Four Programs, One Contact: Fixing Overlap Before Peak Season
Northwind Outfitters runs four Dotdigital programs that can all trigger for the same contact: a welcome series (3 emails over 10 days), an abandoned-browse program (fires 2 hours after a product view with no purchase), an abandoned-cart program (fires 1 hour after cart abandonment), and a weekly best-sellers campaign. Last quarter the average new contact received 11 emails in their first 14 days. Complaint rate rose from 0.09% to 0.31% and unsubscribe rate doubled to 0.44%. The abandoned-cart email has the highest revenue per send (£1.90) but also the highest complaint rate. Contacts who receive both the browse and cart emails within 24 hours convert at 1.1%, versus 2.6% for those who receive only the cart email. The team wants to add a fifth program for back-in-stock alerts, which would trigger on the same product-view data. Leadership wants a fix before peak season in six weeks, and engineering can ship only one data change in that window: a shared last-purchase and last-engagement field.
1. Diagnose what is driving the complaint and unsubscribe increase, then specify the priority rules and suppression you would implement. Justify the ordering using the data given.
Read the worked answer
Diagnosis. The problem is not volume alone; it is uncontrolled program overlap. Four programs read the same behavioural signal with no exit or suppression rules, so a contact who browses, adds to cart, and is also new receives the welcome, browse, cart and campaign emails in the same week. Eleven emails in 14 days is roughly 2.5x a sustainable cadence.
The 1.1% versus 2.6% conversion comparison is the decisive evidence. Contacts who get both browse and cart emails convert at less than half the rate of cart-only contacts. The browse email adds a weaker competing CTA at the moment of highest intent, and it splits attribution so the cart email looks worse than it is.
Priority rules.
- Transactional and purchase-triggered messages always send.
- Abandoned cart outranks browse. A cart event should exit the browse program and suppress browse for 14 days.
- Browse outranks the welcome series for the first 48 hours after a product view. The welcome series pauses and resumes 72 hours later with its remaining emails preserved, not skipped.
- Broadcast campaigns sit below all behavioural programs and are suppressed if the contact received any behavioural email in the previous 48 hours.
Frequency cap. Enforce at contact level, not program level: maximum 1 marketing email per 48 hours and 5 per rolling 7 days. Program-level caps are what allowed this to happen.
Do not launch back-in-stock yet. It triggers on the same product-view data and would be a fifth overlapping program. Ship suppression first, then add it behind the same rules.
2. The team proposes fixing the problem by cutting the welcome series from three emails to one. Evaluate that proposal against the alternatives.
Read the worked answer
It treats a symptom. Cutting the welcome series reduces volume, but the duplication pattern that pushed complaints to 0.31% would persist. A new contact who browses in week one would still receive browse, cart and a campaign email back to back.
It targets the wrong program. New subscribers are the least likely group to complain, because they opted in most recently, and they are in the highest-intent window a retailer gets. Deleting two onboarding emails removes revenue the overlap fix does not touch, while leaving the browse-versus-cart conflict intact.
Better sequence. Keep three welcome emails at day 0, day 3 and day 10, but gate them: any behavioural program enrollment pauses the series, which resumes 72 hours after the behavioural program exits. If the contact purchases, skip the remaining welcome emails and hand off to post-purchase.
How to prove it. Hold out 10% of new contacts on the current three-email series untouched, so you can separate the effect of suppression from the effect of the welcome content. If after four weeks the suppressed group's revenue per contact is flat or better and complaints have fallen, keep all three emails.
Spend the engineering change elsewhere. The shared last-purchase and last-engagement fields enable cross-program suppression, which is the structural fix. Cutting content is cheaper to ship but does not scale: you would face the same decision again when back-in-stock launches.
3. Which metrics go on the weekly dashboard for the six weeks before peak, and what would make you roll back?
Read the worked answer
Dashboard, weekly. Five metrics, each with a target:
- Complaint rate: target under 0.10%, hard ceiling 0.15%. This is the leading indicator and the one that gets a sending domain throttled.
- Unsubscribe rate: back toward 0.22%.
- Emails per contact per rolling 14 days: target under 6, ceiling 8.
- Revenue per contact (email revenue divided by contacts mailed), not revenue per send. Revenue per send rises automatically when you mail fewer, more engaged people, so it will flatter the change.
- Cart program conversion: baseline 2.6% for cart-only contacts.
Rollback triggers. Complaint rate above 0.15% for two consecutive weeks, or cart conversion below 2.0%, or revenue per contact down more than 8% against the pre-change four-week baseline. Any one triggers reverting suppression for the affected program while keeping the frequency cap.
Why not open rate or CTR. Open rate is inflated by machine opens and by exactly the subject lines that drove complaints. CTR can rise while revenue falls if the surviving emails are less relevant. The business question before peak is whether the same contact base generates more revenue at lower complaint cost.
One caveat. Six weeks is short. Put a 10% holdout on the welcome program from day one so you have a comparison if the numbers move against you and leadership asks whether the fix or the season is responsible.