It's more useful to know exactly where Arbyn struggled than to assume everything went smoothly. This page collects the honest record: real questions it couldn't confidently answer, and any place your own rules or policies quietly contradict each other.
What it does
Three tabs cover three different things: Knowledge gaps are real customer questions Arbyn answered with low confidence or had to escalate because nothing in its knowledge covered them well. Rule conflicts are places your escalation rules, guardrails, and learned facts contradict each other. Data conflicts are places your written policy and your live Shopify settings disagree.
Where to find it
Knowledge → Knowledge gaps.
How it works
Knowledge gaps. A gap is a real conversation Arbyn escalated specifically because it couldn't find a grounded answer, not because of a guardrail, a setting, or something outside its control. For each one you see the actual question, what Arbyn understood the intent to be, its confidence right before it escalated, and (if it tried a lookup) how close the best match actually was. An open gap stays open until you resolve it or it reaches a resolved or closed state on its own; it doesn't quietly age out after a few weeks, because a question customers are still asking doesn't get less worth answering just because it's been sitting there a while. The Open gaps count above the list defaults to showing everything, all time, for exactly that reason: you can narrow it to the last 7, 30, or 90 days with the filter if you'd rather look at something more recent.
Five numbers sit above the list: Open gaps (the total, on whatever time range you've picked), Missing knowledge (the subset where nothing you've added came close, the ones a new document or FAQ would genuinely fix), Resolved this week, Avg confidence (how sure Arbyn was right before it gave up and escalated, lower means more uncertain), and Top gap intent (which kind of question, refund, shipping, "where's my order," and so on, shows up most). Together they tell you not just how many gaps exist but what kind of gap they mostly are, so you can spot a pattern (a specific product line generating repeated confusion, say) instead of reviewing one row at a time with no context.
Rule conflicts. Your escalation rules, guardrails, and any facts Arbyn has learned from your team can, in principle, point in different directions on the same situation. This tab surfaces those contradictions directly, for example, a learned fact suggesting one thing while an active escalation rule says another, so you can resolve the disagreement deliberately instead of it quietly picking whichever one happened to apply first.
Data conflicts. This tab compares what your written policy actually says against your live Shopify settings, catching the kind of drift that happens naturally over time: a return policy that says 30 days while your actual return window elsewhere is set to 14, for instance. Each finding shows both sides, which one currently wins (the effective precedence), and whether you've already marked it resolved or it's grown stale since you last looked.
Resolving a data conflict: your four choices. Each conflict row asks How should Arbyn resolve this? and gives you four ways to answer. This is the one place in the Knowledge Center where you're not just reviewing something, you're telling Arbyn which of two truths to speak from.
Make live data the priority. Arbyn answers from your live Shopify setting and treats the written policy as out of date. This is the option Arbyn itself suggests, because your live setting is what a customer will actually experience at checkout, and an answer that contradicts checkout is worse than an answer that contradicts a page.
Make written policy the priority. Arbyn answers from your written policy instead. Pick this when the policy is the deliberate promise and the live setting is the thing that's wrong, for instance a return window you fully intend to honour but haven't configured yet.
Accept both, present honestly. Arbyn doesn't pick a winner. It tells the customer what both say and doesn't pretend to a certainty it doesn't have. Genuinely useful where the two are both true in different circumstances, and far better than a confident answer that turns out to be half right.
Ignore this conflict. Stops it being flagged. Use it when the two aren't really in conflict and you don't want to keep seeing the row.
Until you choose, Arbyn uses the live value and offers to confirm. An unresolved conflict is never a silent coin-flip. Arbyn falls back to your live Shopify setting, since that's the one the customer meets at checkout, and it says so rather than asserting the number flatly. That's the safe default, but it isn't a substitute for a decision: the point of resolving a conflict is that you, not a fallback, decide which of your own two answers is the real one.
Why this all lives in one place. A knowledge gap, a rule conflict, and a data conflict are three different problems, but they share one thing: each is a place Arbyn's honest, current understanding doesn't fully line up with reality, either because nothing covers a question yet, or because two of your own sources disagree. Reviewing them together, rather than as three unrelated pages, is deliberately how this hub is built, and the summary count on the main Knowledge Center reflects the combined total across all three tabs, not just one.
Good to know
A knowledge gap is a real conversation, not a hypothetical. Every entry traces back to an actual customer question that genuinely came up.
Resolving a gap usually means adding to your knowledge, a missing FAQ entry, an updated policy, or a web source that would have let Arbyn answer that exact question confidently.
A rule conflict doesn't mean anything broke. It means two of your own configured rules could disagree on the same situation, worth resolving deliberately rather than leaving to whichever one happens to apply.
A data conflict points at real drift, usually a written policy that's fallen out of sync with a live setting elsewhere; the fix is almost always updating one side to match the other.
The count shown on the main Knowledge Center hub covers all three tabs together, so "no open gaps" there genuinely means all three, not just the first one.
FAQ
Does a knowledge gap mean Arbyn gave a customer a wrong answer? Not necessarily, low confidence usually means Arbyn either escalated the question honestly or gave its best answer while flagging internally that it wasn't fully sure. Either way, it's a signal worth reviewing and closing with better source material.
What's the difference between a rule conflict and a data conflict? A rule conflict is between your own configured rules and learned facts contradicting each other. A data conflict is between a written policy and your live Shopify settings disagreeing on the same fact.
How do I actually close a knowledge gap? Usually by adding the missing information, an FAQ entry, an updated policy, or a web source, that would let Arbyn answer that exact kind of question with real confidence next time.
Why does the main Knowledge Center hub sometimes disagree with what I remember seeing on this page? It shouldn't, the hub's count now pulls from the same single source as all three tabs on this page combined, specifically so the two surfaces can never independently drift apart.