Jepto

@elliot_anderson1 Thanks and a great question. No they are not, we researched a range of Algorithms in our testing and decided that the best option would be to take advantage of the Data Science community and utilise open source algorithms. They have the resources and community around them to improve over time. For the KPI predictions we are leveraging the Facebook Prophet algorithm, which has great accuracy and is considered the current standard for this type of real time prediction (You can select any Google Analytics metric and we will run the algorithm utilising 3 years of your own historical data to give you a live preview of the prediction) For our Anomaly Detection feature we are utilising Twitter's Seasonal Hybrid ESD algorithm. We also did a lot of research in this space and found that there was not one size fits all algorithm because different use cases will want different anomaly thresholds. So to overcome this we have exposed all of the inputs to the algorithm so you as the user can adjust the sensitivity as you seen fit. To make this understandable (as the algorithm hyperparameters can be confusing) the preview is also in real time and shows you how many anomalies would have been detected in your last 3 months of Google Analytics data so you can find the right fit for you. We are also working on adding influencing metrics to KPIs, so you can nominate a metric that affects your KPI metric to achieve a higher level of accuracy.

@errol_mathew Depending on what platform you use for call tracking, many of them utilise the offline measurement protocol in Google Analytics, you could definitely do that.

@dmac sweet! What would that look like on your RAD dashboard ?? - I’m going to look that up and get back to you.

@matsabrahamsen Thanks for flagging this. I will get our Dev team onto it straight away. You can use any domain to sign up and then edit it once the account is created.

@matsabrahamsen Just wanted to let you know that we have deployed an update to the validation on the domain for signup and you should be fine to create an account now. Sorry for the inconvenience.

@somecallmewoo Hi Sarah, Thanks for the kind words! Unfortunately the Call data is not available via the API. If Google does decide to add this to the API we will definitely look to include it.

@louis_meadors2 Thats right Louis. We us the term Marketing Automation and although it's normally used to described how contacts go through journeys and email/sms communication, we mean it in the sense that we are applying automation to the types of tasks that you need to do in order to run your marketing campaigns.

@jmedia9 Prediction algorithms do benefit from increased data and we have looked at a various ways of improving the accuracy and breadth of metrics that we currently offer, however there are some limitations. In general the best way to improve accuracy is by adding regressors to the prediction, which are highly correlated metrics that also have future prediction applied. As an example you could provide advertising spend as a regressor and provide more accurate projections, which would allow you to adjust the spend rates and get a projection to meet a certain client objectives. The short answer is yes, however we would start with the more generalised use cases of taking data that most Clients would have a history of (GMB, Search Console, Ads etc.) to be able to make the ML model work.

@dmac makes total sense on starting general and then honing in on specific use cases as I do appreciate the thoughtful response. Coming from the web design world, the data analytic piece to marketing is more and more important into how I can gleam real insights and turn those into actions I can implement with my clients in short order. Thanks again