Price Drop Alert

A price drop alert earns its place when wishlisted items get cheaper and sales barely move, because users only learn about the cut from a campaign email or by chance. Price change events need a consistent comparison baseline, wishlist and stock status must be known per product, and push permission has to sit on the same user identity; define the drop threshold up front so that small fluctuations never qualify. If prices move several times a day, simplify the pricing pattern first; turning every wobble into an alert wears users down instead of informing them.

Incremental purchase at the new price vs control

A price cut sells on its own; what is measured is how many extra purchases the pushed group produces compared with the control group, not how many clicks.

How it is counted: Among all user and product pairs assigned at the threshold crossing to the journey or control group, the share who buy the item within the same follow-up window from assignment; the gap between the groups is the journey's effect. Delivery or clicks on the push do not change the denominator.

Also track

• Time from alert to purchaseIn the journey group, track how long it takes from the alert to the purchase; confirm that the 48-hour check captures most of the purchases.
• Push clicksReport clicks separately as an engagement signal; a user who sees the new price and does not buy is telling you the drop was not deep enough, and this is not a success measure.
• Price events closed unmessagedPrice events that close without a push because the item is out of stock or no longer wishlisted, as a share of all events; a high share means stock and wishlist data lag behind the price feed.
• Push permission lossTrack users who opt out of push after the alert and read that loss next to the purchase gap; if one user receives price alerts for several items in a row, raise the threshold.

• Trigger and entryEntry happens when a wishlisted item's price falls past the drop threshold you set; movements below the threshold are not triggers. At the crossing, randomly assign users to the journey group or to a control group that gets no push for this price event.
• Who must not enterEach entry is one user and product pair at one price level, so no second entry opens at the same price. No alert is produced for an item that was already bought, taken off the list or is out of stock, and a user who never granted push permission stays out.
• Timing and channelsThe journey first confirms the item is in stock and still wishlisted; if so, one push showing the new price goes out, followed by a 48-hour wait. Recheck the current price right before sending and show the valid price at the destination; there is no further touch.
• Exit and handoffIf a purchase exists at the end of the 48-hour wait, the journey ends; if not, it closes without another message at the same price. An item that fails the pre-check also leaves unmessaged. No handoff exists; a later drop past the threshold is a new event.
• Frequency and consentOne push per price level and no second alert at the same price. Users without push permission do not enter; never promise that the price in an old notification will hold, and explain any offer conditions in the message.

1. 1If the control group buys at the same rate, users were already spotting the discount from their list; even with strong clicks, the journey added nothing. Do not scale the flow when the purchase gap fails to cover the rise in push opt-outs.
2. 2This finding holds for a single push, the 48-hour window and the drop threshold you chose; with no handoff, the purchase is credited to this price event alone. If another journey also messaged the same user about the same item, count the purchase once, not in both.
3. 3The higher the threshold, the rarer the entries. Turn the expected purchase rate and the smallest gap you would act on into a per-group entry count with the sample size calculator, then divide by your daily price events to fix the test length before you start.
4. 4If you see alerts going out on small price movements, the threshold is too low or the baseline inconsistent, and the result measures noise rather than the journey; correct the threshold and the comparison baseline and restart the measurement.