JANUARY 2022 R&D UPDATE

With 2021 now officially behind us, only a few short weeks remain before the bees arrive in the orchards and the almond trees begin to bloom. We spent the past 10 months testing and re-testing model accuracy, building new features and optimizing old ones. With these improvements now in place, we’re ready to deploy Verifli for our fourth almond pollination season.

We achieved a major breakthrough in our hive strength model early this year. The model now exceeds our stringent accuracy target for the weather conditions and hive types most commonly found in almond pollination. Now that we’re confident in our accuracy, we will no longer inspect hives manually to double-check our results, as we have in previous years.

After getting accuracy where we need it for almond pollination, our R&D focus in 2021 was two-fold: boost accuracy in warmer weather and optimize efficiency so customers get results sooner. Here are some of the new things Verifli will feature this year:

Verifli dashboard

Returning Verifli users will notice a new look and feel on the Verifli dashboard. Our marketing team redesigned the graphics and created new branded icons. But that’s just on the surface. Users will notice faster load times thanks to optimizations implemented by our software team.

We’ve also added new metrics to the Verifli dashboard to help users monitor weather conditions and understand how they affect pollination. Two new metrics appear at the top of the summary page: average bees per hive and pollination score. Developed by our data science team, pollination score is a custom formula that indicates expected quality of pollination based on each user’s hive strength results and forecasted weather conditions.

The reporting page includes a new table where users can plot bee flight hours, windspeed, temperature and humidity over a custom time period. This table pulls unique location data for each user, so you’ll only see information relevant to your operation.

Product optimizations

In addition to the user-facing improvements, our software and data science teams spent time optimizing various the behind-the-scenes functions as well in 2021. Though these aren’t quite as flashy as the dashboard updates, they’re critical to our success as we begin introducing Verifli to new pollination events.

Report generation efficiency

Automating the manual steps in our image analysis pipeline was a key priority this year. In last year’s almond pollination season, our back-end teams struggled to match the pace of our image capture teams. Since we couldn’t process images as quickly as we were able to capture them last year, we automated several time-intensive steps to eliminate this bottleneck.

One of the few manual steps we chose not to automate is image quality control. This step helps to ensure that no “mistake images” appear on customer dashboards, and it’s an important step for delivering accurate hive strength results. Though we didn’t automate QC, we built a software platform to help our back-end teams review images efficiently. We expect this new tool to drastically increase the rate at which our teams can perform image QC.

Along with these pipeline improvements, we’ve also worked on increasing the efficiency of our image capture teams. Our software team investigated solutions to the camera connection issues that caused delays for imaging teams last year. They pushed an update to our mobile app this summer removing many of the bugs that created camera connection issues.

Model accuracy

After cracking the code on almond pollination model performance, our modelling focus in 2021 shifted toward improving accuracy on high-strength hives and in warm weather. To get the ball rolling, we sent our data collection team to some of the hottest places in the southwestern US to gather hive strength data. We’ve got a few gaps to fill with more data collection this year, but early model results are encouraging.

Final thoughts

2021 was a year of big accomplishments for our R&D teams. Our core focus was getting the product ready to support significantly greater scale, and we feel good about what we’ve achieved. We’ll put everything to the test in 2022 as we aim to introduce Verifli to new growers and beekeepers around the globe.