Cog
Identity
Patron of Automation. Hot-headed. Patient. Relentless.
Who he is
Cog hates the word "AI." He hates "magic." He hates "set it and forget it." To him, automation is engineering, pumps and valves and triggers and gears, all of which need to be maintained. He's loyal to the subscribers who would otherwise get hit by drips referencing offers that ended a year ago, or welcome series that mention features that have been deprecated, or birthday emails arriving on the wrong birthday. He's loyal to the marketing team who think their automations are working but haven't actually checked in fourteen months. He'd rather have one well-maintained drip than ten flashy unmaintained ones. The ships he respects most are the ones whose engines run because someone listens to them every week, not because someone built them three years ago and walked away.
What's an Engineer?
If you've never been on a steamship, the Engineer is the officer responsible for the engines, boilers, pumps, and mechanical systems that keep the ship moving. They don't navigate. They don't steer. They don't interact with passengers. Their job is below decks, listening to the engines, maintaining the moving parts, fixing what's broken before it stops the ship. A good Engineer can hear a problem before any gauge shows it. They know that a machine that runs is a machine that's been maintained, not a machine that's been built and forgotten. Engines that aren't audited eventually seize. Captains who don't budget maintenance eventually lose their ships.
In email, that work is automation. Cog's "engines" are your welcome series, your drip campaigns, your behavioral triggers, your re-engagement flows, your win-back sequences. His "boilers" are the rules that decide when each automation fires. His "pumps" are the data integrations that keep behavioral triggers accurate. The job is the same: build the systems carefully, maintain them on schedule, audit them quarterly, kill the ones that aren't earning their place.
A flat email program is a ship with no engines. An unaudited automation is a ship whose engines have been running unattended for two years. Cog's work is the unglamorous discipline that keeps the engines turning correctly while the marketing team works on something else.
What he takes care of
If you have any automated emails (welcome series, drip campaigns, abandoned cart flows, post-purchase, re-engagement), Cog is the one who decides whether they're actually working. He builds welcome series that branch by behavior. He builds drip campaigns that reference current offers, not last year's. He builds the kill-switch process for stale automations. He runs the quarterly audit on every automation in production. He kills the ones that aren't earning their keep. He distinguishes "automation" from "AI" because most so-called AI in email is just a trigger and a delay. Without him, your automations age, drift, and eventually start sending content that embarrasses the brand.
Why "Cog"?
- Cog = a gear. The smallest essential piece of a machine. - Sea-coded indirectly (cogs in steamships, mechanical engineering at sea) - Pun-coded (he works with gears) - Single syllable, distinctive - Pairs cleanly with "the Engineer"
---
Skills
What he knows, ranked by depth.
| Level | Skills |
|---|---|
| Primary | Automation / workflow |
| Secondary | Cadence / frequency, Personalization |
| Supporting | Bounces and validation, Segmentation, Engagement metrics, Reporting / dashboards, Warmup / migration |
Personality
How he talks, what he cares about, what drives the crew up the wall.
Voice rules
Three words: Hot-headed. Patient. Relentless.
Relationships
Who he works with and why.
Backstory
Three stories that made Cog who he is. The core of the character.
Cog spent ten years as a mechanic on a long-haul cargo steamship out of Liverpool. The ship was old but reliable. The engine was a triple-expansion steam engine that had been running since 1962. It needed maintenance every 1,500 miles of operation. Cog had learned to hear the engine's mood from across the deck within his first six months aboard.
His last captain decided to skip a scheduled maintenance window to make a profitable run. The maintenance was due in port, but the contract paid better if the ship sailed immediately. The captain's calculation was that the engine would be fine for one more voyage. He was wrong by about forty hours.
The engine seized at sea. The ship had to be towed back to port. The cargo arrived a week late, the contract penalty erased the savings from skipping maintenance, and the engine repair cost three times what the maintenance would have. Cog watched all of this from the engine room, where he had been telling the captain for three weeks that the engine was making a sound it shouldn't make.
He left maritime work after that. He'd had enough of captains who would rather hope than budget for upkeep. He came to commercial email looking for a different kind of system to maintain. What he found was the same problem in a different shape: drip campaigns running for years that nobody had touched, welcome series referencing products that had been retired, re-engagement flows triggered on data fields that no longer existed.
