Smart home devices, installer network · Email

Shelly Group asked 3,900 installers four questions, in their own language

Shelly Group is a publicly listed maker of smart home and building automation devices with a registered installer network in more than 100 countries. A Sellinger agent wrote to installers in their own language, turned the replies into structured insight, and passed the big projects and the commercial questions to Shelly's team. This was not a sales campaign. It was a listening programme.

Client: Shelly Group
Channel: Email, one short nudge in Bulgaria
Window: 11 May to 9 October 2026
Data as of: 9 October 2026
Written by: Asen Levov
Reviewed by: Viktorio Shopov

installers emailed, in 111 countries (some also got one short nudge in Bulgaria)

installers genuinely replied; this includes 30 who wrote in first

installers answered the survey; 172 of the 210 complete surveys came back in one message

big project installers surfaced from the replies

In short

The agent asked four numbered questions in the body of an email, in the installer's language, with nothing to click. A reminder or two followed. Replies were answered in about a minute, repeating two or three things the installer had said.
In the June to August waves, 13.6% of installers replied. With the first email alone the figure would have been 3.3%. Three in four replies came only because of a follow up.
The agent triaged. A person at Shelly took over pricing, support and partnership topics, and the big projects.

9 min read

The key numbers, with their bases

Shelly installer programme, data to 9 October 2026

Number	What it is
3,940	installers emailed, in 111 countries (some also got one short nudge in Bulgaria)
360	installers genuinely replied; this includes 30 who wrote in first
225	installers answered the survey; 172 of the 210 complete surveys came back in one message
64	big project installers surfaced from the replies
17	installers booked an optional call (bookings, not held calls)
21	languages the agent wrote in; 51% of installers replied in a language other than English
1.4 min	median time for the agent to answer an installer, over 243 messages; 70% within 5 minutes

What “conversion” means here. 238 of the 360 installers who genuinely replied did at least one of four things: answered the survey, booked a call, surfaced a big project, or engaged with Shelly's thank you code. It is a measure of engagement. No sales or revenue is claimed.

A quiet portal and no way to read thousands of answers

Professional installers are a key channel for a device maker, and many portal accounts were quiet. Shelly had little signal on who installed its products, on what kind of projects, or what stopped them installing more. A partner team cannot write to thousands of installers in dozens of languages, read every answer, and spot the one quietly fitting out a very large hotel.

Sellinger deployed an agent that writes as “The Shelly Team”, the Installer Development Team, and never as a named person.

It sends a four question survey in the installer's language.
While it ran on its own (June to August) it answered replies within about a minute.
It offers an optional 30 minute call only after the installer has answered.
It hands pricing, support and partnership topics to a Shelly person.

Every reply becomes structured insight: project types, Shelly projects per month, biggest challenge, the one thing Shelly should improve, and a flag on installers with big projects. Shelly reviews it in reports and in a key accounts view.

The playbook: ask nothing but four questions

Day 0
Survey email
Four numbered questions in the body, in the local language. Nothing to click.
Day 7 to 9
Follow up 1
Short and self contained: the questions travel with every email. Median 8.3 days after the first email.
Day 14 to 18
Follow up 2
The last touch. In Bulgaria, one short nudge 45 minutes after the email.
A reply
Answered in about a minute
In the installer's language, repeating two or three things they said.
After the survey
One optional line
It offers a 30 minute call, with a booking link. Never before.
A big project or an issue
A Shelly person takes over
Pricing, distribution, returns and partnerships get a holding reply, then staff.

The four questions (abridged)

What kind of projects do you mainly work on?
Roughly how many Shelly projects do you complete per month?
What is your biggest challenge?
What is the one thing Shelly could do to support you better as an installer?

Rules the agent works by

Team voice, never a named person. Peer to peer, short, no selling.
Write in the installer's language from a 60 country language map. Once they reply, follow the language they wrote in.
No booking link before the survey is answered. Then one optional line.
Do not promise prices, codes, dates or certification status. Commercial and support topics get a holding reply and go to a person.
An opt out closes the thread for good.

