I picked what to sell before switching the ads on
Vasiliev Bronze Foundry wanted to find out whether there was online demand for what it makes. I was given three months and a test budget.
A designer built the site to my brief; a paid-search specialist assembled the campaigns - the structure and the edits were mine.
What was wrong
Nobody had measured the demand, and that was the lesser half of it. A ready-made piece off the shelf and a commissioned casting differ severalfold in price, and their advertising looks identical. Testing everything at once is pointless: the cheap items produce cheap enquiries, the expensive ones get lost in the pile, and a test like that supports no decision at all.
What I did
I chose the direction before launch: made-to-order art casting, with ready-made pieces left out of the test. This was not the first bronze foundry in my work - I had a rough idea of where to go.
Then we built a site for that direction and a campaign in Yandex.Direct. I set up a CRM and call tracking - that is when you can see which ad a particular call came from - and routed every enquiry into it: calls, site forms, emails, messengers. Three quarters arrived as text rather than voice; without a single point of collection half the test would not have counted.
I worked with the sales team separately. I rebuilt the first conversation: the manager finds out what the order is and by when, instead of quoting a price straight away. I listened to the calls myself and went through them. And I pushed for fast replies to email - here a message left for two days goes cold.
The test confirmed demand, with room to spare: one of the four buyers came back for a second batch inside the test window. What did not work was the setup - the test was shorter than the sales cycle, and that was my mistake.
The other thirty-six never reached an order, and some of them were alive: with made-to-order casting, between "interesting" and "paid" sits the approval of a model and a production lead time, and that does not always fit into three months. On top of that the advertising kept bringing in private buyers after a single figurine, while the money turned out to be in batches for organisations. And the sales team could not run a long deal: they advised and waited for the client to come back on their own.
The lesson I took forward: never set a test shorter than the sales cycle. Three months showed the demand and the cost of a buyer, but not how much of that demand reaches money.
I never got to see how it worked over a longer run: the owner of the foundry ran into personal circumstances and the project was stopped. A deal roughly the size of the whole test was sitting in approval - it never closed.