Your Customers Touch You 8 Times — You Only See 3

Last month, an old friend who works in marketing asked me to dinner. The moment we sat down, he slid his phone across the table to me. On screen was his quarterly review: the budget hadn't been cut, the ads still ran, the emails still went out, and the follow-ups were all handled — and yet the number of people placing orders wouldn't move.

He asked, "Guess which step I'm missing?"

I put it back to him: "So tell me — from the first time a customer hears your name to the day they finally hand over the money, how many times do they run into you?"

He thought for a moment. "Three or four."

"Three or four. And what about the other eight or nine?"

He went quiet. Those eight or nine — he'd never once seen them.

That whole dinner, this was exactly what I wanted to talk through.

First, Let's Be Clear on What a "Customer Journey" Is

In plain terms, a customer journey is the entire stretch of road a person travels from "hearing your name for the first time" to "spending money on you." Every time they cross your path along that road is one touchpoint.

That road used to be easy to draw.

You'd search a word, click into the website, fill in a form. Or you'd catch their ad in a social feed, tap through to the landing page, and get gently fed by an email drip.

Three to five touchpoints, all in a straight line — clean, regular, and easy to predict.

Now? Take a guess at a number: on average, how many times does a business customer run into you before placing an order?

Eight to twelve. Even a regular consumer buying something on the pricier side walks through 4 to 8.

But that sentence hides a catch. It refers to "only the handful that traditional tools can track." Just count every touch you can't see, and the real figure is at least 30% to 50% higher.

Why? Because your customers are meeting you in channels you can't see at all.

Customers Get to Know You Where You Can't See

Think back — where have your prospects actually been coming from lately? Of the entries below, have any of them ever made it into your reports?

AI search citations. When people want an answer now, the top of the results page just hands it to them. Those big models answer the question — and casually name-check your brand while they're at it. Google's AI Overviews, ChatGPT search, Perplexity — each can reference you several times in a single day. The customer has an impression of you forming before they've once clicked into your site. Every appearance lays down one more layer of awareness, while your analytics records a big, round zero.

Chatbots. The one on your website, the one on a partner's site, the one on a review platform. A customer arrives with a question, gets their answer, and walks away with more than they'd have gotten flipping through three landing pages. And then? Then they leave. Not a single click, no record anywhere. Traditional attribution can't see any of it.

Voice assistants. Siri, Alexa, Google search. A customer casually says "suggest a reliable brand," and your name floats past. Zero clicks, zero records — but "I've heard of you" has just been planted in their mind.

AI-orchestrated feeds. The sharpest social platforms now sort less by time and more by letting an algorithm decide exactly who this content is right for. Your content reaches the people most likely to respond, with zero ad spend, and your own tools count zero impressions from it. Yet along that stretch, that customer genuinely saw you.

Add all that together, and today it can shape about 35% to 40% of an enterprise customer's decision to buy.

And your attribution model reports a big, fat zero.

Attribution Deserves a New Grading Sheet

What is attribution, exactly?

In one line: it's the assignment of credit. When a deal closes, who gets the point?

Last-touch attribution hands all credit to that final touch. First-touch attribution runs it the other way, crediting the very first. Linear attribution divides credit equally among everyone involved.

That rulebook was written for an older world, where a sale stood on 3 to 5 touchpoints, clearly marked and easy to trace. Today a single order winds through 8 to 12, and about half of those you cannot observe at all. Scoring something you can't see isn't scoring — it's guessing.

A more dependable way is called data-driven attribution.

It doesn't draw some rigid "40% opening, 20% mid, 40% closing" template. Instead, it turns your thousands of real purchase paths into training data, and lets the model work out its own map of which combinations most often end in an order.

A quick example. In most places you'll find that a customer who reads one of your deep articles first, gets name-dropped into a single AI citation, and then finally opens one email — ends up converting at nearly three times the rate of the segment that only received the email.

That deep article, that one unpaid mention, both hold real bargaining value. Capturing that credit is exactly the work for data-driven attribution.

It does have a bar, so I'll spell it out plainly: every month, you need at least 500 orders before the model can sharpen its aim. Until your volume reaches that, fall back on position attribution (40% top, 20% middle, 40% end). It isn't perfect, but it's a whole world better than "last click gets everything."

Bring the Scattered Data Under One Roof

A lot of teams freeze in place — not because they lack data, but because the data lives in scattered corners.

Web analytics knows "who visited which page"; the email tool knows "who opened which email"; the CRM records the follow-ups; the chat tool guards a log of its own. Four datasets all describing the same person — yet switch systems and the customer becomes four different people, four names, four identities.

The work of stitching those fragments into a single card is called a Customer Data Platform, or CDP for short.

The name isn't magic. Think of it as one big warehouse that holds every customer profile you have, all in the same place.