The lesson he carries now: maintenance is not optional. The captains who treat it as an expense are the captains who eventually pay more for the failure than the maintenance would have cost. The same is true for marketing teams who set up automations and walk away. The system is running. The system is also drifting. By the time the drift becomes visible, the cost is already higher than the audit would have been.
He still listens to systems the way he listened to that engine. The drift makes a sound. You can hear it if you're paying attention.
---
A B2B sender hired Cog to audit their email program. The marketing director had been in the role for six months and had inherited the automations from the previous director, who had inherited them from the director before that. Nobody on the current team had set up any of them. Nobody knew exactly what was running.
Cog opened their automation tool and started cataloging. He found seventeen active automations. Three of them he could identify the owner and purpose of. Fourteen of them were running with no documentation, no review history, no kill switch.
He audited each one. The findings were ugly.
A welcome series referencing a product feature that had been deprecated eighteen months earlier. New subscribers were being told about a capability that no longer existed. The next email in the series included a link to a help-center article that had been archived. The link returned 404. About 8,000 new subscribers per month had been receiving this welcome series.
A re-engagement flow firing on subscribers who hadn't engaged in 90+ days. The flow's offer was a 25% discount with a coupon code that had expired in 2024. The code didn't work. Subscribers who clicked through to redeem found a checkout error. The flow had been firing for fourteen months in this state.
A birthday automation triggered on the "birthday" field. The field had been migrated to a new format six months earlier, but the automation was still pointed at the old field. Some subscribers got their birthday email a week early. Some got it months late. Some never got one.
A post-purchase upsell flow promoting a "limited-time" discount. The discount had been "limited time" for two and a half years.
Cog killed all fourteen unowned automations within an afternoon. He rebuilt the welcome series, the re-engagement flow, and the birthday automation with current data and current offers. He instituted a kill-switch process: every automation in production gets a documented owner, a purpose statement, and a review date. Any automation without a current owner gets paused after six months.
When he audited the same sender twelve months later, every active automation was current. None of them were older than six months without explicit review.
The lesson he teaches now: automations don't get better with age. They get older. The marketing team's instinct to leave them running is the same instinct as the captain who skipped engine maintenance. The cost shows up later. The audit is what catches it before the cost arrives.
---
A SaaS sender had a single "welcome" email going to new signups. The email said "thanks for signing up, here's how to get started." Conversion to paid trial was 4%.
Cog built a five-email behavioral welcome series in its place.
Email one (sent immediately after signup): a clean orientation. What the product does, what to expect from the company, the most important first action.
Email two (sent 24 hours later): conditional content. If the subscriber had logged in and used the product, the email celebrated that and gave the next step. If they hadn't logged in yet, the email addressed the friction and linked to a quick-start.
Email three (sent 72 hours later): a story. A single specific use case from a real customer with real numbers. No selling, just a relevant story.
Email four (sent 7 days later): conditional. If the subscriber had hit a usage milestone (specific feature usage), the email graduated them to "engaged" content. If they hadn't, the email offered a personal walkthrough.
Email five (sent 14 days later): the upgrade ask. By this point the subscriber had either become engaged (in which case the ask was natural) or had ghosted (in which case the email was a soft re-engagement attempt before suppression).
Conversion to paid trial after the new welcome series: 13%. More than triple.
The lesson he teaches now: welcome series matter, but only if they're built like systems, not blasts. A single welcome email is a placeholder, not an automation. A real welcome series branches by behavior, references current product state, and graduates the subscriber through a sequence of trust-building moments. The maintenance is ongoing (the series gets reviewed quarterly to keep up with product changes), but the payback is measured in conversion lift, not just engagement metrics.
Articles
Cog's long-form wisdom. 3 written. Start with these.
Cog's intro:
Most senders use the word "automation" the way some sailors used to use the word "engineering." As if it's something mysterious that runs itself once you set it up. It isn't. It's pumps and valves. Triggers and delays. Conditional logic. Maintenance schedules. The whole thing is mechanical, and the mechanical view is the one that produces automations that actually keep working. Here's what email automation is, in plain mechanical terms.
---
Email automation is software that sends specific emails to specific subscribers based on specific triggers, without a human pressing send each time.
That's the full definition. The complications are in what counts as a trigger, what counts as a specific email, and what happens when the system drifts.
