34,351 HOBO/Onset live-chat sessions, January 2021 through June 2026. This is a read of what's actually in the data — where customers come from, what they ask, what already works well, and where the routing layer between the chat widget and a live person breaks down.
74,671 rows total, September 2012 – June 2026. Three distinct businesses share the export: BIO (LI-COR Biosciences reagents/antibodies, ends ~2022), HOBO/Onset (environmental monitoring hardware — routes to the "Onset" team), and InTemp (cold-chain monitoring — routes to the "InTemp" team, a related Onset product line).
The oldest record on file is Sept 20, 2012, a BIO support chat. BIO volume tapers to zero by 2022. This report focuses on Jan 2021 – Jun 2026 — the window where HOBO/Onset is the dominant, then only, business in the data.
Starting with what the data shows is working
Before anything about the routing layer, four things worth noting about the customers themselves.
of sessions include a real, specific question — not just a menu click. Customers regularly name exact model numbers (RX3000, U26, MX1101) and describe their actual setup, unprompted.
door-to-door average for the 13,092 sessions that reach a live person — bot greeting through chat end. Agent-engaged time alone, once transferred, averages 12.6 min.
Door-to-door avg: 17.4 min · Agent-engaged avg: 12.6 min · Agent response time avg: 53 sec
of volume lands Monday–Friday, peaking Tue–Thu — a clean, predictable B2B pattern. This is procurement teams and field engineers working during business hours, not random consumer traffic.
32,566 weekday sessions vs. 1,528 weekend
is the clearest single demand signal — 31.3% of all sessions, more than Indoor and Outdoor combined. Demand is concentrated, not scattered, which makes it easier to know where to invest first.
10,739 of 34,351 sessions tagged Water market
Engagement snapshot
All figures filtered to HOBO-era sessions (Jan 2021 – Jun 2026, n=34,351). Topic counts are keyword matches on customer-side text only; a session may match more than one topic.
Top topics in customer messages · 2021–2026
Market breakdown · 34,351 sessions
Sessions by day of week · every Mon–Sun, Jan 2021 – Jun 2026, summed
Session volume by calendar year · Jan 2021 – Jun 2026
* 2026 partial year, through June 15
Sessions by hour of day (approx. US Eastern) · every hour, Jan 2021 – Jun 2026, summed
Traffic plateaus 10am–1pm ET and stays meaningfully high into the evening — highlighted bars mark the window where a live agent is realistically reachable (see next section).
Why sessions dead-end
58.1% of sessions (19,966 of 34,351) end in "there are no agents available at the moment." We checked whether this tracks with staffing hours rather than a constant failure rate — it does, substantially.
| Window (approx. US Eastern) | Sessions | End in "no agents available" |
|---|---|---|
| ~8am – 6pm ET (likely staffed hours) | 28,460 | 52.3% |
| Outside that window (evenings, overnight, early morning) | 5,634 | 90.2% |
Two things are both true in the data: overnight and after-hours traffic is real (5,634 sessions, 16% of volume) and almost never reaches anyone — which is expected, since no one is staffed then. But even inside the staffed window, just over half of sessions still end the same way. That daytime half is the part staffing hours don't explain.
Patterns in the routing data
All figures from HOBO-era sessions (2021–2026, n=34,351).
Customers show up with real questions — the gap is what happens after, not what they bring
Dead-end sessions last 30 seconds; sessions with a person run 17.4 minutes — because the agent starts over each time
Water-market sessions are mostly existing owners, not new shoppers
When an agent promises a follow-up, the data doesn't show it landing before the customer calls back on their own
Real conversations
Verbatim excerpts from the export — nothing paraphrased or reconstructed.
A customer tries six different ways to get through — the bot repeats the identical line every time
This pattern — free text met with an identical fallback line — appears across the 21,000 sessions where no human is ever involved.
HOBO Bot
HOBO Bot
HOBO Bot
HOBO Bot
HOBO Bot
HOBO Bot
HOBO Bot
HOBO Bot
A genuinely good technical exchange, once a person is on the line
This is the case for the data: agents are good at this. The exchange below — identify the model, check the manual, confirm the channel numbering — is also exactly the kind of catalog-backed lookup that doesn't need to wait for a person.
HOBO Bot
Onset
Onset
Onset
Onset
Onset
A specific, well-handled sales question — resolved entirely by hand
Rachel names the exact product, describes a real deployment (Guam, cellular), and asks a precise technical question before buying. The agent has to escalate internally to get the answer. This is the profile behind the purchase-intent figures later in this report.
HOBO Bot
Onset
Onset
Onset
Onset
Checked directly against the data
This is about live agents, not the bot — the bot's dead-end message never promises a callback, it just points customers to a contact page. We searched human agent messages, mid-conversation, for explicit commitments — "I'll get back to you," "someone will contact you," "a rep will be in touch" — then checked whether the same customer's next recorded contact looks like a proactive follow-up or a customer calling back on their own.
| Promise | Next contact | Gap |
|---|---|---|
| Agent (Onset): "I will prepare and a rep will be in touch with your estimate" — May 2023 | Customer re-enters via bot menu, hits "no agents available" | 52 days |
| Agent (Onset, self-identified as "Kim"): "I will have someone call you this AM… My name is Kim" — Jan 2021 | Customer re-enters via bot menu, hits "no agents available" | 69 days |
| Agent (Onset): "Thank you, they will contact you" — Jan 2021 | Customer re-enters via bot menu, hits "no agents available" | 142 days |
What the buying-intent language shows
Customer messages that name a specific, currently-sold product and use forward buying language ("want to buy," "quote for," "interested in purchasing") — past purchases excluded.
unique sessions named a product and asked to buy it — every one handed off to email, none completed in-chat. (The table below sums to 363 product mentions, since a few sessions name more than one product.)
is the ceiling if all 352 had closed at full price — not a forecast, just "if everyone converted."
| Product | Sessions | List price | At full price | At 10% conversion |
|---|---|---|---|---|
| U26 (dissolved oxygen) | 46 | $1,499 | $68,954 | $6,895 |
| RX3000 | 55 | $1,079 | $59,345 | $5,935 |
| U20L (water level) | 56 | $425 | $23,800 | $2,380 |
| U24 (conductivity) | 21 | $1,075 | $22,575 | $2,258 |
| RX2100 (MicroRX) | 30 | $750 | $22,500 | $2,250 |
| MX1101 | 97 | $159 | $15,423 | $1,542 |
| MX2301A / MX2302A | 44 | $220 | $9,680 | $968 |
| CX402 | 14 | $163 | $2,282 | $228 |
| Total | 363 | — | $224,559 | $22,456 |
Rows sum to 363 product mentions across 352 unique sessions — a handful of sessions name more than one product.