10.10.3 Processing steps
Gaia photometry is unevenly sampled and highly sparse, with an average of $14$ photometric measurements per star each year. A new implementation of the Boxfitting Least Squares algorithm (Kovács et al. 2002) called SparseBLS (Panahi and Zucker 2021a, b) was used, with increased detection efficiency and reduced run time. SparseBLS ran on the initial candidates, taking normalized $G$, ${G}_{\mathrm{BP}}$, and ${G}_{\mathrm{RP}}$ as input, and scanning periods in the range $[0.5,100]$ days with a frequency step of $\mathrm{\Delta}f={10}^{5}{\text{d}}^{1}$. We ranked the results by using the signal detection efficiency (SDE) (Kovács et al. 2002; Alcock et al. 2000a) statistic along with a score we call the transit signaltonoise ratio (TSNR):
$$\text{TSNR}=\frac{d}{{\sigma}_{\mathrm{oot}}}\sqrt{{N}_{\mathrm{it}}},$$ 
where $d$ is the transit_depth, ${\sigma}_{\mathrm{oot}}$ is the standard deviation of the outoftransit points, and ${N}_{\mathrm{it}}$ is the number of points intransit (num_in_transit). In order to find the most promising candidates we applied the following criteria:

1.
SDE$>6$,

2.
TSNR $>7.5$,

3.
transit_depth $$ mmag.
The last criterion was an attempt to avoid cases of eclipsing binaries, or Jovian exoplanets around M dwarfs, which usually have depths greater than $40\mathrm{mmag}$. The top candidates were then inspected visually.