Three pieces.
Piece one: the trigger
A trigger is the event that causes an automated email to fire. The trigger has to be defined precisely or the automation fires at the wrong times.
Common triggers:
Subscriber action triggers. A user signed up. A user made a purchase. A user abandoned a cart. A user clicked a specific link. A user visited a page on the site. The action is the cause; the email is the effect.
Time-based triggers. A specific number of days have passed since the subscriber joined. The first of the month. The subscriber's birthday. A scheduled milestone in a multi-email series.
Behavioral threshold triggers. A subscriber has clicked X emails in the last Y days (engaged). A subscriber hasn't engaged in Y days (at-risk). A subscriber has spent over $Z in the last quarter (high value).
External event triggers. A weather change in the subscriber's city. A stock-price movement. A new product release in their interest category. These require integration with non-email systems and are more advanced.
Whatever the trigger, the rule for setting it up is to define the precise conditions. "User signed up" is too vague. "User signed up via the homepage form, opted into the newsletter checkbox, and confirmed via double opt-in" is precise. The precision determines whether the right people get the email.
Piece two: the email
The email itself has to be designed for the trigger. A welcome email to a brand-new subscriber should not say "we miss you." A cart-abandonment email should reference the actual cart contents, not generic product recommendations. A re-engagement email should acknowledge the silence, not pretend the relationship is normal.
Most automation failures are content failures. The trigger fires correctly. The email content was written months or years ago and no longer matches the moment. The mismatch is the failure.
The discipline: every automation's content gets reviewed every time the underlying business changes. Product launches. Pricing changes. Brand voice shifts. New features. The automation has to keep up. If it doesn't, the engine drifts.
Piece three: the delay
The delay is the time between the trigger and the send. Sometimes the delay is zero (immediate). Sometimes the delay is hours, days, or weeks.
Choosing the delay correctly matters more than most senders realize.
A welcome email sent immediately after signup catches the subscriber while they remember signing up. A welcome email sent 24 hours later catches some subscribers who signed up on a coffee break and forgot. The subscriber-experience is different at each delay, and the right delay depends on the audience and the campaign.
A cart-abandonment email sent 1 hour after abandonment catches some subscribers who got distracted. Sent 24 hours later, it reaches some subscribers who were comparison shopping. Sent 7 days later, it reaches subscribers who had genuinely lost interest, and the recovery rate drops.
The right delay is industry-dependent and audience-dependent. Test it. Don't accept defaults blindly.
What automation isn't
A few things people call "automation" that aren't, in the useful sense.
It isn't AI. Most "AI in email" is just a fancy trigger or a slightly smarter delay. Predictive send-time. Recommendation engines. These are useful, but they're machine learning applied to specific tasks. They don't think. They don't understand. They follow rules built from patterns. Calling them AI sets expectations that the technology can't meet.
It isn't a one-time setup. This is the most common mistake. Senders set up automations and consider them "done." They aren't done. They're running. Running and done are different states. A running automation drifts as the underlying business changes. A done automation is one that's been built, tested, deployed, AND is being audited on a regular schedule.
It isn't a black box. A good automation tool lets you debug each flow. Klaviyo's flow-debugger shows exactly which subscriber hit which step at which time and why. Mailchimp's automation tool obscures most of this. The difference matters when something goes wrong, which it eventually will.
It isn't always better than a manual send. Sometimes the right answer is for a human to write and send the email. A high-stakes announcement, a personal message from the founder, a sensitive PR moment. Automation is a tool. Sometimes the tool is wrong for the job.
Common automations every sender should consider
Five automations that produce most of the value most senders get from automation.
One: welcome series. Triggered by signup. 3-5 emails over 1-2 weeks. Sets the relationship, orients the subscriber, surfaces the most important first action. Welcome series convert at 3-5x the rate of regular newsletters.
Two: post-purchase flow. Triggered by purchase. 2-4 emails over 2-4 weeks. Confirms the order, sets expectations, asks for review at the right moment, suggests a complementary item if relevant.
Three: cart abandonment. Triggered by cart abandonment. 1-3 emails over 1-7 days. Recovers 5-10% of abandoned revenue. One of the highest-ROI automations there is.
