Six Systems, One Visitor: An Automated Lifecycle for a New Institution

The brief

Six systems each held a partial view of the same human being: web forms, ACME ticketing, ACME membership, Shopify, Clover point-of-sale, and DonorPerfect. None of them agreed on who anyone was.

That is survivable at a thousand contacts. At opening scale it produces the failure every institution recognizes: a $10,000 donor receiving a first-time-visitor welcome email, a member being asked to become a member, a family getting four versions of the same message because they exist four times.

One profile, permanent signals

Matching runs on email first, then source-system contact ID, with a manual review queue for anything ambiguous. The rule that makes it durable: source is a permanent tag that is never overwritten. Signals stack rather than replace. Someone who arrives as a shop customer and later becomes a member is both, forever, and both facts stay available to segmentation.

The named data-quality metric is match success rate — because a lifecycle program silently degrades when matching degrades, and nobody notices until the wrong person gets the wrong ask.

The priority hierarchy

The core design decision is what happens when a person qualifies for several journeys at once, which is most people. Nine levels, evaluated in order. The counterintuitive one is near the top:

Five journeys

• Soft Lead Nurture — 4 emails, for people who raised a hand but haven’t transacted.
• Visit Prep — 5 emails before, 6 after.
• Member — 14 emails across the membership year.
• Donor — 9 emails, tiered.
• Cross-cutting — 3 more that run alongside the others.

The lead-time gate

The detail I’m proudest of is that the pre-visit sequence compresses to fit the buyer instead of firing on a fixed calendar. Book five or more days out and you get the full five-touch cadence. Book two to four days out and it skips the T-5 email. Book same-day or next-day and T+0 merges with the T-24h content, leaning on a morning-of SMS instead.

A visitor who books tomorrow should not receive an email telling them what to pack next week. That sounds obvious and almost no automated program does it, because sequences are usually written as fixed timelines rather than as a function of how much runway you have.

Two membership asks, deliberately different

The post-visit sequence makes the membership case twice, in two registers. T+5 is tactical — apply the price of today’s ticket toward a membership. T+30 is narrative — come for the man, leave with a mission. Those land with different people, and sending only one of them means losing whichever half doesn’t respond to that frame.

Wealth screening that never blocks a send

Capacity screening runs as a nightly asynchronous batch, by design. Today’s email uses today’s capacity tags; tomorrow’s uses updated ones. Nothing waits on a third-party lookup, because a screening API having a slow night should never delay a visit-prep email.

It produces two flags for human review — member_prospect and upgrade_candidate — which go to a major gifts officer rather than triggering an automated ask. The argument for doing it at all:

Without ongoing screening, a high-capacity member sits in standard automation giving $150 a year when they could be giving thousands.

Guardrails

The rules that stop the program from doing damage, enforced platform-wide rather than per-sequence — which matters, because per-sequence limits are exactly how people end up receiving four emails from four journeys that each individually behaved:

• Maximum one email per 48 hours, and eight per 30 days, across everything.
• Tier 5 donors suppressed from all automation.
• Donors at $5,000+ never receive an automated ask below their lifetime average gift.
• New members protected from upgrade asks for 60 days.

That third one is the rule I’d put on a wall: the rule that prevents $100 asks from preventing $10,000 gifts. An automated ask that anchors low is not neutral — it actively costs money, and it does so invisibly.

250 miles

Member content splits on a distance band, on the premise that a substantial share of members will rarely or never visit. Inside 250 miles, membership is framed around return visits and events. Outside it, the digital benefits lead — because selling “free admission all year” to someone in Atlanta is selling something they cannot use, and they will notice.

Configuration, not engineering

All of it runs in Constant Contact rather than a custom build. The work is configuration, not engineering — which is the only responsible choice for an institution whose communications team is two people. A bespoke marketing-automation platform would have been more elegant and would have become unmaintainable the moment I left.

What I’d carry forward

• Decide the priority hierarchy before writing a single email. Everything downstream is an argument about precedence, and it’s cheaper to have that argument on paper.
• Never overwrite source. Signals should accumulate; the moment you replace one, you have destroyed a segment you’ll want later.
• Enforce frequency caps globally. Per-journey limits guarantee the failure they’re meant to prevent.
• Suppress your best relationships. The most valuable thing automation can do for a major donor is leave them alone.
• Fit the sequence to the runway. Lead time is a variable, not a constant.

Why it mattered

An institution’s email program is the only conversation it has with almost everyone who will ever care about it.

Most of the 150,000–200,000 people expected annually will visit once. What happens in the six weeks around that visit determines whether they become a member, a donor, an advocate, or a person who went to a museum once. That is worth designing properly, and it is worth designing before the doors open rather than after the first season’s data disappoints.

Related: the launch campaign that fills the funnel · the wearable that triggers the recap email · the opening that generated the volume