Most onboarding flows are really just apology letters dressed up as tutorials.
The Problem I Kept Ignoring
I watched seventeen user sessions in a row one Tuesday morning. Not skim-watching. Actually watching, with a coffee going cold next to me, as real people clicked into MentionFox, poked around for thirty seconds, and then either bounced or started doing something I had never intended. Seventeen sessions. Maybe four of those users found the feature they came for. The others just... drifted. They were not confused in an obvious way. No error messages, no broken flows. They were confused in the quiet way, the way you are when you walk into a room and forget why you came. The product was technically working. The product was also failing.
The Thesis That Changed How I Built
I had convinced myself for a long time that good UI was sufficient. If the interface was clean and the copy was clear, users would figure it out. That belief is expensive. It costs you churn you never get to diagnose, because the user who drifts away at minute two does not file a support ticket. They just disappear. What I started to understand from those seventeen sessions is that there is a gap between "knowing what a feature does" and "knowing where to reach for it right now, in this moment, given what I am trying to accomplish." Documentation closes the first gap. It does almost nothing for the second.
The thesis I landed on was this: users do not need more information about the product. They need a guide who watches what they are doing and says "try this, here, now." That is a fundamentally different thing to build. And once I believed it, I could not un-believe it.
What I Actually Built and Why It Went Wrong First
The first version of what became FoxChat was embarrassing. I shipped a chatbot that sat in the bottom corner and answered questions about features. Classic FAQ wrapper with an LLM underneath. Users ignored it at roughly the rate they ignore cookie banners. I checked the interaction logs after three weeks. Almost no one opened it voluntarily. The people who did were already power users looking for keyboard shortcuts. The people who needed it most, the drifters from my session recordings, never touched it.
The failure taught me something specific. A reactive chat widget puts the burden on the user to know they are lost. But the whole problem is that lost users do not feel lost in a way they can articulate. They feel vaguely unsatisfied, like the product is not quite doing what they hoped. They do not think "I should ask the chatbot." They think "maybe this tool is not for me."
So I rebuilt it around a different premise. Instead of waiting for the user to ask, the AI needed to observe context and offer navigation. Not pop-up notifications. Not tooltips that appear on a timer. Actual context-aware suggestions tied to what the user was doing in that moment. If someone had just connected a monitoring topic and was sitting on the results page without taking any next action, the AI should notice that and offer a specific path forward. Not "here are our features." More like "you have fifteen new mentions from the last 24 hours, here is how to filter for the ones most likely to be buyer intent signals."
Measuring Whether It Actually Worked
I am allergic to vanity metrics at this point, so I set two specific numbers to watch. First: did time-to-first-meaningful-action drop? By meaningful action I mean something that would realistically lead to a use case completion - not clicking a menu, but actually running a search, saving a lead, or exporting something. Second: did the percentage of users reaching what I internally call "the aha moment" increase within the first session?
Before the new version of FoxChat was in place, median time to first meaningful action was around nine minutes for new users. That sounds fine until you remember that most people give a new B2B tool about three minutes of genuine patience before they start mentally composing their cancellation email. After we rolled out the context-aware guidance, that number came down to under four minutes. The aha moment rate in session one went from something I was embarrassed to say out loud to a number I am now willing to reference in a pitch deck. Not because the product changed. The product was identical. The guide changed.
There was also a failure mode I had not anticipated. Some users found the AI suggestions intrusive. I got feedback, direct feedback from a customer I respect, that it felt like the product was "hovering." I toned down the frequency and added a simple way to suppress suggestions for a session. Retention among the users who had complained actually improved after that adjustment. The lesson there is that guidance should feel like a knowledgeable colleague, not a surveillance system.
The Investor Research Angle I Did Not Expect
One place where this mattered in a way I had not planned for was the investor research use case. Some of our users come to MentionFox not to generate leads but to track signal about markets, competitors, or funding activity. That workflow is genuinely non-obvious on first contact. The steps are different, the filters are different, the outputs they care about are different from a sales-led user. Without guidance, those users consistently got stuck trying to use the lead-generation flow for something it was not designed for.
When the AI started recognizing the signals of an investor research session - the types of keywords being tracked, the sources being prioritized - it could route them toward the right configuration. That improved completion rates for that specific use case by a margin that would have taken me months to achieve through UI redesign alone.
What This Means If You Are Building Something
The practical version of everything I just said is this. If you are watching session recordings and seeing people fail quietly, your instinct is probably going to be "I need better onboarding docs" or "I need a product tour." Resist that. Docs and tours are static. Your users are not static. They come in with different intent, different context, different levels of patience. The guide has to be dynamic or it is just more noise they will learn to ignore.
Build something that watches and responds, not something that broadcasts. The bar is not "did the user receive information." The bar is "did the user take the action that was right for them, at the moment they were ready to take it." That is a harder bar. It is also the only one that shows up in your retention numbers.
If you want to see how FoxChat handles contextual guidance inside MentionFox - specifically how it surfaces next-step suggestions based on what a user is actively doing - the best way is to poke around the product directly. And if you are evaluating whether the platform fits your team, the MentionFox pricing page has the current plan breakdown without any contact-us-for-a-quote friction, because I find that friction annoying as a buyer and I refuse to do it to you.
If you found this useful, I write about solo-founder distribution, B2B SaaS, and what's actually working in the AI-search era over on my Substack (one post per week, no spam).
I'm building MentionFox - a B2B intelligence suite that combines brand mention tracking with AI-visibility (GEO) measurement, investor research, and outreach automation. There's a free tier and a 5-day trial of Pro at mentionfox.com/pricing.
Top comments (0)