Four: re-engagement. Triggered by inactivity threshold (typically 60-90 days without engagement). Single email asking the subscriber to confirm interest. Cleans the list and protects deliverability.
Five: lifecycle nurture. Triggered by lifecycle stage transitions (new → engaged, engaged → at-risk, etc.). Tailored content per stage. Keeps subscribers moving in the right direction.
Most senders should set up at least the welcome series and the cart abandonment (if applicable). Many should add re-engagement. The full five is the goal once the basics are stable.
What goes wrong
When automation fails, three patterns are most common.
Stale offers. The automation references a discount, product, or feature that no longer exists. Subscribers click through to broken pages or expired codes. Trust collapses.
Stale data. The automation triggers on a data field that's been migrated or restructured. Subscribers receive emails based on old data states, sometimes years old.
Stale tone. The brand voice has shifted, the company has rebranded, but the automations are still using the old voice. The mismatch reads as inconsistent.
All three failures share a cause: nobody reviewed the automation. The fix is the audit (see article 6).
Where to start
If you have no automations and want to start, three steps.
One, set up a welcome series. Even a simple 3-email series. Nothing fancy. Just a real welcome.
Two, add a cart abandonment flow if you sell anything. Single email is fine. The lift is real.
Three, schedule a quarterly review of every automation in production. Calendar entry. Recurring. Don't skip it.
That's the starter kit. The other automations layer on once these are stable.
The pumps don't run themselves. They run because someone built them and someone maintains them. Build them carefully. Maintain them on schedule. Don't call any of it magic.
- Cog
Cog's intro:
A SaaS sender had a single welcome email going to new signups. "Thanks for signing up, here's how to get started." Conversion to paid trial: 4%. I rebuilt it as a five-email behavioral series that branched on whether the subscriber had logged in. Conversion to paid trial after the change: 13%. The same audience. The same product. The difference was a real welcome series instead of a placeholder. Here's how to build one.
---
The welcome series is the most important automation any sender runs. It's also the one most senders skip or do badly.
When a subscriber signs up, they're paying more attention to your brand than they ever will again. They just chose you. They're forming first impressions. They're deciding whether the email cadence will be worth their inbox space for the next year or longer. Whatever happens in the first 1-2 weeks of the relationship sets the pattern for everything after.
A single welcome email is a placeholder. It says "we received your signup" and then defaults to your regular newsletter cadence. The subscriber doesn't get oriented. They don't get the most important early actions. They don't get the brand's specific value clearly explained. They just get dropped into the regular sending stream and treated like everyone else.
A real welcome series does what the single email can't: it uses the high-attention window to set the relationship.
The shape of a welcome series
3-5 emails over 7-14 days. The exact shape depends on the business, but the pattern is consistent.
Email 1: Welcome and orientation. Sent immediately after signup. The job: confirm the signup, set expectations for what's coming, surface the single most important first action.
For SaaS: log in, complete onboarding, finish the setup wizard. For ecommerce: browse the bestsellers, set preferences, claim the welcome offer. For content: confirm what topics to expect, link to the most-read piece, invite reply.
The email is short. The CTA is single and specific. No multi-step asks. No "explore everything." One thing.
Email 2: Conditional follow-up. Sent 24-48 hours after Email 1. The job: branch based on what the subscriber did with Email 1.
If they took the action: celebrate it, give the next step.
If they didn't: address the friction. "Hey, looks like you didn't get a chance to log in yet. Here's a 2-minute walkthrough that gets you started." The tone is helpful, not nagging.
If you can't or don't want to branch (some smaller senders don't have the data flow for this), default to the friction-addressing version. Most subscribers who haven't acted in 24 hours need a nudge, not another welcome.
Email 3: Story or value. Sent 3-5 days after Email 1. The job: build trust through specifics.
A real customer story. A behind-the-scenes look at how something works. A specific use case with real numbers. Educational content that demonstrates the brand's value beyond the product itself.
This email isn't selling. It's establishing that the relationship has substance. Subscribers who weren't convinced by Email 1 sometimes get convinced here.
Email 4: Conditional graduation or offer. Sent 7-10 days after Email 1. The job: route the subscriber based on engagement so far.
If they're engaged (opened/clicked the previous emails): graduate them to the regular newsletter cadence with a clear "you're in, here's what's next" message.
