Potential Challenge Ahead: Balance between a lot of (75%) small leads (XS, S) and a few large (value) leads (L, XL)

[NicoHambauer]

A potential upcoming challenge, as described during the sprint-09 retro, is the following:

• The AI model fits well and results in high predictive power for (extra-)small merchants
• Potentially it could be harder to accumulate more data that represents samples for larger merchants.
• Based on the opportunity costs we can create an optimal version of this?
• Andi will either send the "boundaries" for the revenue data or multiply it with some secret factor "alpha"