From June to August the engine sent up to five emails. Shelly then capped it at the first email plus two follow ups.

From a quiet portal to 225 answered surveys and 64 big project installers

The funnel

Step	Installers
Emailed	3,940
Genuinely replied (includes 30 who wrote first)	360
Converted, by the definition above	238
Answered the survey	225
Big project installers	64
Booked a call	17

Survey counts come from Sellinger's insight extraction and still undercount the October wave. 9 of the 17 booked calls came from the optional call line, three of them within 33 to 40 minutes. 66 converted threads were handed to a person.

Reply rate by wave

Wave, by month of first email	Emailed	Replied	Rate
May, Bulgaria (email and a short nudge)	114	31	27.2%
June, rest of world	1,452	201	13.8%
July, rest of world	516	67	13.0%
October, regional relaunch (one week old)	1,857	38	2.0%

The October wave is one week old, so its replies are still arriving. Read June and July as the matured rate. The rows show the main waves; the headline totals count installers differently (the 360 includes installers who wrote in first), so the rows do not add up exactly to them.

Big projects, described in general terms. The big project replies included a hotel fit out, a residence with a very large number of devices, and an IT integrator that serves business clients, many of them hotels. Most of these installers were not top tier in the portal data. Their size only showed up in what they wrote.

Follow up efficiency: three in four replies came from a follow up

The June to August waves, 1,966 installers, are the cleanest view, because every installer could receive the whole sequence.

Reply rate per touch, of installers who got that email without having replied yet

Touch	Reply rate
First email	3.3% (64 of 1,966)
Follow up 1	5.3% (100 of 1,902)
Follow up 2	3.0% (45 of 1,513)
Follow up 3	3.6% (33 of 920)
Follow up 4	6.3% (26 of 414)

With one email only, 3.3% would have replied. With the follow ups, 13.6% did, which is 4.2 times as many.
76% of replies came only because of a follow up. Share of all 268 replies by touch: first email 24%, follow up 1 37%, follow up 2 17%, follow ups 3 and 4 22%.
Median time from first email to first reply: 8.8 days. Median time from the triggering email to the reply: 10 hours. Installers answer when it lands.
Follow up 1 was the best single touch.
A quick nudge after the email works for a local base. In Bulgaria a reminder email followed by a short nudge got 18% of recipients to reply on that touch.
Touches 4 and 5 still produced 22% of replies, but they also produced the only 2 “please stop” complaints, across 1,966 installers and more than 6,700 emails. For a base of existing partners Shelly chose two follow ups, and the data supports it.

Speed and language

Installers reply from job sites, in the evening and at weekends.

Speed and language

Measure	Result
Agent answered within 5 minutes	70% (243 installer messages)
Within 1 hour	84%
Within 24 hours	91%
Installer messages that arrived outside office hours (Sofia time)	50%
Converted installers who got a reply within 15 minutes of their first message	67%
Languages the agent wrote in	21
Installers who replied in a language other than English	51% (178 of 351)
Agent replies that matched the installer's message language	92% (309 replies); 98% matched either the message or the country language
Replies after an installer's first message sent by the agent on its own	82%

Languages included Bulgarian, Polish, Spanish, Portuguese, Greek, Czech, Croatian, Hungarian, Romanian, Hebrew, Turkish, Ukrainian, Macedonian, German, Italian and French. Greetings follow local grammar, such as the vocative in Czech and Polish. One Brazilian installer answered only once the Portuguese version arrived.

