Why most of my campaigns failed
What control groups taught me at Indosat.
At Indosat we sent fewer than 50 campaigns a month to more than 100 million subscribers. Every campaign kept about 10% of people as a control group, so we could see what really changed.
I expected the main problem to be switching: people moving from a package they already had to the new offer. So every business case had a clear line for how many buyers were just switching.
But that wasn't the most common problem. Most failed campaigns simply had very few buyers. People didn't want the offer, and clicks and views stayed flat.
The other thing I learned was about ARPU. A campaign can push ARPU up and still not be a win. I started checking ARPU together with data usage. If ARPU went up but usage stayed flat, people were paying more for the same thing, and that usually didn't last. Many campaigns stopped there.
So now I ask three questions, in this order. Does anyone want this? Is the revenue new, after switching? Does it pay back after launch, not just in the plan?
One more thing. The spreadsheet itself mattered less than I thought. What helped most was writing the key assumption where everyone could see it. Then people could question it, and those conversations were more useful than the forecast.