Consider a multi-homing complementor who has an existing track record on an
incumbent platform (
i) and has additionally started to operate on an
entrant platform (
e). Reputation on both platforms is conceptualized as a rating score ranging between 1.0 and 5.0 stars (i.e., the most common system). The complementor has acquired a rating of
\({r}_{i}\) on the incumbent platform and may (or may not) have already acquired an on-site rating of
\({r}_{e}\) on the entrant platform. We consider the complementor’s signaling decision, that is, whether to import their rating
\({r}_{i}\) from the incumbent to the entrant platform. The stylized function
\(f\left({r}_{e},{r}_{i}\right)\) describes the relation between the complementor’s (on-site and imported) rating scores and their trustworthiness in the eyes of prospective consumers. To be able to account for their individual and interactive effects, the two ratings
\({r}_{e}\) and
\({r}_{i}\) are not conflated into a single score but displayed separately. This yields four conceptual cases: (1) no on-site rating exists and no rating is imported; (2) on-site rating
\({r}_{e}\) exists and no rating is imported; (3) no on-site rating exists and a rating
\({r}_{i}\) is imported; (4) on-site rating
\({r}_{e}\) exists and a rating
\({r}_{i}\) is imported. The “trust function”
\(f\left({r}_{e},{r}_{i}\right)\) can hence be formalized as
where the parameter vectors
\(\alpha ,\beta ,\gamma ,\delta\) capture the effects of the rating scores
\({r}_{e}\) and
\({r}_{i}\) on the complementor’s trustworthiness across the four cases (
\(j=\mathrm{1,2},\mathrm{3,4}\)).
3 For the first case (1), a single coefficient
\({\alpha }_{1}\) suffices. For the cases, in which
either an on-site rating (2)
or an imported rating (3) exists, a linear relationship captures the association between rating and trustworthiness. If both on-site and imported ratings exist (4), we allow for interaction between ratings (
\({\delta }_{4}\)). A pair-wise comparison of all four cases results in several decision boundaries depending on the availability of rating scores
\({r}_{e}\) and
\({r}_{i}\). First, to decide whether to import a rating of
\({r}_{i}\) if no on-site rating
\({r}_{e}\) is available, the complementor compares cases (1) and (3). Equating and solving for
\({r}_{i}\) yields the import threshold
$$r_{i} > r^{*} = \frac{{\alpha_{1} - \alpha_{3} }}{{\gamma_{3} }}.$$
(2)
Building on the results of extant literature on on-site reputation and the emerging work on reputation portability, we consider on-site and imported ratings as signals for the complementor’s trustworthiness from the consumer’s perspective. Indeed, as previous research has shown, “good” on-site reputation is an effective trust signal in online transactions (Dellarocas et al.
2009; Qiu et al.
2018; Tadelis
2016). Also, a high imported rating can facilitate consumer trust in the complementor (Otto et al.
2018; Teubner et al.
2020). Yet, existing work has neither considered the effectiveness of different rating
values, nor the interplay of on-site and imported ratings in promoting trust. Based on the overarching theoretical framing of (cross-platform) signaling and previous work, we argue that imported ratings operate similarly to on-site ratings in that higher imported rating scores will, ceteris paribus, yield higher trustworthiness. Importantly, however, this does not imply that importing a rating will always be beneficial compared to not displaying any rating at all.
Importantly, as captured by case (4), a complementor may not only have an imported rating but also an on-site rating. This raises the question of how the consumer will respond to the availability of two (potentially different) ratings. According to signaling theory, the availability of two ratings requires the consumer to assess two
different signals for the
same quality (i.e., trustworthiness). Both signals are meaningful in the sense that maintaining a good reputation on either platform is costly for the complementor. Hence, it would not be reasonable for the consumer to disregard either of the two signals. At the same time, it is the nature of signals that they are “inherently noisy” and serve as an
indicator for the signaled quality rather than as
proof (de Haan et al.
2011). Thus, in the presence of two signals (rather than one), either one renders the respective other more reliable. While it is not clear, ex ante, whether both signals will receive similar weighting by the consumer, we expect that an improvement in one rating has a – ceteris paribus – positive effect on the trust-building effect of the respective other. In other words, we hypothesize a positive interaction
\({\delta }_{4}\) between the on-site and imported rating scores.