We analyzed the first word of tens of thousands of support tickets across multiple SaaS products. The distribution tells you something about how customers think when they reach out for help.

"I" is the most common first word at 38% of tickets. "I can't log in." "I was charged twice." "I need help with..." The customer is centering themselves in the problem. These tend to be individual issues with a specific person's account.

"My" comes in at 15%. "My account is locked." "My order hasn't arrived." "My team can't access the dashboard." Similar to "I" but with a possessive emphasis. These customers feel ownership over the thing that's broken.

"The" accounts for 8%. "The app is crashing." "The export function is broken." "The website shows an error." These are systemic reports. The customer isn't talking about their account. They're describing a product-level issue. These tickets are more likely to be bug reports than account-specific problems.

"How" makes up 7%. "How do I export data?" "How do I add a user?" "How does billing work?" These are information requests, not problems. They're the easiest to auto-resolve because they have standard answers.

"Why" is 5%. "Why was I charged?" "Why can't I log in?" "Why did you change the pricing?" "Why" tickets tend to have more emotional weight. The customer isn't just asking for information. They're expressing frustration or confusion about a decision or outcome.

"We" is 4%. "We need to migrate our data." "We're having trouble with the API." "We" signals a team problem, often from a B2B customer. These tickets tend to be higher value (the customer represents an organization) and more complex (the problem affects multiple people).

"Can" is 3%. "Can I get a refund?" "Can you help me with..." "Can you fix this?" These are permission or possibility questions, and they sometimes mask the real question. "Can I get a refund?" really means "I want a refund."

What This Tells You

The first word isn't just linguistics. It predicts the ticket type.

"I" and "My" tickets: account-specific, individual, usually addressable with account lookup and standard resolution. AI handles many of these well.

"The" tickets: product-level, potentially affecting multiple users. Route to technical support or engineering triage.

"How" tickets: information requests, highest deflection potential. Self-service or AI auto-response handles most.

"Why" tickets: emotionally charged, require empathy before information. Route to experienced agents.

"We" tickets: organizational, potentially high-value. Prioritize and consider the account value.

You could build a lightweight triage heuristic from just the first word. It wouldn't be as accurate as full intent classification (like Supp's 315-intent model), but for teams without AI, it's a free improvement over random queue assignment.

The Length Signal

First-word analysis is just the beginning. Ticket length itself is a signal.

Tickets under 20 words tend to be simple, direct requests. "I need to reset my password." These are your highest auto-resolution candidates.

Tickets of 20 to 100 words are typical support requests with enough context to investigate. "I was charged $49 on March 3rd but I cancelled my subscription on February 28th. Can you refund this?"

Tickets over 200 words are either detailed bug reports (good) or frustrated customers writing their whole story (needs empathy). The very long tickets (500+ words) are almost always from angry customers or from people explaining a complex technical issue. Both need a human.

The bimodal distribution of ticket length (many short, many long, fewer in the middle) maps to different support strategies: automate the short ones, investigate the medium ones, prioritize the long ones.

Applying This Without AI

If you don't have AI classification, a simple rule-based triage using first words and ticket length gives you a crude but useful prioritization:

"Why" + over 100 words = high priority, needs empathy "The" + mentions error/crash/broken = bug report, route to tech "How" + under 50 words = FAQ candidate, check knowledge base first "We" + any length = flag as potential B2B account

It's a proxy. Intent classification does this better by a wide margin. But for a team of 2 agents with no budget for AI tools, first-word triage is free and instant.