CRM Marketer

When every marketing hire is pointed at bringing in new people, nobody is talking to the ones already inside. The person running ads watches today's signups. The person running content watches monthly visitors. The 3,000 people who signed up and stopped coming back after two days belong to neither number, so nothing happens to them.

In a collaborative document product the gap looked like this. A thousand people sign up in a month and about three hundred ever come back. The addresses of the other seven hundred sit in the database, and no one has decided what those addresses should receive. The only mail going out was a blast to every registered user whenever a feature shipped, and each blast brought a wave of unsubscribes. Spending on acquisition climbed every month while the list already paid for quietly burned down.

Sending messages to customers you already have, in order to manage the relationship, is CRM marketing, and the CRM marketer owns that work. The audience is not someone who has never heard of the product but someone who has already been inside it once. Done well, that brings users back far more cheaply than acquisition does. Done badly, it costs you the customers you already had.

The day opens with yesterday's sends. Not just the open rate, the share of delivered mail that got opened, but what happened after the open — how many people actually came into the product, and how many unsubscribes the send produced. A 40 percent open rate that yields twenty visits means the subject line worked and nothing else did.

From there the work breaks down roughly like this.

1. Define the audience — write the recipients as a condition. "Signed up more than seven days ago and has not created a single document" is specific enough to produce a list. Dormant customers, the ones with no activity for a while, get pulled out as their own group at this step.
2. Write the message — each group needs a different one. People with no documents get a path to their first one; people working alone get a reason to invite a teammate. The same sentence sent to both groups reads as generic to each of them.
3. Sequence it — instead of one send, messages get bundled into a drip campaign, a set of messages spaced at fixed intervals. Day one, day three, and day seven each carry different mail, and creating a document stops the rest of the sequence.
4. Automate it — the sequence goes into a marketing automation tool, which fires messages when conditions are met. Then it gets tested with a real account, because one badly written condition sends the same person the same mail three times.
5. Run experiments — subject lines and send times get split into two versions with an A/B test so the difference is measured rather than argued about. Changing one thing at a time is what makes the result readable.

Each quarter brings its own campaigns: winning back customers who left, or nudging accounts toward a higher plan. The measure of the role isn't how much mail went out; it's retention, how reliably people who used the product keep coming back.

Almost nothing here can be done alone. The data that builds an audience, the channel that carries the message, and the judgment about whether it may be sent at all all sit with other people.

• Marketing operations — the closest working relationship is with marketing operations, who run the tools and the data plumbing. Whether "has not created a document" can actually be turned into a list gets settled here. Behavior nobody records cannot become a condition, so new campaigns often start with a request to log an event that isn't captured yet.
• Performance marketer — with the performance marketer running paid ads, the job is to keep audiences from overlapping. Paying for an ad aimed at someone who already signed up means buying the same person twice.
• Product manager — the split between what the product should say on screen and what belongs in a message gets negotiated constantly, since anything sent by mail is read a day late.
• Designer — mail templates and push notification copy get worked out together, line by line.
• Legal and compliance — marketing privacy rules, the constraints on using customer data for marketing, get checked before a send. Mailing an address that never opted in is the kind of mistake that comes with a fine, which is why the check happens before the send rather than after a complaint.

"Let's not blast this one. Narrow it to people at day seven who haven't started."

A tighter audience makes the message specific and keeps unsubscribes down.

"Opens went up and conversions didn't move, so all we changed was the subject line."

Getting someone to open and getting them to act are separate problems.

"This should be in the product, not an email. By mail they read it tomorrow."

Pick the channel to fit the moment.

At Pagedeck, Ravi was handling this alongside paid ads. People who hadn't created a document in their first week received exactly one email. He split it into three, spaced across the week, and set it to stop the moment someone created a document. Two months later the share of signups creating a first document had gone from 31 percent to 44 percent.

Unsubscribes went down over the same period. Feature announcements that used to go to the entire list now go only to the people the feature concerns.

The most common belief is that more sends produce more results. There is a stretch where volume and clicks really do rise together, which makes the belief easy to hold. But mail sent to uninterested people comes back as unsubscribes and spam complaints, and those degrade deliverability — the share of your mail that reaches an inbox at all. Then the messages that should have landed go to spam too. Repairing a damaged sending reputation takes months.

The second is seeing a CRM marketer as someone who writes email copy. The visible output is writing, so the reading is natural. Most of the hours actually go into deciding who receives something and when; the copy comes after that decision. Reverse the order and you get a beautifully written email sent to the wrong list. The same confusion shows up in hiring. A strong writer brought in without access to the data tools ends up stuck producing one blast a week, which is the job the role was created to replace.