From 1992 until it was sold in 1996 I ran an internet service provider (ISP) called The Little Garden, later TLGnet, Inc. I had four other partners of varying degrees of involvement.
TLG started in 1992 as a "share" of a then-arcane and expensive internet connection between three pioneer businesses; more detail is available here. In 1992 the Internet was still crawling out from under the ARPAnet entanglements, and was not quite commercial. TLG was a business formed out of need; in hindsight, the cart before the horse.
TLG started in this primordial time without business plan, money, or really, any plan at all. That changed rather quickly and by 1994 we were caught up in the commercial-internet undertow and on our way to being a serious regional player.
From zero cash flow in 1992 it grew to over $125,000.00 per month when we sold TLG to Best Internet Communications (now part of Verio). Obviously I had help -- substantial, indispensible talent: employees, mainly Edgar Nielsen and Deke Nihilson and the crew we hired and trained -- but I remained chief business and network architect from start to finish.
It's safe to say that Little Garden/TLG would not have been more than a blip on the map without the help of Randy Bush. Though he wasn't part of TLG in any direct way, he'd been a friend of mine for some time already (having done the RFC-like state machine documentation for the FidoNet protocol and set the standard for that protocol's success) and besides giving me invaluable advice on network technical stuff, as part of the Rainy-Garden (rg.net) consortium? got Little Garden it's first "real" connection to the net, after Rick Adams kicked us off Alternet (the infamous "find yourself another service provider" remark after an argument with John Gilmore) via Sprint Government Systems Division. (In 1993 the internet the "commercial vs. research" split use of resources was ending.) A real T1!
From then on, from Randy I learned: routing via BGP, policy routing, policy databases, peerage (and got us peers at MAE-WEST and elsewhere), at his continual poking we did NACRs and DNS and WHOIS right.
From dba to Inc
TLG existed in three stages; prehistory, d/b/a, and TLGnet, Inc. Prehistory is socially complex and business simple; read about it elsewhere. Essentially, it was a means to share an expensive 'net connection amongst some savvy users. The idea of TLG-as-business was thrust upon us externally.
The d/b/a phase (1993-1995): TLG started out, business-wise, as a d/b/a of Tom Jennings. During the build-up phase I relied on our existing, loyal and socially-coherent customer base as a source for operating and startup cash, to incrementally build our infrastructure. I was essentially the only day-to-day talent, but I had substantial technical support from certain "customers" (eg. Stu Grossman). Our business plan was set by the end of 1993, and at that time very radical in its goals and approach -- briefly,
Essentially, other ISPs restricted use and resale of their connections, in a sort of zero-sum approach. By concentrating on bulk connectivity we at once created a market for our customers to provide the vertical services we didn't want or couldn't afford to provide, and built a hard-to-beat solid rep that for a long while locked out direct competitors to our core business; having our prices online and breaking down the leased-line costs and equipment gave us a major one-up economically, technically, and in credible reputation over nearly all other ISPs, big or small.
TLG was run on QuickBooks. I kept tight control over invoicing and expenses from the start. In 1994 I hired two employees (soon made minority partners), Deke Nihilson (jack of all trades) and Edgar Nielsen (network architect and heavy sysadmin). Payroll was done in-house (foolishness!). By the end of 1994 TLG grossed about $10,000/month.
We were profitable from the start. We paid off a private $10,000 loan in a year, as we built the network, solely from operating cash.
By 1995 we had a 24/7 network operations center (NOC), by 1996 every node in the network was SNMP monitored. We had about 2000 Class C addresses (the coin of the realm), we had automated DNS in 1995, a complete database of every customer, service and network interface from which we generated router code and logins. We were purchased by Best for our income and our NOC (still pretty nifty by 1999 standards).
Last but not least, we created our own talent -- we hired smart and hard-working people without skills -- sometimes with no computer skills at all -- and brought them up to speed within a year. Not only did we get relatively inexpensive and dedicated talent, we got to give back to the local community, too. Almost without exception all are now "in the industry" at competitive salaries, and marketable skills.