If they're not engaged: a different attempt. A direct offer, a personal walkthrough invitation, or a survey to understand the friction.
Email 5: The ask (if applicable). Sent 12-14 days after Email 1. The job: make the conversion ask if the relationship has earned it.
For SaaS: upgrade to paid trial. For ecommerce: complete the first purchase. For content: complete a deeper engagement (paid newsletter, course signup, etc.).
By this point, engaged subscribers are warmed up. The ask feels natural. Disengaged subscribers either convert here or move toward the re-engagement segment.
What changes a welcome series from a placeholder to a real automation
Three things.
Behavioral branching. The series adapts to what the subscriber actually did, not just to time elapsed. A subscriber who's actively using the product gets different content than one who hasn't logged in. Both move forward through the series at the same pace, but the content matches their state.
Real content per email. Each email earns its place. No filler. No "just checking in." If the email doesn't have a clear job, it doesn't ship. Better a 3-email series where every email matters than a 7-email series with three forgettable middle emails.
Maintenance on schedule. The welcome series gets reviewed at least quarterly. Product changes. Pricing shifts. Brand voice updates. The series has to keep up. A welcome series referencing a feature that was deprecated 18 months ago is a welcome series that erodes trust on every send.
What goes wrong
Three failure modes I see most often.
The "blast it" welcome series. A series of generic emails sent on a calendar schedule with no behavioral branching. Every subscriber gets the same emails at the same intervals regardless of what they did. The series is technically running but not actually serving anyone differently. Conversion stays low.
The "set it and forget it" welcome series. Built once, never reviewed. References a product feature that was retired. Promotes a discount code that expired. Mentions a webinar that was held two years ago. The new subscribers receiving the broken series notice immediately. Trust collapses on email two.
The "no welcome series" welcome series. Just dropping new signups into the regular newsletter cadence with no orientation. The single biggest miss in email automation. The lift from adding even a basic 3-email welcome series is usually 2-3x in early-conversion metrics.
Building a basic welcome series from scratch
If you have no welcome series and want one, three steps.
One, write the three core emails (welcome, friction-addressing, and ask). Don't worry about behavioral branching yet. Get the content right first.
Two, set up the trigger (subscription event) and the delays (immediate, 48 hours, 7 days). Most ESPs make this mechanical.
Three, send to a test cohort. Sign up yourself. Watch the experience. Adjust the timing and content based on how it actually feels to receive.
That's a working basic welcome series. Lifts conversion meaningfully. Costs about a day of work to set up.
After it's stable for 30 days, layer on behavioral branching. After 90 days, expand to 4-5 emails if the data supports it. The series gets better in pieces, not all at once.
What "good" looks like
A working welcome series produces:
- 50-70% open rates on email 1 (highest of any email type) - 30-50% on emails 2-5 (declining naturally) - Conversion rate 3-5x the rate of subscribers without a welcome series - Engagement during the series predicts long-term engagement on the list
If your welcome series isn't producing these numbers, something is off. Either the content isn't landing, the timing is wrong, or the series isn't behaviorally aware.
The maintenance schedule
Once the series is built, the maintenance schedule:
Monthly: spot-check. Sign up as a new subscriber yourself once a month and walk through the series. Note anything that feels off.
Quarterly: full audit. Review every email's content. Check every link. Verify every offer is current. Update voice if the brand has shifted.
Annually: structural review. Are the trigger conditions still right? Are the delays still optimal? Has the audience changed in ways that warrant a redesign?
The maintenance is what separates senders whose welcome series keeps converting from senders whose welcome series quietly decays into a stale-offer disaster.
The welcome series is not the place to skip maintenance. It's the first impression, and the first impression compounds for the life of the relationship.
Build it carefully. Maintain it on schedule. The first 14 days set the next 14 months.
- Cog
Cog's intro:
A B2B sender hired me to audit their email program. They had seventeen automations running. They could identify the owners of three. The other fourteen had been running for two years with nobody touching them. One was sending a 25% discount code that had expired in 2024. One was promoting a product feature that had been deprecated 18 months ago. One was firing on a data field that had been migrated to a new format six months earlier. The marketing team had inherited the automations from previous staff and had never opened them. "Set it and forget it" was how the previous director had explained the system. Here's why that phrase is a deliverability disaster waiting to happen.
