Outage Postcards

Most marketing runs on the company's schedule. A newsletter goes out on Tuesday. A campaign launches in September. A sales list gets worked when somebody has time.

Some offers have a better clock.

For a standby-generator company, that clock is the weather. A household on a private well can lose more than lights during an outage. No power means no pump, which means no taps, no shower, no water for the heating system, and sometimes a frozen house if the owners are away.

I built a pipeline around that moment.

It watches public outage and weather feeds across the service area, then scores each event by type, severity, geography, and the number of relevant homes nearby. Severe thunderstorm warnings, high-wind warnings, winter storms, and major outages can all count. Broad watches and low-value alerts usually disappear before anyone sees them.

The audience is mapped before the storm arrives. Public well-completion records identify homes that likely depend on a private drinking-water well. Those records are cleaned into a small local database with address, town, coordinates, well type, drilling year, and source metadata. A campaign can filter to the exact segment: drinking-water wells, recent enough records to trust, inside the service area, outside the suppression list.

That preparation matters because the event window is short. After an outage clears the threshold, the system matches the outage geometry against the address pool, removes current customers and recent contacts, deduplicates the list, estimates the cost, assigns a tracking number, publishes a campaign landing page, and submits the mail batch.

No one needs to draw a polygon at midnight. No one needs to export a CSV while the phones are already ringing.

What Makes It Work

Weather data is loud. The system has to be good at silence.

Each alert runs through tiered rules. Critical events always surface. Major warnings need enough relevant homes in scope. Watch-level alerts need both scale and a narrow enough geography to be meaningful. The pipeline groups similar alerts by event type and calendar day, then applies a daily cap so one active storm line cannot create a pile of duplicate starts.

Every skipped event has a reason: wrong alert type, too few homes in scope, too many counties, already seen today. When the system acts, the run is auditable after the fact.

The postcard is generated per recipient. The greeting falls back safely when an assessor record looks like a trust, LLC, or business. Town names are formatted for natural copy. The QR code carries a recipient tag through a short redirect, so scans can be attributed without printing a brittle final URL.

Each campaign also gets its own local phone number. Calls forward normally, but inbound activity belongs to the campaign that produced it. The landing page publishes before the cards go out. Postcards are submitted with idempotency keys, so a retry skips duplicate cards for the same recipient.

Why This Pattern Matters

Direct mail has a useful shape here. It arrives after the immediate disruption, when people have had time to cool down and start thinking about prevention. It reaches the home as a physical reminder, with enough space to speak to the neighborhood and the specific problem the household just experienced.

The automation makes the message quieter, more specific, and better timed.

That operating pattern is the useful part. A real event occurs in the world. The system understands whether the event matters, who it matters to, what should happen next, and where the guardrails are. The visible output might be a postcard, a call list, a landing page, a repair reminder, a customer check-in, or a sales brief.

For businesses with local service areas, high-consideration purchases, seasonal risk, or a strong "I should really deal with this" moment, that kind of automation can feel less like marketing and more like good timing.