Background. Tea is one of the most economically important crops in China. However, the tea geometrid (Ectropis obliqua), a serious leaf-feeding pest, causes significant damage to tea crops and reduces tea yield and quality. Spiders are the most dominant predatory enemies in the tea plantation ecosystem, which makes them potentially useful biological control agents of E. obliqua. These highlight the need for alternative pest control measures. Our previous studies have shown that tea saponin (TS) exerts insecticidal activity against lepidopteran pests. Here, we investigate whether TS represents a potentially new alternative insecticide with no harm to spiders. Methods. We investigated laboratory bioactivities and the field control properties of TS solution against E. obliqua. (i) A leaf-dip bioassay was used to evaluate the toxicity of TS to 3rd-instar E. obliqua larvae and effects of TS on the activities of enzymes glutathione-S-transferase (GST), acetylcholinesterase (AChE), carboxylesterase (CES) and peroxidase (POD) of 3rd-instar E. obliqua larvae in the laboratory. (ii) Topical application was used to measure the toxicity of 30% TS (w/v) and two chemical insecticides (10% bifenthrin EC and 50% diafenthiuron SC) to two species of spider, Ebrechtella tricuspidata and Evarcha albaria. (iii) Field trials were used to investigate the controlling efficacy of 30% TS against E. obliqua larvae and to classify the effect of TS to spiders in the tea plantation. Results. The toxicity of TS to 3rdinstar E. obliqua larvae occurred in a dose-dependent manner and the LC 50 was 164.32 mg/mL. Activities of the detoxifying-related enzymes, GST and POD, increased in 3rd-instar E. obliqua larvae, whereas AChE and CES were inhibited with time by treatment with TS. Mortalities of E. tricuspidata and E. albaria after 48 h with 30% TS treatment (16.67% and 20%, respectively) were significantly lower than those with 10% bifenthrin EC (80% and 73.33%, respectively) and 50% diafenthiuron EC (43.33% and 36.67%, respectively). The highest controlling efficacy of 30% TS was 77.02% at 5 d after treatment, which showed no