I tried the broad match switch everyone argues about. 90 day results.

Someone asked last summer whether anyone had switched from phrase to broad and not regretted it. I did it in December on a client account and promised myself I would report back either way, so here it is. Setup: lead gen, around 400 conversions a month before the switch, target CPA bidding that had been stable for a year. I converted one of three campaigns and left the other two as a control, which is the part I would insist on if anyone tries this. Weeks 1 to 3. CPA up about 40 percent, volume up about 25 percent. Search terms full of rubbish. This is the phase where most people revert, and I nearly did. Weeks 4 to 8. I added negatives twice a week rather than monthly - that's the real cost of broad match and nobody bills for it properly. CPA came back to roughly 10 percent above the control campaigns, volume held at plus 20. Weeks 9 to 13. CPA within 3 percent of control, volume plus 18 percent. Net more leads at effectively the same cost. What I think mattered: the account already had a lot of conversion data, the negative list was already good before I started, and I did not touch the target during the learning period even when it looked bad. What I would not conclude: that this works everywhere. On a 20 conversions a month account I suspect the noisy phase never ends, because there's never enough data to get past it. That is the version of the question I still can't answer.