Traffic allocation determines how much eligible traffic is assigned to the experiment treatment. A fifty-percent split gives the treatment and control equal intended shares. A smaller treatment split can reduce exposure while collecting less treatment data. It doesn't mean the treatment is inherently safer or that the raw conversion totals can be compared without considering exposure.
Keep allocation separate from the success metric. A treatment receiving less traffic can have fewer conversions even with a better conversion rate. Real experiment types may offer different assignment mechanisms and product constraints. The compact browser uses one percent split and a deterministic fictional readout. It doesn't model user-level assignment, cross-device identity or a real experiment's power calculation.