How to choose marketing channels and test them properly, including paid.

Tech founders often spread a small budget and a little time across ten channels, see weak results everywhere and conclude that “marketing does not work for us”. The problem is usually not the channels but the method. Most businesses find that one or two channels drive the bulk of their growth. Your job is to find yours quickly and cheaply.

Step 1: list the candidates

Write down every channel that could reach your ICP: direct outreach, LinkedIn, content and search, events, partnerships and resellers, communities, referrals, marketplaces and app stores, paid search, paid social, sponsorships and PR. Our post on channels you should consider is a useful prompt.

Step 2: score and shortlist

Score each from one to five on: are your buyers there, can you reach them at a cost your contract value supports, how quickly will you know if it is working, and does your team have the skills. High-value B2B deals can afford human, high-touch channels. Low-priced self-serve products need scalable, low-cost channels. Pick the top two or three.

Step 3: design each test

For each channel write a one-paragraph test plan:

• Hypothesis: “Paid search on [phrase] will produce demo requests from ICP companies.”
• Budget and time box: for example six to eight weeks and a fixed spend.
• Success test: a target cost per qualified lead or per customer, worked back from your unit economics.
• Owner: one person accountable.

Step 4: run paid channels with care

Paid search captures existing demand: it works when people already search for your category. Paid social, including LinkedIn ads, creates demand through targeting by role and company, but is usually more expensive per click. Start small, send traffic to a dedicated landing page with one call to action, track conversions properly, and switch off broad or automated targeting until you have data. Never judge paid on clicks; judge it on qualified leads and customers.

A good landing page for a paid test says, in this order: the problem in the buyer’s words, what you do about it, one piece of proof (a customer quote or result), and a single next step such as booking a 20-minute call. Remove the main navigation so visitors are not distracted. Add negative keywords for job seekers, students and “template” searches if they are not your buyers. Check your search terms report every week in the first month; it is where most wasted spend hides.

Step 5: decide and scale

At the end of the time box, put each test into one of three buckets: scale (it met the success test), iterate (promising, change one variable and rerun) or stop. Only scale one winner at a time, and watch whether cost per customer rises as you spend more; it usually does.

Worked example

Fieldnote, a fictional Bristol tech startup selling inspection apps to facilities managers, shortlisted outbound email, LinkedIn content and paid search. Paid search was cheap per click but brought mostly students and job seekers, and was stopped after six weeks. LinkedIn content from the founder brought steady conversations. Outbound to a list of 400 facilities firms produced the most qualified meetings per pound. Fieldnote scaled outbound first, kept LinkedIn as a supporting channel and parked paid until it had more brand awareness.

Common mistakes

• Testing too many channels at once with too little effort in each.
• No success test agreed upfront, so every result is “interesting”.
• Stopping a channel after two weeks, or keeping a failing one for a year.
• Copying a channel because a large company uses it.

What to measure

• Cost per qualified lead and per customer by channel.
• CAC payback period by channel.
• Lead-to-customer conversion rate by channel, since cheap leads that never buy are expensive.
• Time from first touch to customer.