Your account has 15 campaigns and none of them have enough budget to learn. Or it has one big campaign and you can't tell which ad is actually working. Both are broken, just in opposite directions. Here's how to decide.
Why the middle ground matters more than you think
Most account damage comes from drifting into one of the two failure states without noticing. Every new idea feels like it deserves its own campaign, so campaigns multiply. Or results get shaky, so everything gets merged "for stability," and the signal drowns. A good account is a machine for creating signal, and both drifts destroy signal in different ways.
Consolidation, done right, concentrates learning power. Testing, done right, protects your main campaign from having to answer every question at once. You want both, in specific places. The skill is knowing which one the current moment calls for.
Consolidate when the data isn't readable
Splitting budget across many campaigns means none of them get enough spend or conversions to tell you anything. You end up with 15 campaigns all returning mediocre, noisy results, and no clear signal about why. That's not testing. That's fragmentation.
The tell is simple. Look at your campaigns and ask what each one taught you this month. If the honest answer is "not much, the numbers are too thin," you're fragmented. Every campaign at Rs.1,000 a day on a purchase event is a machine that produces noise, and fifteen noise machines don't average into signal, they just multiply the work of reading it.
The fix is concentration. Fewer campaigns, more budget behind each, broad enough audiences that the system can actually find buyers. One campaign spending Rs.5,000/day teaches you more than five campaigns spending Rs.1,000 each, because only one of them can exit learning and produce a clean read. This comes down to the learning-phase math: each structure needs enough conversions per week to stabilize, and fragmentation guarantees none of them reach it.
Consolidate when a campaign is starved, when you have more than a handful of campaigns sharing a modest budget, or when you genuinely can't tell which lever is moving results.
Test separately when the question is specific
Testing needs isolation. If you want to know whether a new creative angle beats the current winner, or whether a different audience converts better, put the test in its own campaign with its own budget so the only variable changing is the thing you're testing. Run the winner against the challenger with equal spend and let the data decide.
Why isolation matters: inside your main consolidated campaign, question variations interfere with each other. Delivery algorithms share context, and a new ad placed beside a dominant winner either gets starved of spend or cannibalizes it, either way the read is muddy. A separate small campaign, equal budget, clear question, gives you an answer you can act on.
Define the test before you run it. A question, two variants, a budget, and an end point. "Does the objection-handling creative beat the problem-agitation creative at Rs.500/day for two weeks, judged on cost per initiate checkout?" That's a test. "Let's try this new idea and see" is a donation.
The mistake is turning every test into a permanent campaign. A test answers a question, then you fold the winner back into your main consolidated campaign and kill the loser. Tests are temporary. Consolidation is where the money gets made. I've inherited accounts with forty campaigns where thirty were tests from two years ago that nobody ever closed. The account was a museum, not a business.
The structure that does both
One consolidated campaign doing the bulk of your spending, built to produce a clear read. A small, separate testing campaign where you try new creatives and angles with defined budgets and an end date. When something wins the test, it graduates into the main campaign. When it loses, you kill it and move on.
A 90/10 split is a good starting shape. Ninety percent of budget in the main concentrated campaign, ten percent in a testing lane, one to three live tests at a time, each with a defined budget per variant.
This is concentration plus a testing engine. Most brands either have neither or have so many permanent test campaigns that the whole account is unreadable. Keep the main account boring and concentrated. Keep the tests small and temporary. Boring main campaign, sharp tests, one promotion lane. That structure reads clean on the worst months and scales without drama on the good ones.
The budget math before you decide
One way to make the decision mechanical instead of emotional. Estimate what each campaign needs to learn: 50 conversions a week per ad set, times your real cost per result, and that's the weekly budget a healthy campaign needs. Cost per purchase Rs.600 means each ad set needs roughly Rs.30,000 a week. Total ad budget Rs.60,000 a week means you can afford two learning campaigns. Not six, not fifteen, two.
Then check the other side, the readability question, with the same kind of arithmetic. A campaign testing a new angle needs enough spend to produce maybe 30 to 50 results on the test metric before the read means anything. Below that, any difference you see is noise. Working the two sums before touching structure keeps you out of the classic mistake: consolidating past the point where you lost the ability to test, or testing past the point where nothing could learn. Money per campaign per week is the number that moves you, in either direction, and it's knowable in five minutes on a spreadsheet.
I'm Hirdesh Matai, founder of HDM GRO-ads, the science of e-commerce advertising.