplift claims out of the guaranteed part of the budget.
Sources & editorial notes
Primary references checked on 10 September 2026. Calculations and proposed workflows are our editorial examples, not independently observed provider results.