the journey everyone owned a piece of, and nobody had seen whole

customers had to pass compliance before they saw any value

Salesmsg is a texting platform, but in the US a business can’t text its customers until it passes compliance: the rules a business has to sign up to and follow before carriers let it send texts, including getting each customer’s consent first.

Customers had to fill in linked forms, one depending on the other, and share their business details before the product had shown them any value.

the first message was easy; the real one was blocked

In any given month, about half of paying customers sent a two-way text. The journey rewards the first message and blocks the real one.

Customers who got help had staff do the setup and compliance work with them. The documented compliance model is a sales-assisted model. Self-serve has no equivalent. Where the product didn’t let customers do the job themselves, people filled the gap.

every team held a different picture

The product serves two kinds of people: a buyer persona, the person who decides to pay for the tool, and a user persona, the person whose company picked it for them.

Product, marketing, sales, onboarding, compliance and customer success each saw their own part of the journey, with their own metrics and tools. None of this is anyone’s fault. It is what happens when each team measures and messages its own part of the journey. Earlier attempts to map it had started, and never been finished.

the question nobody had asked

How do you get a business through long, demanding compliance before it has seen any value?

Three answers at sign-up (integration, country and team size) decided, once and invisibly, whether a customer would ever get an onboarding manager. The rest faced compliance alone.

personas first, because a journey needs someone to walk it

The journey was always the goal, and a journey needs someone to walk it. We started by interviewing the teams, and their answers didn’t agree: no two people described the same moment when a customer first sees value. So we built two proto-personas from marketing’s ideal customer profile (ICP, a description of the customers a company wants most) and from recorded calls, and checked those instead.

The product’s pace didn’t leave room to observe customers, but we had a rich library of recorded calls where staff helped customers in the app. We used those calls as evidence, and set up AI to respond the way those customers did. Most pain points came from there.

Proto-persona

The owner-operator

Runs sales for a small team with lots of leads

Lost a lead last week to a competitor who texted first. Signs up to close that gap by Friday, and gives up if a session doesn’t move them forward.

• A small team, lots of leads
• Lives in the CRM
• Starts the trial alone; someone else approves the money
• Skims instructions and clicks around when stuck

Their goals

• Text every new lead within minutes
• Keep every text in the CRM without extra work
• One shared inbox for the reps

What gets in the way

• Business checks before the first text, with no reason given
• A wait with no status, which looks like the tool is broken
• Fees that show up late
• Help that only comes as long text

How they meet compliance

Expects to be texting within the hour. The business checks are the first surprise, and with no reason given they read as red tape. A review that takes days with no status looks like the tool is broken.

from a strategy doc to a ranked list of problems

AI worked through the data; the judgment stayed with us

AI helped us work through a large amount of data and call transcripts. Judging which parts were real insights, and what to do with them, stayed with us as designers.

It also got things wrong when numbers had to be weighed. It worked from what was written down, and couldn’t know that a decision had changed in yesterday’s planning meeting. That context had to come from a person.

six findings, once the pieces sat side by side

• Activation counted a test text.

  Activation was counted when customers sent a test text to themselves, so the metric showed success before most of them reached the job they came for.

• Six people, six “aha” moments.

  We asked six people across the teams when a customer first sees the value. We got six different answers, and none of them was sending a first real message.

• Compliance failed outside the product.

  Most compliance rejections had nothing to do with the product itself. They came from the customer’s own website pages, and from business details that had to match official records.

• Five timelines.

  Customers were told how long approval would take in at least five different ways, from 24 hours to several weeks, depending on which email, banner or person they heard it from.

• Three tools, no shared stop.

  Three separate tools messaged new customers, each on its own fixed schedule. When we canceled a test account ourselves, onboarding emails kept arriving days later.

• A quiet trial end.

  The end of the trial was a quiet moment in the journey. No screen in the product had been designed for it yet.

approval came with no next step

When the last approval arrived, the email offered one button: log in. One customer replied:

“Thank you, Lets do this, so what is my next steps”

even the data needed a shared picture

From mid-May, roughly twice as many new customers reached compliance submission, and the cause wasn’t recorded anywhere.

what the map led to

The project stopped after this first phase, so we didn’t redesign onboarding ourselves. Two things came straight out of the map.

• Consent examples

  The in-app examples for setting up customer consent were out of date. The fix had been waiting a long time; after the map, it went into a sprint and is nearly shipped.

• The compliance redesign

  Salesmsg’s design manager is now redesigning compliance from scratch, starting from the map.

Three more issues were raised and handed over: the onboarding checklist some new customers never saw, messages that kept arriving after a customer canceled, and sign-up answers that onboarding never used.

what we handed over

The findings write-up in the team’s knowledge base, and the board with the journey map, the affinity map and the prioritization matrix.

• Share findings early

  Tie each one to a number the business already tracks. Research with people takes time, so the team needs to see value along the way.

• Bring the design system in from the start

  So ideas can turn into prototypes quickly.

• Use AI freely, and check everything

  Use AI to analyze data, and check every result before it goes anywhere. The judgment stays with you.

• Bring people back into the context

  Findings that make sense from inside the research can lose people who weren’t in it. Bring the business back into the context every time you share.