An owner tells me they want to try LinkedIn ads. Or a trade show. Or a cold email programme, or a podcast, or paid search in a new city. The question is always some version of the same one: will it work?
Nobody can answer that in advance. What you can do is design the test so that after a set amount of money and a set amount of time, you know. Most small companies skip that design step entirely. They spend, they watch, they get a feeling, and then they decide based on the feeling. That is how a channel that would have worked gets killed in week three, and how a channel that will never work stays on the books for a year.
Here is the sequence, in order:
- Set a pilot budget you can afford to lose. Decide the number before you look at what the channel costs to run.
- Write the success test before you spend a dollar. One sentence with a date and a number, so nobody can reinterpret the result later.
- Give the channel the time its type needs. Four to six weeks for paid search, eight to twelve for paid social, a fixed volume rather than a fixed time for cold outreach, six to nine months for content.
- Do not kill it in week three. Every channel has a learning period. Act only on the date you wrote down, unless tracking itself is broken.
- Do not let it run for a year either. Put a quarterly review date on every channel and cut what cannot justify itself out loud.
- Get tracking and follow-up in place first. A pilot without a way to trace leads to their source, and a team that can respond fast, wastes the spend before the test even starts.
- Run one pilot at a time. Two at most, and only if they need attention from different people.
Set a number you can afford to lose
The pilot budget is a risk decision. Before you look at what a channel costs to run, decide what you are willing to write off entirely if the answer is no.
For most 10 to 50 person B2B companies in Metro Vancouver, that number sits somewhere between $3,000 and $15,000 for a single channel test, depending on the size of the business and how much cash sits in the account. The exact figure matters less than the fact that you picked it deliberately and can lose it without changing any other plan.
Two things go wrong when you skip this. The first is that you start a test you cannot afford to finish, so you pull the money out halfway and learn nothing. The second is that the money hurts, which makes you emotional about the result, which makes you read early noise as a signal. An owner who is worried about the spend will find a reason to stop at day twelve.
Include your own time in the number. If a pilot needs eight hours a week from you or your best salesperson, that is a real cost and it competes with revenue work. Put an hourly figure on it and add it in. My rule is that if the time cost is larger than the cash cost, the channel is probably wrong for a small team.
Write the success test before you spend
This is the single habit that changes the outcome, and it takes twenty minutes.
Write one sentence, in a document with a date on it, that says: if by [date] this channel has produced [number] of [thing], we continue. Otherwise we stop. Then send it to whoever else has an opinion, so the test cannot be reinterpreted later.
A real example of the shape: by March 31, with $8,000 spent, this channel needs to have produced at least 12 qualified conversations and 2 proposals. Below that we stop.
Pick the measure that sits closest to money while still being reachable inside the test window. Closed deals are the honest measure, but if your sales cycle runs four months, a three-month test will show zero closed deals for a channel that is working fine. In that case the test measures qualified conversations, and you accept that you are measuring one step earlier than you would like. Say that out loud in the document so nobody pretends later that it measured revenue.
The number itself should come from your own arithmetic, not from a benchmark article. Take the pilot spend, divide by what you currently pay to win a customer through your best existing channel, and you get a rough count of what this channel would need to produce to be worth keeping. If you do not know what you pay to win a customer, work that out first. It is the foundation for every spending decision you make, and it is covered in more depth in the post on acquisition cost and payback period.
How long a fair test takes
Channels do not reveal themselves on the same schedule, and using one timeline for all of them is where most pilot programmes go wrong.
Paid search on high-intent keywords shows you something in four to six weeks, because people searching for what you sell are already in market. If the phone does not ring at all after 300 clicks on a keyword that describes your service exactly, the problem is your offer or your landing page.
Paid social and LinkedIn ads need eight to twelve weeks. You are interrupting people who were not looking for you, so the first version of the ad is usually wrong and you need two or three rounds of changes before you see the channel's real performance.
Cold outreach needs a fixed volume rather than a fixed time. Six hundred well-targeted contacts, properly sequenced, will tell you whether your message lands. Sixty will not, no matter how long you spread them over.
Content and search visibility need six to nine months before the numbers mean anything, which is why they are a poor fit for a short pilot. If you want to test this, test the production process rather than the results: can your team actually publish twelve useful pieces in a quarter? Most cannot, and finding that out is worth the quarter.
Trade shows and events are a single-shot test with a long tail. One show tells you whether your buyers are in the room. It tells you nothing about return until you have worked the list for three to six months afterwards, which is why the follow-up system matters more than the booth.
Killing a channel in week three
This is the most common failure and it is expensive, because you pay the full setup cost and collect none of the return.
Every channel has a learning period where the results look bad. Ad platforms need conversion data before their targeting improves. Your ad copy is wrong on the first attempt. Your landing page has a problem you have not found yet. Your sales team does not know how to handle a lead from this source, so the first ten get mishandled.
Week three is where all of that is visible and none of the fixes have landed yet. An owner looks at the dashboard, sees money out and nothing in, and stops it.
The written test is what protects you here. When you have a document that says the decision happens on March 31 with $8,000 spent, week three is simply not a decision point. You can be unhappy on week three. You cannot act on it. That constraint is the whole value of writing it down.
The exception is a genuine execution failure rather than a weak result. If tracking is broken, if the ads were never approved, if the landing page was down for nine days, stop, fix it, and restart the clock. A broken test is not a failed test.
Running a losing channel for a year
The opposite mistake is quieter and costs more. A channel produces a trickle of leads, never enough to justify itself, and it survives because nobody ever looks at it directly.
It survives for three reasons I see repeatedly. Somebody on the team likes it and defends it. It produced one good customer eighteen months ago and that story keeps getting retold. And the spend is small enough each month that it never triggers a review.
Two hundred dollars a week is $10,400 a year. That is a real budget line hiding inside a small number.
The fix is a calendar entry. Every channel gets a review date, every quarter, where somebody states the cost and the result out loud. Channels that cannot survive that conversation get cut. This is the same discipline that makes a simple marketing dashboard worth building: it forces the numbers into the same room as the decision.
What has to be in place before the pilot starts
Two things, and a pilot without either of them is wasted money.
The first is tracking that connects a lead to its source. You need to be able to answer, for every enquiry, where it came from. That means a form that records the source, a phone number or a question that identifies the channel, and somebody who actually enters it. It requires that the last step happens every time, which is usually the part that fails.
The second is follow-up capacity. If your team already takes three days to respond to an enquiry, adding a channel will not help you. The new leads will get the same three-day response and convert at the same poor rate, and you will conclude the channel does not work. Fix the response time first. It is cheaper than any pilot and it improves every channel you already run. The inbound conversion guide covers what that looks like in practice.
How many pilots to run at once
One, if you have no dedicated marketing person. Two at the most, and only if they need attention from different people.
The limit is attention. A pilot needs somebody watching it weekly, making changes, and handling the leads properly. Split that attention across four channels and every one of them gets a half-hearted run, which produces four inconclusive results and no knowledge.
Sequence them instead. One channel per quarter, tested properly, gives you four real answers in a year. That is a better position than trying eight things badly and still not knowing which one to scale.
When you do get a yes, the next decision is how much more to put into it. That is a separate exercise and it deserves its own planning round rather than an impulse to triple the spend the week the pilot ends.
The takeaway
Pick a number you can lose without flinching. Write the success test down with a date on it before any money moves. Give the channel the time its type actually needs, then make the decision on the day you said you would.
Run one pilot at a time and make sure your tracking and your follow-up can handle the leads before they arrive. A channel test is worth doing when it produces a clear yes or no. Most of them produce neither, and that is a design problem rather than bad luck.