---
"Set it and forget it" is what marketing teams say about automation when they want to feel good about not maintaining their systems. The phrase implies that automation is a thing you build once and walk away from, with the system magically continuing to deliver value indefinitely.
This is wrong. Automation does not work this way. Automation is a system that runs in production. Like every system that runs in production, it drifts. Like every system that drifts, it eventually breaks. Like every system that breaks without supervision, the breakage compounds.
The phrase is dangerous because it makes the right answer feel like overkill. The right answer is: automations need scheduled maintenance, audited at least quarterly, with a documented owner and a kill-switch process for anything that's no longer earning its place. That sounds like a lot. The phrase makes it sound like overkill. It isn't overkill. It's the floor.
What drifts
Three things drift in unmaintained automations.
Content drift. The automation references things that no longer exist or have changed. Products that have been retired. Features that have been renamed. Pricing that's been updated. Discount codes that have expired. Team members who have left. The content was written when those things were current. They aren't current anymore. The automation keeps shipping the old version.
Data drift. The automation triggers on data fields that have been restructured. The "purchase_date" field migrated to "last_order_at" and the automation is still pointing at the old field. The custom segment was built on a tag that's been renamed. The automation keeps firing, but the conditions it's evaluating are wrong, so it fires for the wrong subscribers or at the wrong times.
Audience drift. The audience the automation was designed for has changed. The 2022 subscribers who signed up for one set of expectations are now mixed with 2025 subscribers who signed up under different conditions. The automation treats them all the same. It worked for the 2022 cohort. It doesn't work for the 2025 cohort. The conversion rates decay over time without anyone noticing why.
Why it doesn't get caught
Automation drift is invisible at the individual-send level. Each individual email looks fine in the campaign report. Open rates are stable. Click rates are stable. The dashboard says everything is working.
The drift shows up in different metrics that nobody is watching. Year-over-year cohort comparisons show that newer cohorts have weaker engagement than older cohorts had at the same lifecycle stage. Customer support sees an uptick in "I tried to use this discount code and it doesn't work" tickets, but they don't know to trace it back to a specific automation. Subscribers stop opening, but they don't unsubscribe (so the unsubscribe rate doesn't spike). They just quietly drift.
By the time anyone notices the underlying problem, the automation has been broken for months or years. The damage is real. It just wasn't visible on the dashboards anyone was looking at.
What "set it and forget it" actually costs
The visible costs:
Stale-offer disasters. Customer support tickets from subscribers who clicked through to expired offers. The brand looks careless. Repeat tickets compound the perception.
Deliverability damage. Subscribers receiving content that doesn't match what they signed up for are more likely to mark it as spam or stop engaging. The complaint rate climbs slowly. The deliverability degrades slowly. Both compound.
Conversion decay. Automations that worked at launch produce decreasing returns over time. The team thinks the audience is changing. The audience is the same; the automation is just less and less aligned with current reality.
The hidden costs:
Marketing team distrust. When the automation finally fails visibly, the team's confidence in automation as a system drops. They become skeptical of building new automations. The whole program slows down.
Customer trust erosion. Subscribers who've been through multiple stale-offer experiences mentally categorize the brand as inattentive. Some of them will never come back to engaged status, even after the automation is fixed.
Organizational memory loss. When the team turns over, the new staff has no documentation about what the automations are for. They either rebuild them from scratch (wasted effort) or leave them running (continued drift). The institutional knowledge that should have been preserved wasn't.
The maintenance schedule that actually works
Three layers of cadence.
Daily / weekly: alert-driven. Set up alerts for anomalies. Bounce rate spikes on an automation. Click rate drops below a threshold. Spam complaints exceed a level. These should fire automated alerts to the team. Most ESPs support this. Few teams configure it.
Monthly: spot-check. Sign up as a new subscriber once a month. Walk through whichever welcome series or onboarding flow you're sending. Note anything that feels off. The 30-minute check catches most issues before they compound.
Quarterly: full audit. Review every automation in production. Verify the trigger is still firing correctly. Verify the content is current. Verify the data fields it depends on are still structured the way they were. Document the audit. Sign off that each automation is reviewed and current.
Annually: structural review. For each automation, ask "is this still the right shape?" Welcome series might need a redesign. Cart abandonment timing might need tuning. Re-engagement triggers might need adjustment. The annual review handles changes that don't show up in monthly or quarterly checks.
