Experimental Design Details
We will run a randomised field trial (RFT) in 2024 involving 13,233 domestic British consumers. Carried out in collaboration with Octopus Energy — an energy retailer in Great Britain (and other markets) — our RFT will be used to test how much it costs to switch Octopus' UK-based, EV-owning customers to a unique time-of-use (ToU) tariff combining managed charging with a static daily off-peak unit price for one's entire home (i.e., £0.075 per kilowatt hour [kWh] from 23:30-05:30 versus a region-specific static peak charge of approximately £0.30/kWh from 05:30-23:30). In so doing, we wish to determine these customers’ "price sensitivity" — i.e., their reaction to a service or product when its cost changes. Thus, we will assess customers' willingness to accept the Octopus-controlled charging of their EV via the "smart" ToU tariff in exchange for one of three randomly-assigned amounts of money (i.e., £0/Month, £5/Month, or £50/Month). Effectively subsidising uptake of Octopus-controlled charging, this money will be paid conditional upon switching to and remaining on the "smart" ToU tariff and, for customers in a fourth treatment group that will also receive £50/Month, conditional on "not" overriding supplier control (e.g., by immediately providing their EV with power or suspending Octopus telemetry). Following research on effectively engaging British EV owners, our incentives will be offered via emails emphasising general reduction in electricity expenditure and a theoretical £700 savings on charging costs due to the "smart" ToU tariff, where this figure is based on Octopus Energy's internal modelling of customers' domestic consumption and annual mileage. We will test the efficacy of this rhetorical aspect of our treatments using a £0/Month experimental group and a "pure" control group receiving neither email-based encouragement or pay.
We used rich pre-treatment information (e.g., historical electricity consumption) for the 13,233 Octopus Energy customers to actually assign these individuals to one of the five conditions (i.e., Control Group, Email + £0/Month, Email + £5/Month, Email + £50/Month, Email + £50/Month [No Supplier Override]) after grouping them based on their characteristics. That is, our random assignments were performed within “blocks” (i.e., strata) of customers with similar values for pre-treatment variables we expect to be predictive of potential outcomes. We used the R library “blockTools” (https://www.ryantmoore.org/html/software.blockTools.html), to construct blocks of twelve customers by minimising the multivariate distance between all possible pairs of customers given their values for relevant pre-treatment covariates. Blocks were constructed within the 14 electricity-supply regions of the United Kingdom. Thus, our blocks group customers who live in the same geographic area who have similar values for their pre-treatment covariates. Within-block treatment-group assignment was handled by “blockTools” automatically during block construction.
Note that, despite our five trial arms, we constructed blocks of size twelve instead of size five to accommodate our earmarking of different total numbers of customers to each condition (a cost-saving measure). Thus, we “overloaded” our blocks by placing multiple customers in
the control group and the £0/Month group. Specifically, in each block or “12-set”, two customers were assigned to the control group, seven were assigned to the £0/Month group, and one was assigned to the £5/Month, £50/Month, and £50/Month (No Override) groups.
Finally, our study is longitudinal, where each of the 13,233 Octopus Energy customers are tracked for 195 days.
Please see our pre-registered analysis plan for additional details.