Signals the algorithm uses
- Failure type and reason code — a card-expired case needs the customer to act first; an insufficient-funds case needs to wait for the right moment to retry.
- Customer value, history, and behavior — a long-standing, high-value customer is handled differently than a brand-new subscriber.
- Available outreach channels — the strategy is constrained to channels you have connected and that have valid contact data for this customer.
- Customer’s timezone and country — every touchpoint is timed to local business hours.
- Payday patterns by country — for insufficient-funds failures, the algorithm accounts for when customers in that country typically receive their salary.
- Historical performance of similar cases — anonymized outcomes from comparable failures across all Airdun clients.
What the strategy determines
- Channels to use and their order — which channels are deployed, and in what sequence, based on predicted effectiveness for this customer and failure type.
- Number of touchpoints — how many messages and actions the sequence contains before the case is considered exhausted.
- Timing of each message — the delay between touchpoints, anchored to the customer’s timezone and any relevant payday or behavioral signals.
- Tone and content per touchpoint — early messages are gentler, escalation steps are more direct.
- Retry schedule — when Airdun attempts to charge the card again, coordinated with the outreach so the customer has had a chance to act.
Personalization per message
- Language — Airdun detects the customer’s language and writes in it.
- Timing — messages land during local business hours.
- Failure reason — an insufficient-funds message differs from a card-expired message.
- Customer segment — tenure and account value shift the tone.
The 1-click card update
Every message embeds the official Stripe invoice link, so the customer can update their payment details and settle the charge in a single click, without logging in or navigating menus.Communal learning
Every recovered payment — and every failed attempt — feeds back into the model. The algorithm learns what works for each failure type, country, customer segment, and channel combination, across the entire client base. All signals are anonymized; no identifiable customer data is used in any cross-client context.You do not configure the strategy — Airdun builds it. Your inputs are which channels are connected and, for high-value accounts, the escalation threshold. Configure that under Settings → Notifications → Escalation.