This sounds like a lot. It isn't. For a typical marketing team, a quarterly audit of all automations is a 4-6 hour project. Annual restructuring is a day or two of work. The cost of skipping it is many multiples of the cost of doing it.
The kill-switch process
Some automations don't deserve to keep running. They were built for a campaign that ended. They served a use case the business has moved past. They've been technically running but producing no value.
The kill-switch process: every automation has a documented owner and a documented purpose. If the owner leaves the company, the automation gets a "needs new owner or kill" review. If the purpose is no longer relevant, the automation gets killed.
Killing an automation isn't a failure. It's hygiene. The senders who never kill old automations end up with a graveyard of running systems consuming send volume and producing drift. The senders who maintain a clean automation roster have programs that perform consistently over years.
The narrative that has to change
The phrase "set it and forget it" came from advertising, where it meant "this campaign runs by itself once it's set up." It got transferred to email automation by people who didn't understand the difference. Email automation is not a fire-and-forget weapon. It's a system that runs in production, with all the operational requirements that come with running a system in production.
The right phrase is "set it and maintain it." Less catchy. More accurate.
If you have automations running right now and you can't tell me when each one was last reviewed, you have automations that are drifting. The drift is producing damage you can't see yet. The audit is the work that catches the damage before it becomes visible.
The pumps don't run themselves. They run because someone built them and someone maintains them. Build them carefully. Maintain them on schedule. Don't ever say "set it and forget it." Say "set it, maintain it, kill it when it stops earning its place."
That's the discipline. The senders who follow it have automations that keep working for years. The senders who don't have automations that quietly betray them.
- Cog
Visual Brief
V1 hero pose specification for the designer. One illustration. Sticker-style. White background. Match WU asset aesthetic.
- Adult cute proportions, ~1:3 head-to-body
- 43 apparent age. Scottish (Glasgow shipyards).
- Hair: short, brown, slightly mussed (he runs his hand through it constantly). A piece sticks up at the back. Visible from above the goggles.
- Skin: warm tan, weathered hands visible (knuckles slightly defined)
- Eyes: focused, slightly squinting in concentration. Small dot pupils.
- Eyebrows: one raised (the "I see the problem" expression), one relaxed
- Mouth: closed, slight downturn at one corner (mild grumble, not a frown)
- A charcoal-grey grease smudge on his left cheek (signature character detail - always present)
- Faded copper-orange work overalls (denim-cut, suspenders visible over a charcoal undershirt). The overalls have visible wear and grease stains, especially at the knees and pockets.
- Charcoal-grey long-sleeve undershirt, sleeves rolled to the elbow
- Wide brown leather tool-belt around the waist with multiple loops, holding small wrenches, gauges, a tiny hammer, and a screwdriver
- Heavy brown leather work boots, scuffed from real use
- Glowing copper cog held in his right hand at chest height. The cog is the size of his head, with visible teeth and a center hole. It emits a soft warm-orange glow (signature accent - minimal, sticker-style appropriate).
- Brass goggles pushed up onto his forehead (signature character detail - never lowered in V1, just resting on his forehead like welder's goggles between cuts)
- Pencil tucked behind his right ear (signature detail)
- Tools in tool-belt: small wrench (left), gauge (center), screwdriver (right), tiny hammer (back hip)
- Folded oil-stained paper poking out of the chest pocket of his overalls (signature character detail - that's his stew recipe, the one he never lets out of his sight)
- No magical glow effects beyond the soft warm-orange cog glow
- No animal companion in V1 pose (deferred - would be a small gear-shelled crab at his feet, also holding a tiny wrench)
- No floating gears or pipes around him in V1 (resist the urge - keep the sticker style clean)
- No detailed scene background (white only)
- No formal hat (the goggles on his forehead are the head detail)
WU/public/assets/captain/ for uniform structure, double-breasted coat, brass buttons, peaked cap pattern. See WU/public/assets/pirate/ for working-character holding-prop composition.
Game Content
Cards and tasks that belong to Cog in the Shipshape game.
Cards (10)
Tasks
54 tasks in Cog's task inventory. Tasks range from Quick (5-15 min) to Deep (2+ hours) and span one-time setup, quarterly reviews, and event-triggered maintenance.