Marketing Technology

Marketing used to run entirely by hand. The customer list lived in a spreadsheet on somebody's laptop, onboarding emails went out by copying addresses one row at a time, and ad results were read off the ad platform's screen and retyped into a separate document. With 200 names and one channel, that works fine. With 30,000 names and five channels, it stops working.

It stops in two separate ways. The first is volume. Say the team decides that people should get a nudge three days after signup. Somebody now has to open the list every morning, filter for rows where the signup date is exactly three days back, and send. Miss a day and nothing marks the gap. The 400 people who signed up that day simply never get the message, and two months later, when the team goes looking for why conversion is soft, that hole is not even on the list of suspects.

The second is fragmentation. One person exists as an anonymous click record in the ad platform, as an email address in the sending tool, and as a company name in a salesperson's notes. None of those three places holds the single story: this person came in from an ad, opened three emails, and asked about pricing last week. Without the story there is no way to decide what to send next, and the team falls back on sending the same message to everyone, which reads as noise to the person who already asked about pricing and as a cold pitch to the person who has never heard of the product.

The place this surfaces is usually a meeting about results. Pagedeck hit it in the quarter right after its Series A, when the team turned on several channels at once. Ravi asked where last month's 120 paid conversions came from. The ad platform claimed 90 of them. The email tool reported 70 people who had opened a campaign. Sofia said she had personally closed 40. That adds up to 200. What actually happened was 120, with three tools each counting the same people once. No tool could catch the double counting on its own, because no tool knew the other two existed.

Marketing technology is what happens when both problems get pushed onto machines. Repeating work gets written down once as a condition and runs every morning without anyone remembering it. Scattered records get stitched together per person so the same human is not counted three times. None of the individual tools is a new invention. What changed is that they started being treated as one connected set rather than as separate purchases.

The tools under this heading fall into five groups: collection, which makes behavior leave a record; analytics, which turns those records into something a person can read; data, which stitches scattered records into one person; execution, which actually sends messages to the resulting list; and experimentation, which splits an audience to compare two versions. These five are an ordered chain rather than a flat list. Empty collection means empty analytics screens, and unstitched data means messages going to the wrong people.

Collection is where records start. A tag manager puts the job of adding and removing tracking code onto a single screen. Ad platforms and analytics tools each want a small piece of JavaScript on the site, and if every one of those requires an engineering ticket and a deploy, connecting one channel takes two weeks. Install a tag manager once and a marketer adds the snippet from a tool screen and checks the result half an hour later. What changes is who does the installing, not whether installing is needed. Nothing is recorded that nobody decided to record. The real time in this group goes not into the thirty-minute setup but into the few days spent listing which buttons and screens deserve a record and agreeing on what each name means. If one person tags the signup button as a completed signup and another tags it as a started signup, every number downstream disagrees with every other number.

Analytics turns stored records into readable screens, and splits in two.

1. Web analytics — collects how visitors arrived, which pages they passed through, and where they left. This is the group that produces sentences like "4,000 people viewed the pricing page and 180 of them reached the signup screen."
2. Mobile measurement partner (MMP) — reconnects app installs to the ad that caused them. Unlike the web, an app install routes through an app store, and the origin information breaks along the way, so a separate tool exists to repair that link. When three campaigns run in the same week and installs rise by 1,200, dividing that 1,200 among the three is this tool's job.

Data is the group that stitches records into people, and it also holds two. A customer data platform (CDP) merges records scattered across tools onto a single person. The same human shows up as a browser cookie, as a device identifier in the app, and as an address in the email tool; the platform joins those into one line saying this person clicked an ad, installed the app, and opened a message. A customer relationship management (CRM) system holds the contact history people write down by hand — when someone called, what proposal went out, when the next meeting is. The difference is where the record comes from. The first fills itself from what users do on screen; the second fills only when a rep types something in. As a company grows, the first keeps growing on its own while the second starts thinning out whenever the people responsible get busy.

Execution is the group that actually speaks to the assembled list. Marketing automation lets a team draw a condition and a sequence once and then leaves it running. Send a guide to anyone who is three days past signup and has created zero documents, follow up two days later with a different subject line for anyone who did not open it, and stop entirely if there is still no response — all of that becomes shapes connected on a canvas. The morning list-filtering ritual disappears here.

Experimentation is the group that decides which of two versions is better. An A/B testing tool randomly splits incoming visitors, shows one half the current screen and the other half the changed one, and compares the outcomes. The reason a tool is needed has less to do with the splitting than with the verdict. Whether a gap between 3.1 percent and 3.4 percent counts as a real difference depends on how many people are in each group, and doing that arithmetic by hand usually goes wrong. The tool takes the group sizes and conversion counts, decides whether a gap that size could plausibly be chance, and says so plainly when the data is still too thin to call. Without it, teams tend to pick a winner off two days of numbers and then quietly revert that screen a month later.

Strung together, the chain reads: the tag manager leaves records, analytics turns them into screens, the customer data platform stitches them per person, automation messages the resulting list, and the testing tool compares two versions of that message. The hard part in practice is not choosing tools but keeping the connections alive. Six tools mean more than six links between them, and when one tool renames a field, a link goes quiet. A broken link does not raise an error — it shows up as a number that is slightly smaller than it should be, which is usually noticed weeks later. That is exactly why a dedicated role grew up around this work.

"You can add that straight from the tag manager. No need to wait on a deploy."

It can be handled today by a marketer instead of being queued behind the engineering schedule.

"That segment can't go to automation yet — it isn't wired into the data platform."

The team can identify the right people but the list has no path to the tool that sends messages.

"We pay for six tools and three of them are connected. The rest are each looking at their own numbers."

Fix the connections between what you already own before buying anything else.

At Pagedeck this comes up in a quarterly stack review. Ravi brings a single page showing the tools in use, and Mia and Theo go through it with him. The page has six boxes with arrows between them: solid arrows are connections where data actually moves, dotted ones are gaps nobody has closed. Last quarter the page had two dotted arrows. This quarter it has four, because two more tools got added.

Theo's questions never change. How many engineering days does it take to turn a dotted arrow solid, and who keeps that connection working afterward? The second question carries more weight. Wiring something up once is not the end of it, since both tools keep shipping their own updates. If Ravi says he will own it, that ownership takes half a day of his week, every week, and campaign work shrinks by exactly that much. It is not unusual for the meeting to end with a decision to cancel a subscription rather than connect it. The test Mia applies is simple: name the decision this tool has changed in the last quarter. A tool nobody can answer that question for is paying rent on a dotted arrow, and the honest move is to drop it and give the half day back to Ravi.

The most common assumption is that owning the tools means the marketing is handled. Product pages show full dashboards and tidy automation flows, so it looks like signing a contract makes that screen yours. What arrives is an empty screen. Which behaviors to record, how to group people, and what to say to each group are all decisions a person supplies. Defer them and keep buying, and six months later four unused subscriptions are billing monthly while each tool reports a different conversion count. By then even cleaning up is hard, because nobody remembers which tool is the source for which number.

The expectation that automation reduces work is half right. The morning list-filtering really does go away, so the feeling is earned. But different work moves into the empty slot: designing the conditions, checking that they behave as intended, and watching for connections that quietly dropped. Skipping the checking is what causes damage. A signup-date calculation that is off by one day can run for two months, sending 40,000 people their guide at the wrong moment, while every dashboard reads normal the whole time because the emails kept going out. The more of a flow that runs by itself, the more deliberately someone has to schedule time to open it and look.