Using App Downloads to predict SolarEdge Ltd.'s Quarterly Financials

Mobile Phone Application Data ("App Data") correlates to certain company financials and given its high (daily) measurement frequency, it can be used to regularly update and predict estimates for these metrics.


The following article using SolarEdge Ltd. as an example, explains:

  • how to identify relevant Apps

  • how to evaluate the quality of the App Data

  • how to predict financial metrics as stated in company quarterly reports using App Data

SolarEdge Ltd.

SolarEdge is a company active in the Solar Energy Industry. Their system is designed to maximize power generation of Solar Panel Systems. The SolarEdge Solution consists of:

  • power optimizers

  • inverters

  • communication and smart energy management solutions

  • cloud based monitoring platform

In brief SolarEdge's system allows users to optimize their Solar Power Generation System by attaching a SolarEdge Power Optimizer to each panel and connecting these to one SolarEdge Inverter. This setup automatically maximizes the power generation of the system.


Here is a brief view of SolarEdge Ltd.'s Income Statement.

SolarEdge’s Power Optimizers (POs) and Inverters Shipped Strongly Correlated to Revenue


The plots show the evolution over time of Items Shipped and Revenue.

Power Optimizers and Inverters Shipped as stated in SolarEdge's Quarterly Reports are 98% correlated to Total Revenue. A regression of Inverters to Revenue gives an R-Squared of 98%.

The plot shows the scatter plot of Total Revenue vs Power Optimizers Shipped (in orange) and Total Revenue vs Inverters Shipped (in green).

“SolarEdge Monitoring” is SolarEdge’s main App

Clients can set up their SolarEdge system using mobile apps published by SolarEdge. It has 4 main apps which are summarized in the table below (this info can be easily found on the Google Play Store or the Apple App Store):


In the analysis that follows we evaluate the SolarEdge Monitoring App as it has the highest number of both Downloads and Ratings. We use the Apple (iOS) version of the app has it more ratings than the Google Version (Android). In further analyses the Android version and other apps should also be included.



Evaluating App Data Quality for the “SolarEdge Monitoring” app

Apps need to be ranked in the public Apple Store Ranking for app data providers to be able to estimate downloads. Thus we need to know how often the app is ranked in the period we are interested in.

To understand this we first purchase historical app rankings data for the Apple SolarEdge Monitoring App in the US app store (this is because most of SolarEdge's clients are based in the USA):


The table shows the % of days the SolarEdge Monitoring App is ranked within each quarter.

The “% of days app is ranked” column indicates on how many days downloads data will be available. In this case on average 93% per quarter, so we move forward and purchase the app downloads data.


SolarEdge Monitoring iOS App Daily Downloads

Below we show how the downloads data looks:

In blue are shown the estimated daily downloads of the SolarEdge Monitoring App and in green the estimated precision of the downloads data. Downloads data is only available when the app is ranked in the App Store's public ranking.

The table below combines quarterly app and financial statements data:

The precision data (supplied by this data provider together with the downloads data) gives us a further indication of how good the download estimates are. In this case the average precision is 77% per quarter (note that the precision becomes higher in later quarters).

The data also shows that on average for every 16 Inverters shipped there is a Monitoring App download.


App Downloads are Highly Informative of Inverters Shipped

We finally analyze the usefulness of the App Data in predicting revenue via Inverters Shipped as this is the most natural connection. To do this we perform a regression of quarterly app downloads to quarterly Inverters shipped. The results are shown in the image below:


The left chart shows the evolution of quarterly app downloads and Inverters shipped over time. The plot on the right shows the scatter plot of App Downloads to Inverters shipped, a clear linear relationship is observable. The table below the right plot gives the results of the regression: it shows that App Downloads are very significant when trying to predict Inverters shipped (P-value 0.015%).

The Regressing of Quarterly Inverters Shipped to App Downloads is highly Informative:

  • R Squared is 94%

  • Coefficient of App Downloads is very significant (P-Value 0.015%)

  • the average error of the Fitted Model is 5.8%


This demonstrates that App Downloads can be used to Predict Inverters Shipped and thus Total Revenue on a Quarterly basis.

DDQIR: Alternative Data-Driven Quantitative Investment Research

© 2020 by DDQIR