Research.
What we learn from running SYNC on our own outbound and with client teams, written up with the numbers, including what failed.
case study · September 2026One vacancy, one reply in 58 minutesHow a single dated signal on a hotel's careers page became a call the next morning, and what the message did not say.Read the case
58 minfrom the first message to a reply, and a call the next morning.
All writing
guide · 7 October 2026→What to verify before you buy an AI sales agentEight questions to ask any vendor of an AI sales agent, including us, and how to check each answer. With SYNC's own answers, and where it falls short today.case study · 29 September 2026→We asked our own product who should buy itWe ran SYNC's research with ourselves as the seller: 594 companies, 35 flagged as wrong matches, 244 dossiers of which most say wait. What it got wrong, and why this is weak evidence.guide · 29 September 2026→Mapping the buying group: who really decides?Published group sizes run from three or four people to thirteen or more, because they count different things. A way to map the group in seven roles, and what our own maps can and cannot show.perspective · 29 September 2026→What a human approval is worthResearch on automation says an approval step is weaker than it sounds. Why we keep ours, what our numbers show so far, and why we expect it to wear down.research · 29 September 2026→Advice a model may skip is advice it does skipIn our clearest example, a rule that was only stated to the writing model drifted. The rules that matter are now checks that refuse a draft. What those checks caught, and what that does not prove.research · 29 September 2026→What reps change before they sendOf 266 drafts that a person released, 249 left with a changed text, and most got longer. What that shows about review, and what it cannot show.perspective · 29 September 2026→Why SYNC refuses to write without a reasonCompanies buy when something changes, so SYNC waits for a reason to write. The evidence is indirect, the case against is strong, and our software does not require the date our rule asked for.research · 29 September 2026→The accounts we left aloneOf 2,225 researched companies, 423 have a ready dossier. And 267 of the 570 ready dossiers have no dated reason on record, although our rule asked for one when we counted.perspective · 29 September 2026→Information foraging, and why SYNC reads before it writesA theory from Xerox PARC says people follow a scent to information and leave when it fades. We use it as an analogy for account research, and show where our own records run out.research · 29 September 2026→What we cannot measure yetWe publish no open or reply rates. Most recorded first opens arrive within a minute of sending, replies were incomplete until mid-September, and one meeting is on record.guide · 29 September 2026→What mailbox providers actually requireGoogle and Yahoo ask for proof of identity, an easy way to stop, and almost no complaints. Microsoft requires the first. None of them says that sending slowly helps.
research · no date yetComingWhat 250 agent-written emails taught usWhat our own outbound taught us, including what failed. Waits until every reply has been recorded for a full month.
guide · no date yetComingIs cold B2B email allowed in the Netherlands?What the law and the regulators say. Not published until a lawyer has reviewed it.
See who should buy from you this quarter.
Leave your work email. SYNC reads your company’s website and sends you a first look: what you sell, who probably buys it, and up to five companies scored A to D, each with its source.