Skip to main content

Churn, Conversion and pLTV/UA are the three most critical problems that most companies have.

Churn – your consumers/customers leave after a while for some unknown reason. You can speculate (hypothesis) but you are not sure why

Conversion – Your consumers/customers buy your products. You are happy and you think that you know why and what levers to pull, but no matter what you do, your numbers do not change dramatically or at all.

LTV and UA – Acquiring consumers/customers has a cost. You think you have a positive ROI on your campaigns, but you are not sure until 30, 60, 90, 180 days out when you might (or might not) know whom of those users converted and how much the spent. Some of those users did not spend enough to have a positive ROI, thus, your campaigns we far from profitable and not optimized.

Contact us today & Get a FREE Discovery Call! Start Here

Churn

  • Perform data exploration to understand your case for Churn – create different “why” hypotheses
  • Build a churn model (ML) that predicts who will churn in the next churn cycle – we need to determine what the cycle is (i.e.: in mobile games is 14 days)
  • Run a series of semi-blind validations to fine tune the model – minimum 75% accuracy
  • Run the model live and pick one of the “why” hypotheses and run a retention campaign against the churning consumers. Inspect the results
  • Asses the results of the retention campaign – what consumers that had been id’ed as churners did not churn
  • Pick a different hypothesis and try again
  • Hypothesis creation is a guided exercise and based on understanding your data and needs to happen continuously
  •  In the end, you will know why and how to prevent churn. It’s time to automate

Conversion

  • Perform data exploration to understand your case for Conversion – create different “why” hypotheses
  • Build a conversion model (ML) that predicts who will buy in the next cycle – we need to determine what the cycle is (i.e.: in mobile games is 7 days)
  • Run a series of semi-blind validations to fine tune the model – minimum 80% accuracy
  • Run the model live and pick one of the “why” hypotheses and run an acceleration campaign against the converting consumers. Inspect the results
  • Asses the results of the campaign – what consumers that had been id’ed converted sooner than predicted
  • Pick a different hypothesis and try again
  • Hypothesis creation is a guided exercise and based on understanding your data and needs to happen continuously
  • In the end, you will know why and how to accelerate your revenue. It’s time to automate

pLTV & UA

  • Build a predictive Life Time Value model to predict LTV 30, 60, 90 and 180 days out
  • Use historical data to do a semi-blind validation
  • Use historical campaign data to match the validation against channels and campaigns
  • Using the results above, build a model that predicts outcome based on selected parameters (revenue, engagement, retention, product selection, etc.) This is the basis for multi-armed-bandit models
  • Experiment and fine tune the model
  • Include message and creative optimization on your experimentation
  • Automate

FESSEX Consulting can help you move to a more effective methodology to manage and use your data to generate actionable insights and dramatically improve your operations.

Contact us today & Get a FREE Discovery Call! Start Here