55 downloads, one accepted lead, no pipeline
We had never evaluated a campaign after it went live, so I ran the first one, on a whitepaper that was on our internal list as a success.
It was on that list because of its download count, which is a number that looks like a result and arrives early. Here is everything the campaign actually produced.
| What it produced | Count |
|---|---|
| Downloads of the whitepaper | 55 |
| New contacts created from them | 14 |
| Leads that sales accepted | 1 |
| Pipeline | 0 |
The cost side is missing as well. 11 of the 14 contacts came through a paid channel with no cost data attached to it, so I could not say what that one accepted lead cost either. Nobody had noticed the gap, because nobody had opened the calculation.
The follow-up sequence had not sent a mail since 3 August
The sequence meant to follow up with everyone who downloaded it had been built, configured and switched on, and its last send was on the third of August. Seven weeks.
No error, no alert, nothing on a dashboard. From the outside it looked exactly like a sequence with nobody in it.
Seven weeks. A finished follow-up sequence sat switched on and sent nothing, and nobody noticed, because nothing was broken.
Campaign live, sequence on, form published, tracking installed. Every one of those was true.
From six months inside a B2B GTM organization with 36,000 accounts in its CRM.
I want to be careful about what that explains. The silence and the missing pipeline are two findings from the same campaign, and I cannot tell you how much of the second came from the first. What I can tell you is that neither was visible from anything anybody was looking at.
Why an existence check misses this
The check everybody runs is whether the sequence is built. The check that would have caught this is when it last sent something.
Those two questions look interchangeable and they behave differently. A built sequence is a state, and it stays true after the thing stops working. A last send is an event, and it goes stale on its own, which is what makes it useful.
Every dashboard we had answered the first question. All of them stay green while the process underneath produces nothing, because green describes the configuration.
A download counts the cheapest thing a reader can do
A download tells you the promise on the landing page was good enough to trade an email address for.
It cannot tell you whether the person wanted the thing once they had it, whether anybody followed up, or whether sales saw something worth taking. A number built out of cheap actions goes up in situations where nothing else does.
The download count was accurate and still misleading, because it was the only number being produced, so it became the verdict by default.
What to do instead
- Decide the one number before the campaign goes live. It has to be something a reader can only reach by doing something that costs them time or attention. Downloads do not qualify, and neither does an open rate.
- Ask when each moving part last ran, not whether it exists. For every sequence and every scheduled step in a campaign, the useful field is the date of the last send.
- Put that date where the result is read, rather than in a log nobody opens. It belongs next to the number it is supposed to explain.
- Put the cost data in place before the spend starts. A lead with no cost attached to it cannot be judged, and the gap is cheap to close beforehand and impossible to close afterwards.
- Read every campaign once after it is live. This one had a good headline number and a broken middle for weeks, and the only reason it got found is that somebody ran the evaluation for the first time.
The same failure shows up in scheduled work that has nothing to do with campaigns. A routine of mine failed silently and went on looking healthy, for a different reason and with the same shape. And on a much longer timescale, the measurement half of this is what happens when nobody splits organic traffic by brand and non-brand.