flowpeekresources / guides / Automations
Intent-based remarketing: how to use Flowpredict before the cart
You'll plan a workflow that uses purchase intent without waiting for a product to be added to the cart.
Short answer
The key answer
Flowpredict can trigger an automation once the intent score meets a threshold. For an identified contact with the right consent, Flowpeek passes on a supported action, such as a tag in GetResponse. You configure the follow-up communication in your own platform. A score of 70/100 doesn't mean a 70% chance of purchase.
- A workflow started by intent, with no cart required
- Contact identification as a condition for a marketing action
- Controlling tagging, frequency and the post-purchase exit
1. Decide what information an interested customer needs.
Remarketing means getting back in touch with someone who has come across your offer. Before the cart, they may still be choosing a category, comparing variants or checking out the store. Start by helping with that decision: product information, guidance on choosing or purchase terms. A high model score doesn't necessarily mean a discount is needed right away.
Write down which group you want to reach and what communication will make sense for it. Keep interest separate from identification: you can analyze an anonymous visit, but the prediction alone doesn't give you a recipient's address. A contact workflow requires a link to a contact in the platform and meeting the right consent conditions.
2. Choose a threshold, but check the profile context.
Flowpredict returns a relative score from 0 to 100. A higher score signals stronger intent as the model sees it, not a promise of a purchase. New interactions change the available context; on the next recalculation the score can go up or down. Don't expect an update immediately after every click, or a steady rise.
The profile shown has a score of 40, five visits, 23 page views and no cart. It's an authentic, separate profile at the moment the screenshot was taken. With an example threshold of 70, it wouldn't qualify for the workflow yet. We don't present it as an earlier stage of a higher-scoring profile or as proof of a campaign's effectiveness.
3. Check the platform connection and contact identification.
In Settings → Integrations you choose the connection with the platform you use. The Automations tab becomes available once a supported marketing platform is connected. If you have several connections, check which one you're building the workflow for. The available actions depend on the integration: a tag in one tool may correspond to a group or a profile property in another.
For the first check, use an agreed test contact with the right consent and a confirmed link to the store. A tag alone doesn't create a list of anonymous recipients. In this guide the example is GetResponse: Flowpeek can add or remove a tag, and you set up the follow-up message on the platform side. We don't assume every connection has identical capabilities.
4. Build one path from prediction to tag.
As the starting point, choose “Purchase propensity score (ML)” and the condition greater than or equal to your chosen threshold. For the exercise, that can be 70. This start doesn't require an earlier cart. Next, you can add a purchase condition within a set window to route contacts with and without a recorded purchase separately.
In the NO branch for purchase, add a tag, e.g. wysoka_intencja (high intent); in the YES branch, end this communication. Optionally, add a “Condition: product interest”: any product, type “Add to cart”, minimum 1, 7-day window. The NO branch means no cart was recorded in that range. This is a configuration of an existing condition, not a separate “no cart” block. It doesn't prove the customer never had a cart outside the available tracking.
5. Plan re-entry and the exit after purchase.
Choose a re-entry policy: once, again after a break or every time. Repetition should match the goal of the communication. The same high score persisting for several days doesn't mean the customer needs another message at every check. Sending limits in the marketing platform are set independently.
Prepare a separate workflow after a recorded purchase that removes the earlier intent tag. On the platform side, check that removing the tag actually stops the relevant communication. Tagging alone doesn't launch ads, set a budget or sync audiences with every ad system. Those connections require separate configuration.
6. Test both branches before you judge the result.
A workflow has a draft version and an activated one. Before activating, go through the connections, threshold, purchase window, action and re-entry. After a controlled launch, check the execution history and whether the tag appears in the target platform. The engine runs in cycles, so don't expect a reaction in the same second as the interaction in the store.
Test a contact that meets the conditions, a profile below the threshold and a post-purchase situation. Assess correct operation and the usefulness of the communication separately. More purchases in the high-score group don't by themselves prove the messages had an effect — that group already showed intent. To assess the effect, you need a properly planned comparison.
Common mistakes
70 points means a 70% chance of purchase
The score is a relative model assessment. You choose the threshold for your process and assess the effectiveness of your actions separately.
A tag replaces the whole campaign
Flowpeek passes a supported action to an identified contact. You set up the message, frequency and further integrations in the relevant tools.
For your next review
Checklist
- I know the communication goal and what the threshold means.
- I have an identified contact, the right consent and a connection.
- The workflow has a separate path after purchase.
- I've checked re-entry and the rules for stopping communication.
- I've confirmed the action history and the result in the platform.
Now your store
Map out a workflow without sending a campaign
- 1
Write down the ML trigger, the threshold and the purchase window.
- 2
Draw the YES/NO branches and name the action in each.
- 3
Describe three test cases and the expected tags.
From report to decision
Configuration example — not a campaign that was run
We assume an illustrative threshold of 70 and a one-day purchase window. These settings aren't a recommendation for every store.
- Score 75, no purchase in the window
- Tag wysoka_intencja (high intent) if the contact is identified and eligible
- Score 40
- No entry via the trigger with a threshold of 70
- Purchase after the tag is added
- A separate workflow removes the tag; the platform ends the relevant communication
You're confirming the rules and the flow of information. You don't attribute the effect of a message or a future purchase to the sample profiles.
Questions and answers
Can an automation start before the cart?
Yes. The prediction trigger doesn't require a product to be added to the cart. The follow-up action still depends on the other conditions, contact identification and consents.
Will Flowpeek send a message to an anonymous person?
The model score alone doesn't provide a contact address. An action sent to the platform requires a reliable link to a contact and the right permissions.
Will adding contact scoring points in GetResponse change Flowpredict?
No. The platform's scoring and Flowpredict's relative model score are separate pieces of information. They shouldn't be presented as the same metric.
From knowledge to report
Open the right module.
Flowpredict
Flowpredict scores every visitor's interest in buying on a scale of 0–100. Find people with a high score before they add a product to the cart, and pass them on to marketing automation.
Automations
Combine the Flowpredict score with the offer being viewed and the visit history. Automatically tag recognized contacts who have the right consent in your marketing tool, even before they add a product to the cart.
Flowprofile
A card for each recognized customer: visits, purchases and consents in one place. Check the context before you choose an audience or plan your communication.
Read the next guide.
All guidesAI / 13 min
Purchase prediction: how to read the Flowpredict score
Read profiles scoring 40, 71, 95 and after purchase. Translate the relative score into conditions and tags, spot pre-cart intent and check a small test.
Automations / 14 min
Abandoned cart automation, step by step
Set up the event, the delay and contact tagging. Tell pre-cart intent apart, test both branches and remove the earlier tag after a purchase.
Setup / 7 min
Setting up Flowpeek: how to prepare your store and sign off the setup
Gather your store details, check the first visit and order, agree on the data scope. A simple sign-off checklist — no self-installation needed.
Free · 30 minutes · online
GA4, your store and your ads show different numbers? We'll check it with you.
We compare one month of sales across your tools and point out three fixes to your measurement. No strings attached. What we check
30 days on your own data
Prepare a workflow for your consultation.
Bring the goal, threshold, conditions and three test cases. Together it's easier to work out how to use intent in your communication process.
- No fixed-term contract
- We set everything up for you in 48 h
- App servers in Germany (EU)
- Cancel anytime
30 days for PLN 0 · first payment after the trial · cancel anytime
See pricing after the trial