How the agent handled what installers raised

Situations and what the agent did

Situation	Threads	What the agent did	Result
“I already answered” or “I never got your email”	about 12	Re sent only the questions within a minute	3 of 7 resend requests ended in a completed survey
Partial answers	24	Asked only for the missing numbered questions, in the same thread	12 ended complete
Pricing, buying direct, distribution	most common	A short holding reply in the installer's language, then a Shelly person	Handled by staff
Support problems raised in the survey	about 20	Holding reply, escalation to Shelly support	Handed to Shelly's team
“What's in it for me?”	2	Explained what the call covers and why	One installer booked six days later
“Please stop emailing”	2	Closed the survey thread	No further survey emails

Several installers wrote back to say they appreciated that someone read their answers. No installer is quoted here: installers are described, not named, and their words are not reproduced.

What did not work

Too many touches. Touches 4 and 5 produced the only two stop complaints. Shelly capped the sequence at two follow ups.
A reminder that dropped the questions. Seven installers asked for the questions again. Since 30 July every email carries them.
The October wave is only a week old. Its 2.0% is not a rate yet, and survey counts still undercount it.
Improvised answers. Before October the agent sometimes improvised on pricing and made promises that were not its to make. Since then, commercial and support topics get a holding reply and go to a person.
The agent is faster than the hand off. The agent answers in about a minute, but a thread that needs a person is only as fast as that person. Calls, partnerships and pricing need a human owner who shows up.
Calls booked, not held. 17 installers booked a call. The study counts bookings. The calendar marks a booking completed when its slot ends, so we have not verified how many of those calls took place.
No sales claim. The programme measures answers, calls and projects flagged. It does not measure sales or revenue and claims none.

Six patterns behind the 238 converted installers

Ask, don't sell.

No pitch, no link, no login. Four numbered questions in the body of an email, so the installer can answer inline.

172 of 210 complete surveys came back in a single message.

Win on the follow up.

Installers opened with an apology for the delay and then answered everything. To them the follow up read as care, not pressure.

76% of converted installers replied only after a follow up.

Answer in a minute, with their words.

Each reply repeats two or three specifics: the products, the project type, the numbers, the place.

67% of converted threads got an AI reply within 15 minutes.

Survey first, call second.

The booking link appears only after the installer has answered, as one optional line.

9 of the 17 booked calls came from that line.

Local language is the precondition.

Nearly half of converted installers wrote back in their own language.

46% wrote in a language other than English.

The agent triages, a Shelly person closes.

Partnerships, device deals for content creators, certificates and cash back were finished by Shelly staff in the same thread.

66 converted threads were handed to a person.

What made it work

1
Lead with a question, not an offer. Partners answer four honest questions. They ignore promotions.
2
Write in their language, every time. Half the replies were not in English, and some installers answered only once the email was in their language.
3
Plan for the follow up, and cap it. It earned three in four replies. Cap it where your partners start to feel spammed: here, two follow ups.
4
Answer in a minute and repeat what they said. The agent answered 70% of installer messages within 5 minutes.
5
Read every answer for size. The biggest projects came from accounts the portal data rated low.
6
Keep a person ready for the warm ones. The agent triages and a person has to close.

How we measured this

Source. Sellinger production data for the Shelly workspace, read only, snapshot 9 October 2026. Warm installer programme only. A separate pilot of cold outreach in the UK that started on 5 October is excluded.
Emailed means received at least one outbound email (3,940 installers).
Replied means wrote back after our first email. Out of office and delivery notices are excluded.
Touch is the nth email before the installer's first reply. The reply rate per touch is the share of installers who got that email without having replied who replied before the next one.
Converted means answered the survey, booked a call, surfaced a big project or engaged with the thank you code, as defined above. It is a measure of engagement, not of sales.
Survey answers and big project flags come from Sellinger's insight extraction on full replies. 150 threads were also read by hand.
Response time is the installer's message to the agent's next email. Replies typed by Shelly staff and human approved drafts are excluded from agent speed.
Installer names are changed or omitted and identifying details are generalized. Threads were translated from Greek, Spanish, Portuguese, Croatian and other languages. This page has no illustration.