Onset / HOBO — Five Years of Live Chat

What five years of conversations tell us about Onset's customers

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.

34,351Sessions, 2021–2026
94.6%Type a real question, not just a click
38.1%Reach a live person
~19,600Unique email addresses seen overall 4,935 of those are linked to an existing CRM contact

What's in this file

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).

Where we drew the 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

Customers are specific, engaged, and easy to serve well

Before anything about the routing layer, four things worth noting about the customers themselves.

94.6%

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.

17.4 min

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

96%

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

Water

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

Where volume comes from — and what customers are asking about

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

Temperature Monitoring
2,856
Bluetooth / HOBOconnect
2,333
HOBOware Software
2,329
Setup & Install
1,970
Pricing / Quote
1,558
HOBOlink / Wi-Fi
1,041
Water Level Logger
926
Warranty / RMA
795

Market breakdown · 34,351 sessions

Water31.3%
Indoor14.4%
Outdoor9.7%
Env. Research7.9%
Agricultural4.1%
Other / unlabeled32.6%

Sessions by day of week · every Mon–Sun, Jan 2021 – Jun 2026, summed

Mon6,450
Tue7,183
Wed7,058
Thu6,645
Fri5,230
Sat865
Sun663

Session volume by calendar year · Jan 2021 – Jun 2026

20218.1k
20227.2k
20237.8k
20243.8k
20254.2k
2026*3.2k

* 2026 partial year, through June 15

Sessions by hour of day (approx. US Eastern) · every hour, Jan 2021 – Jun 2026, summed

12am3am6am9am 12pm3pm6pm9pm

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

Most dead ends happen when no one is staffed to answer

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)SessionsEnd in "no agents available"
~8am – 6pm ET (likely staffed hours)28,46052.3%
Outside that window (evenings, overnight, early morning)5,63490.2%

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

Important things the data reveals

All figures from HOBO-era sessions (2021–2026, n=34,351).

01

Customers show up with real questions — the gap is what happens after, not what they bring

02

Dead-end sessions last 30 seconds; sessions with a person run 17.4 minutes — because the agent starts over each time

03

Water-market sessions are mostly existing owners, not new shoppers

04

When an agent promises a follow-up, the data doesn't show it landing before the customer calls back on their own

Real conversations

Three sessions that tell the whole story

Verbatim excerpts from the export — nothing paraphrased or reconstructed.

Never gets a reply

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.

Nov 16, 2021 · 53 seconds, start to end · no human ever involved

HOBO Bot

Hi, I'm HOBO Bot, a digital assistant, please select a department.
Do I need to be constantly connected to log data?

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
jesus

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
tech support

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
real person

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
PERSON

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
Not you

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
I'm just gonna take a bat to these dumb things

HOBO Bot

Sorry, I'm still learning. I'm afraid I didn't understand that, can you please try again?
Handled well

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.

Apr 28, 2023 · 9.1 min · Technical Support

HOBO Bot

Hi, I'm HOBO Bot, a digital assistant, please select a department.
Technical Support
Transferred to Onset

Onset

Welcome to Onset. How can we help you today?
I was hoping you could help me clarify some of my monitor data
I have two MX1104 monitors set up with surface temp. attachments. The exported data shows two sets of temp. readings, but it isn't clear which is for the surface sensor and which is for the monitor itself

Onset

I'm looking now, just a moment
Thanks

Onset

I found this in notes appended at the end of the manual: For MX1104 loggers, an external sensor is channel 1
Ok! That may help
The only difference in the export data is that one set is labeled "SD-TEMP-P 21536407:21536407-1" and the other is labeled SD-TEMP "21536407:21536407-2"

Onset

*nod* the -1 is channel 1 and the -2 is channel 2. I was just looking at the readout from my own MX1104 to confirm
Amazing
Well, that does it for me. Thanks for your impeccable help

Onset

I'm glad i could help, have as much fun as you can for the rest of the day and a great weekend
Hey thanks! You too
Genuine buying research

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.

Mar 19, 2021 · 9.2 min · Sales Support · Outdoor market

HOBO Bot

Hi, I'm HOBO Bot, a digital assistant, please select a department.
Sales Support
Transferred to Onset

Onset

Welcome to Onset Computer Corporation. How can we help you today?
Hello, I am interested in purchasing a RX3000 to be used on the island of Guam (a U.S. territory). We would be using a cellular plan. I don't think Guam is on the AT&T network, but it is also not listed under Global Coverage.
So I just wanted to check that our best option is a User-supplied SIM card

Onset

Hello. Let me just ask this question of a sr person. One moment please

Onset

Yes, the best option is a user-supplied SIM card.
ok thank you!

Onset

Thank you

Checked directly against the data

When a human agent promises a follow-up, does it happen?

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.

What we found

950 sessions contain a specific promise, with an email on file. 341 of those customers have a later session we can look at. In every sampled case, that later session opens with the customer going back through the bot menu from scratch — not a message that references a prior callback. The remaining 609 promises don't reappear in the chat channel at all, which isn't proof either way — a phone call wouldn't show up here.
PromiseNext contactGap
Agent (Onset): "I will prepare and a rep will be in touch with your estimate" — May 2023Customer 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 2021Customer re-enters via bot menu, hits "no agents available"69 days
Agent (Onset): "Thank you, they will contact you" — Jan 2021Customer re-enters via bot menu, hits "no agents available"142 days

What the buying-intent language shows

Customers who named a product and said they wanted to buy it

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.

352

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.)

$224,559

is the ceiling if all 352 had closed at full price — not a forecast, just "if everyone converted."

A more honest number

At a conservative 5% conversion, that's ~$11,200; at 10%, ~$22,500 — likely a floor, since the 58% who hit a dead end never got the chance to name a product at all.
ProductSessionsList priceAt full priceAt 10% conversion
U26 (dissolved oxygen)46$1,499$68,954$6,895
RX300055$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
MX110197$159$15,423$1,542
MX2301A / MX2302A44$220$9,680$968
CX40214$163$2,282$228
Total363$224,559$22,456

Rows sum to 363 product mentions across 352 unique sessions — a handful of sessions name more than one product.