Evaluating Campaigns: Which Numbers Count After Four Weeks

Why is the four-week evaluation so often wrong?

A campaign is a time-limited series of measures with a common goal and its own budget, such as search ads for one service, an ad series on LinkedIn or a mailing to existing customers. After four weeks, the first monthly report is in. It shows impressions, clicks, costs and perhaps seven inquiries. Far-reaching conclusions are often drawn from these numbers: the campaign is stopped, doubled or rebuilt.

This guide therefore answers the question: Which campaign numbers support a decision after four weeks, which do not yet, and how does a small business decide without being fooled by chance? We first clarify the terms and a model from decision research that explains why small samples are so convincing. Then we weigh two attitudes, build a check in three questions and work through an example.

After four weeks you know whether the campaign runs. Whether it pays, only the orders will tell you.

Leading and lagging indicators: what each number shows

A campaign’s numbers can be arranged along the measurement chain described in the article on how to measure marketing success: visit, inquiry, conversation, order. At the start of the chain are leading indicators. They arrive quickly and in large numbers, such as impressions, clicks and click-through rate. At the end are lagging indicators. They arrive slowly and in small numbers, such as proposals and orders. Leading indicators say a lot about the campaign itself and little about its result. With lagging indicators it is the other way round.

The thinking model: the law of small numbers

Tversky and Kahneman (1971) described what they called belief in the law of small numbers. The law of large numbers states that a large random sample resembles the population it comes from. But many people expect the same of small samples. They take seven inquiries as a scaled-down image of what the campaign will bring permanently.

In the Quarterly Journal of Economics, Rabin (2002) analyzed this error formally and worked out a consequence that is central to evaluating campaigns. Believers in the law of small numbers infer overly extreme values from short sequences. And anyone who sees short sequences from many sources attributes differences to the sources that do not exist. Rabin calls this fictitious variation and considers it one of the economically most important consequences of the error. His example is investors who, after two good years, believe they have found a particularly able fund manager, although all managers are average.

Applied to campaigns this means: anyone who compares search ads with seven inquiries and LinkedIn ads with four after four weeks sees a winner. In fact, both channels may be equally good, and the difference is chance. The article on sales bottleneck analysis gives a rule of thumb for this: with fewer than about twenty cases per stage, rates are very uncertain.

How large is chance with small numbers? A worked example

The example builds on the article on search ads with a small budget. There, a quarter with 3,000 euros at 4 euros per click brings 750 clicks, 22 inquiries and 3 orders. After four weeks, about a third of that has accrued: 1,000 euros, 250 clicks and 7 inquiries. The inquiry rate is 2.8 percent.

How precise is 2.8 percent? Calculating, with a common statistical method, a range in which the long-run inquiry rate lies with high probability gives about 1.4 to 5.7 percent for 7 inquiries from 250 clicks. The cost per inquiry can therefore lie between roughly 70 and roughly 290 euros. A second campaign with a 2 percent inquiry rate would be indistinguishable from this one.

That does not make the number worthless. If even the least favorable cost per inquiry lies below what an inquiry may cost according to the back calculation, that is a good sign. If even the most favorable lies above it, that is a bad one. If the affordable limit lies in the middle of the range, after four weeks you simply know nothing final yet.

Decide early or wait? Two views

The first view says: budget is scarce, so decide early. Anyone who lets a weak campaign run for three months burns money that is missing elsewhere. In its favor is that many mistakes do show up quickly: wrong search terms, an audience that does not fit, a form that does not work.

The second view says: results take time, so wait. In business-to-business markets, weeks often pass between inquiry and order. Anyone who stops after a month measures only the beginning of the measurement chain. In its favor is the finding from attribution research mentioned in the article on marketing success. Gordon, Moakler and Zettelmeyer (2026) analyzed 2,226 advertising experiments. Last-click attribution explained the actual effect measured in the experiment only weakly, with a coefficient of determination of 0.19.

Both views are right, but for different questions. I consider a division correct: questions that leading indicators can answer are decided after four weeks. Questions that only lagging indicators can answer are decided after a deadline set in advance, for small B2B businesses usually after one to two quarters. The mistake lies not in deciding early or late but in answering a question with the wrong number.

The four-week check: three questions in a fixed order

Model and assessment yield a check that I call the four-week check. It asks three questions. The order matters: a later question is asked only once the earlier one is answered.

1. Question 1: Does it work?

   Is the campaign being delivered, do clicks arrive, does the form work, does the inquiry’s source reach the CRM? Leading indicators fully support this. A fault is fixed immediately, not evaluated.

2. Question 2: Does it hit?

   Does the campaign reach the right people? Search terms, industries and job functions of those clicking, share of suitable inquiries according to the 4P filter, notes from the first conversations. This is where you sharpen: exclusions, targeting, ad copy.

3. Question 3: Does it pay?

   Is the cost per inquiry within the range the back calculation allows? After four weeks only as an interim status with a range. The decision is made after the deadline fixed before launch.

For the second question, the share of suitable inquiries is often more telling than their number. How an inquiry is classified by fit, problem, person and plan is described in the article on lead qualification. Two conversations with exactly the right buyers say more after four weeks than ten inquiries from students and competitors.

What to check first when numbers are weak

If a campaign looks weak after four weeks, the cause often does not lie in the ad. The article on how to align ads and landing pages shows that the same ad at the same cost per click can cost 100 or 400 euros per inquiry depending on the landing page. Before giving up a channel, it therefore pays to follow an order that goes from the cheapest to the most expensive change.

1. Technology: test the form, check the source in the CRM, check the landing page’s load time on a phone.
2. Landing page: Does the headline confirm the ad’s promise? Does the effort in the form match the offer?
3. Targeting: exclude search terms or audiences that do not fit, lower bids for unsuitable times or regions.
4. Message: align the ad copy with the target group’s problem.
5. Channel: only when everything else has been checked and the deadline has passed, move budget to another channel.

How to improve a website’s inquiry rate with few visitors, without relying on tests with large visitor numbers, is described in the article on how to measure the website inquiry rate. Its rule also applies to campaigns: one change at a time, a fixed comparison period, conversations as a second source.

What follows in general

Evaluating campaigns shows a general problem of small businesses. Tools built for large companies produce reports that look precise even when the numbers behind them are small. The danger lies less in the numbers than in the conclusions. Rabin’s fictitious variation describes exactly this: anyone who reads differences into many small samples finds winners and losers where there is only chance.

The limits of this guide are clear. The range in the worked example is an approximation, and real campaigns also fluctuate by weekday, season and competition. The studies by Tversky and Kahneman and by Rabin concern judgment in general and capital markets, not marketing. The study by Gordon and colleagues comes from a large advertising platform. The direction is robust: small numbers say less than they seem to.

What remains and how to start

We began by asking
campaign numbers support a decision after four weeks. Leading indicators do: whether the campaign is delivered, whether technology and form work, whom it reaches and which search terms or audiences it brings. Inquiries do with caution, as a range and together with conversations. Orders and cost per order do not yet, when they consist of single-digit numbers. Anyone who asks the three questions of the four-week check in their order and ties the decision about success to a deadline set in advance protects their budget from bad campaigns and good campaigns from hasty judgments.

• Note the deadline for the success decision for every running campaign.
• Test the form and source field yourself.
• Go through the search term or audience report and exclude what does not fit.
• Call two people who inquired via the campaign.
• Write down the cost per inquiry as a range, not as a single number.