Targeting sales incentives

2023-02-01

2023-04-30

Number of relevant bookings (i.e., bookings for the specific target destinations), revenue

Number of passengers per booking, various survey measures (e.g., sales agents‘ perception, transfer of voucher and commission payments to sales agents), gross margin

In treatment (ii) (to be described below), a travel voucher is sent to the agency owner and the owner is strongly encouraged to transfer it to the sales agents. We measure in a survey the owners’ and sales agents’ beliefs and perceptions about our travel company and its products and their booking behavior and strategy (e.g., bookings at other tour operators). We also measure in the survey whether the owner kept the voucher/commission for herself or shared the voucher with some or all sales agents in her agency.

The experiment will be conducted with a leading travel company and tour operator in Germany. The firm sells its products online but also through a large number of mostly independently owned travel agencies. These travel agencies work together with several tour operators and the respective tour operators pay the agencies through commissions typically set once a year. Common commission payments are defined as percentages of sales (where typically percentages increase in the overall sales volume of an agency).

We study whether top-up payments made for bookings to specific target destinations increase respective sales. We vary whether the top-up payment is paid to the agency owner or the team of sales agents (through travel vouchers).

More concretely, we compare the performance (as measured by revenue and the number of bookings) in a control group of unaffected agencies to two treatment groups where
(i) a top-up payment of 6 Euros for each passenger is paid to the agency owner (for every booking of at least 1,000 Euros to Egypt, Greece, Spain, or UAE),
(ii) a top-up payment of 6 Euros for each passenger is paid to the team of sales agents through a travel voucher (for every booking of at least 1,000 Euros to Egypt, Greece, Spain, or UAE).

We further study whether the top-up payments have an impact on the number of passengers per booking. In treatment (ii), the travel voucher is sent to the agency owner and the owner is strongly encouraged to transfer it to the sales agents. We conduct surveys, in which we ask, among other things, whether the voucher was passed on to the sales agents.

To understand underlying behavioral mechanisms we consider potential heterogenous treatment effects in several dimensions such as the agencies’ total number of past bookings at the study firm, their baseline commission levels paid by the study firm, the prior total number of bookings to the targeted destinations, the "tightness" with which the agencies are linked to other tour operators (e.g. because of competitors’ incentive contracts, ownership, long-term cooperation, premium access to competitors' products and services). We will also study heterogeneous treatment effects with respect to the existing incentives systems within the agencies (i.e., the incentives for employees implemented by the agency owner) and agencies' team characteristics (e.g., number of employees).

We conduct two other RCTs in collaboration with the travel company in the same time frame (see the respective pre-registrations for details on the other experiments). One of these RCTs studies whether better service quality affects the agencies' performance. The other one studies different sizes of top-up payments to the agency owners. The control group that we use in the current RCT will also be used in the two other RCTs. Furthermore, treatment group (i) will be used in the other RCT on top-up payments.

As described in the above, we run three RCTs with different research questions in the same time frame. As the three RCTs consist of altogether five different treatment arms (and the control group), we randomly assigned all target stores to one of six treatment groups. To do so, we ran a stratified randomization procedure using the randomize package in Stata. As the target group consists of agencies organized in six different chains, assignment was stratified by chain.

Store/Agency

No

We only include agencies with separate independent owners in the treatments (i.e. we do not include owners who own multiple agencies). Standard errors will therefore be clustered on the level of an agency which corresponds to clustering on the level of the owner.

In total, 1821 agencies were intended to participate in all trials we conduct in cooperation with the company. Although these travel agencies are mostly independent, they are part of larger buying syndicates. Our 1821 agencies belong to six different buying syndicates (e.g. chains). We drop in total 54 agencies from the trial, because they either have more than one owner (which would comprise non-independent observations) or the agencies are blocked for sales due to payment arrears. This leads us to a total number of 1767 valid observations for all trials. For this trial, we use 1008 observations and the remaining 759 are allocated to treatment groups of other trials.

252 sales agencies in treatment (i), 250 in treatment (ii), 506 in control