A few players in the space:

• Segment: the strongest when it comes to plug-ins — it ships with well over 400 ready connectors, a match for engineering-led teams that like to build things themselves.
• HubSpot CDP: if your company already lives deep inside the HubSpot stack, this one is the least work — it may even save you most of the data wiring.
• mParticle: solid on enterprise data governance and privacy compliance.
• Salesforce Data Cloud: knits your Sales, Marketing, and Service clouds into one customer graph, and scouts a model named Einstein to predict the customer's next move.

Which is best? There's no standard answer. Look at what your own stack already uses and pick the one that fits naturally.

There is one hard rule: the CDP has to be locked tightly to your CRM and your automation flows. Otherwise all that warehouse stays forever a shelf of "how we saw them," with nothing it can (or does) steer in the future.

Run the whole thing smoothly, with a CDP plus AI to join identities, and cross-channel attribution can get about 30% sharper.

One Road, Four Segment, a Traffic Light at Each Turn

Data and tools sorted. Then what? Now it's time to hang some lights over the journey.

A customer journey splits into four sections, and each one mounts its own kind of automation.

The first, know them. Their touchpoints run AI citations, natural search, and voice recommendations. Just play out the moment they show even a flicker of interest — educational content flows to them, one piece after another. The instant your branded search volume starts climbing, that's the cue to hurry up.

The second section, hesitating. They're reading your blog, talking to your in-company bot firm, checking the comparison pages. At this point, don't load more puffery onto the screen. Hand them a version of your demo shaped for their situation, so they can actually see exactly what thing you can do for them. Once their overall score passes the line, the system routes the contact to sales.

The third section, buying. The buyer drifts across retargeting ad units, tailored emails, and pricing pages. Spot that kind of movement, push a time-limited offer out immediately, fire the note written exactly for them — and Sales lands the final nudge.

And the fourth, keep them. Thirty days of silence triggers the churn alarm. The moment they pass a usage milestone, "suggest add-ons" steps on, and periodic NPS surveys go out asking what could be improved.

These all sound very clean in descriptions. But the genuine minefield comes not across sections but where sections touch.

Which section bleeds the most people? Between section two and section three — over a single junction, you can lose as many as 7 out of 10.

Where exactly do those people lose track of you? In practice, it tends to come down to three things.

First, your material never answers the question they're holding. They already took you up against the other guy across the street, but you're still saying "this is who I am" while never answering the honest, direct question: "why choose you?"

Second, your calling cadence was wrong. Using one pace means they soften and cool; with the other, you just push things further out.

Third, the link cuts the instant they switch channel. They left your pricing page feeling the useful, and then get a drip mail, "you didn't finish the page you opened," and a huge chunk of that group simply walks off.

AI dropoff analysis does the detective work: it lines up the full behavior of the people who convert against the people who walk away, and tells you exactly which touchpoint, which single post, which instant you truly lost them.

Run for a quarter or two, even the "from hesitation to the sale" slice can fall a further 15 to 25 points.

Make that ledger strong, and every single touch point afterwards stands on hard ground.

Five Steps You Can Follow Directly

Here's what to do in practice. Walk the five steps in order, and skip no single one.

Step one: drag every touchpoint into daylight.

Website, social, email, ads, chatbot, plus your brand getting cited inside some AI answer or the recommendation handed over by a voice helper — put all of them on the table. Most teams jump when they finish: the true total runs 40% to 60% higher than the number they'd guessed in their gut.

Step two: wire up a CDP and treat everyone as the same person.

Pour all your systems in. Fold "the stranger who peeked at your homepage," "the registered returning member," and "the person who texted your bot" into one and the same customer. The proof is simple to run: pick out 100 older customers at random, match one order at a time, and only after every single one lands does the CDP pass the test.

Step three: let the data draw its own main lines.

Park that slide about "the funnel is a straight line" over in the deck, and let the data cluster on its own. Look at what comes out: are the real main trajectories only, say, 4 to 6? Chances are you'll be surprised to see, yes — and those 4 to 6 main lines are what produce 70% to 80% of the wins.

Step four: set the traffic lights.

Begin with the main line that brings the most of your traffic. Define clearly "what sign shows they're ready to step up a level?" and bolt automation onto it. Once that road gets smooth, go back and set the second and the third.

Step five: keep the numbers in view and adjust as you go.

Count from first touch to the close, how much credit each touch contributes, and how many people bleed out at each seam — collect the bills once a week. In an AI-era world, customers keep shifting; your journey map has no choice but to be a living, updating map.

Back to the Friend at the Start

That night, he left half-believing and half-doubting, and he wrote down, one by one on a sheet of paper, all those dozen or so "invisible touchpoints" he'd never counted before.

Two weeks later, he sent back a short message with just a few words: "It really did go up."

A single real regret of the marketer is drifting through this job, across that ordinary-looking stretch of the journey, and simply letting the customer pass by their side.

Start by knowing you have 8 to 12; only after that can you count your easy 3.

May you, a little earlier than the flock, go out and pull those invisible 9 touches back out of the place nobody else is looking, one by one.