There are three main reasons why our models would recommend reactivating a profile into your marketing list.
- Direct signal that that customer is back. We see an event in Klaviyo's event stream that indicates the person is interested in something. Website visit, email open, purchase, etc.
- Predict seasonality. We use historical data, lifecycle data, etc. - should our data tell us that next season/quarter these customers are likely to come back and interact in some way, our models will turn them back on now and warm them up.
- Samples - we intelligently sample, all the time. Our models will still have some probability of turning customers back on even if they've been unresponsive for a long time. It will always consider users in the user suppressed category as temporarily dormant to some degree and randomly sample when they should be probed to see if they are back in market.
We flip the approach email marketers use on its head. Often, email marketers will take some signal (an open, an active on site, etc.) and then use a rule of thumb (or analysis) to sweep everyone that falls within a threshold
Orita takes an opposite approach with The Orita Platform. We build models based on the conversation event(s) of interest as our starting point, and build a case for each metric, and combination of metrics.
We then do math - a lot of math - to predict the likelihood of that conversion event happening.
- The current state: “people who clicked in the last 90 days are likely to buy”
- The Orita Way: “people who bought were X likely to click in the last Y days”
When you invert the problem this way, what happens is that you can quickly figure, for a specific brand at a specific point in time, which signals are important; which signals are not important; and which signals are only important in the context of another signal.