The reason volume stopped working is mechanical. Filters improved considerably, sending reputation now carries more weight than message content, and recipients receive more unsolicited email than they did while having less patience for it. A business sending five hundred generic messages a week sees almost nothing except a damaged domain, and the damage outlasts the campaign.
Small scale works because the message can be specific. If you have spent ten minutes on somebody's website and found an actual problem, you can write a sentence that could only have been sent to them. That sentence is what earns a reply, and no amount of clever templating substitutes for it.
The honest test is whether you can name why this business rather than any business. If the answer is that they appeared on a list, the message will be generic, the recipient will recognise that immediately, and the send is a waste of a domain that took time to build.
The arithmetic favours quality more than people expect. Twenty researched messages producing three replies beats five hundred templated ones producing two, and the twenty take about the same total effort once you account for the list building the five hundred required.
Where it works best is not what most people attempt. Approaching businesses that serve the same customers you want, at a different point in the same journey, is considerably more productive than approaching potential customers directly. An accountant who fields questions about websites is a better contact than a business that has not yet decided it needs one.
Trigger events are the second strongest use. A business that has just opened, moved, expanded, or won something has a reason to be hearing from you this week rather than at some point. Public records, local news, and business registrations all surface those, and the message writes itself.
The third is reconnecting with people you already know, which is not cold at all and outperforms everything else. Former colleagues, past clients, people you studied with, anybody you have dealt with professionally. Most businesses never send this message and it converts better than any cold approach.
Understand what the channel cannot do. It will not build a brand, it will not work while you are unclear about what you sell, and it will not compensate for a weak offer. A poor offer sent to five hundred people is still a poor offer, and the volume simply spreads the evidence more widely.
Then decide honestly whether you will do the research. If ten minutes per recipient is not something you will sustain, this channel is not for you, and the same hours spent on referral relationships or local visibility will produce more.
Keep the two channels separate in your own mind, because they behave differently. Outreach to potential partners is a relationship activity with a long horizon. Outreach to potential customers is a sales activity with a short one, and confusing them produces messages that do neither well.
Measure replies rather than opens, since open tracking is increasingly blocked and was never a decision anybody could act on. A reply, positive or negative, is the only signal that tells you whether the message worked.
Consider whether the same hours would produce more elsewhere before committing to the channel. For most first year businesses, asking existing customers for referrals and building local visibility both convert better per hour than any cold approach.