You are currently in step 1 Create and commit a prediction model of the Get started checklist. To identify high-value customers, you will first start generating the predictive model data, using the Prediction Models module in Spotler Predict.
For the Nordic Outdoor example, we will create a Customer Lifetime Value prediction model. Nordic Outdoor wants to use expensive channels, such as a proactive WhatsApp campaign and Direct Mail. To increase ROI, they only want to target customers with the highest predicted value.
- Step 1.1 Open Prediction Models and create a new model
- Step 1.2 Choose the Customer Lifetime Value template
- Step 1.3 Define your model information
- Step 1.4 Configure the Target Variable
- Step 1.5 Define the Target Group and commit the model
Step 1.1 Open Prediction Models and create a new model
Go to Prediction Models in the left-hand menu. This is where you create new and manage existing prediction models in your account. Click Create new Prediction Model to start with a new model.
Step 1.2 Choose the Customer Lifetime Value template
In this tour and our example for Nordic Outdoor, we will use a Customer Lifetime Value model, to predict the long-term value of customers based on past behaviour. This helps you prioritise customers for campaigns where you want to focus budget, attention or channel capacity on the people who are expected to bring the most value.
Select Customer Lifetime Value as the model template and click Continue.
⏩ Skip if ready
Already have an active Customer Lifetime Value model in your account? For example, one that was created by a colleague? You can continue with the next article: Get started: Turn prediction into a user segment.
If you still want to complete this step yourself, choose the Second Purchase Prediction template. The setup logic remains the same, but the prediction goal will be different. With a Second Purchase Prediction Model, the target group is restricted to one-time buyers in addition to the information from the CLV model. This can be valuable if you have a large number of One-Time Buyers, whom you wish to turn into Repeat Customers.
Step 1.3 Define your model information
Enter the basic information for your model.
- Name: choose a clear name that explains the prediction goal, for example Nordic Outdoor CLTV 14 Days.
- Labels: optionally add labels to keep your models organised.
- Description: optionally explain what this model will be used for.
-
Back test: choose if you want to create a model for a back test. A backtest offers the possibility to train a model on past data and get a forecast for a past point in time (as opposed to a live model). As the forecast is based on a point in time in the past, the model performance can be evaluated immediately.
Step 1.4 Configure the Target Variable
The Target Variable tells Spotler Predict what outcome it should learn to predict. In simple terms, it answers the question: What does success look like?
The Target Variable consists of two components:
- Temporal component: when the predicted behaviour should happen.
- Contextual component: which behaviour or event should be predicted.
Together, these settings determine how historical customer behaviour is interpreted, which transactions count as successful outcomes, and how the model learns to identify future high-value customers.
🎯 Nordic Outdoor example
Nordic Outdoor wants to identify customers who are most likely to generate value after receiving a proactive WhatsApp campaign and a Direct Mail catalogue. The Target Variable ensures the model learns from the right transactions and predicts behaviour within a timeframe that matches the campaign.
1. Configure the Offset
The Offset represents the time between selecting customers and the moment your marketing activity actually reaches them. This setting ensures that operational processing time is taken into account.
For example, Nordic Outdoor's Direct Mail campaign requires printing and postal delivery. In that case, an Offset of around 14 days would ensure the model only evaluates purchases that happen after customers have actually received the catalogue.
| Channel | Recommended Offset | Reason |
| 1 day | Customers receive the message almost immediately. | |
| 1 day | Messages are delivered instantly. | |
| Direct Mail | 14 days (or actual delivery delay) | Allows for printing, production and postal delivery. |
If there is no meaningful delay between audience selection and campaign delivery, use 1 day.
Why Offset is important
If the Offset is too short, purchases that happen before customers have actually received your campaign may incorrectly count as successful outcomes. This can distort model performance and teach the model patterns that are unrelated to campaign exposure.
Always align the Offset with the real operational timing of your campaign.
2. Configure the Prediction Period
The Prediction Period defines how long after the Offset the model should evaluate customer behaviour.
For example:
- Offset: 14 days
- Prediction Period: 30 days
In this scenario, the model checks whether the target behaviour occurs between Day 15 and Day 45. The Prediction Period determines the future time window for which purchase probability and expected revenue are forecasted.
| Prediction Period | Impact |
| Shorter period (e.g. 14 days) | Focuses on immediate purchasing behaviour and stricter conversion criteria. |
| Longer period (e.g. 60 days) | Captures more conversions but may be less directly related to campaign influence. |
A good rule of thumb is to align the Prediction Period with the time frame in which your campaign is expected to influence customer behaviour. For example, if a voucher expires after 30 days, a 30-day Prediction Period is usually a logical choice.
3. Configure the Transaction Filter
The Transaction Filter determines which transaction events are taken into account when calculating the Target Variable. For each order type, you define how it should influence the prediction:
- Positive: events the model should predict.
- Negative: events that offset positive events. Only the remaining net value counts towards the Target Variable.
- Neutral: events used to calculate variables but that do not directly influence the prediction target.
- Ignore: events that are excluded from the dataset before model preparation starts.
4. Configure the Target Filter
Use the Target Filter when you want the model to focus on a specific subset of transactions, for example:
- Predict purchase probability and expected revenue for a specific product category, such as T-shirts.
- Predict purchase probability and expected revenue for a specific combination, such as Nike T-shirts.
By narrowing the target, you can create more specialised prediction models that align closely with specific products, brands or campaign goals.
For the Nordic Outdoor example using the Customer Lifetime Value template, no further restrictions are necessary.
Step 1.5 Define the Target Group and commit the model
Continue to the Target Group step. Here, you can define which users should be included when training and applying the model:
- Number of purchases a customer made before today.
- Time passed since the last purchase.
- Country a customer lives in.
When no conditions are chosen, the model will by default be trained on all available customers. So only select a Target group when it is needed to segment your customer database for this use case.
Review your settings and when everything is ready, commit the model. Spotler Predict will now process the model. The model looks for patterns in historical behaviour, such as purchase frequency, recent purchases, average order value, product categories bought and purchase intervals.
📚 Need more guidance?
Your prediction model is now created, committed and available for segmentation. 👉 Continue to turn prediction into a user segment and convert the prediction output into a usable audience.