Channel mix analysis: which channel should the budget move to?

It is 9:40 on a Thursday morning. A marketing lead opens the mid-month spend sheet: half the budget went to search, a third to paid social, the rest to email and content production. The source breakdown of the records that came in over the same period tells a different story. Paid social produced almost twice as many records as search, yet most of the opportunities that reached a quote came from search. The figures here are illustrative; the question is not. If part of next month's budget is going to move, which way should it move, and on what evidence?

This piece is about answering that with a method instead of a hunch. We start by defining what a channel mix analysis actually measures. Then we look at why the spend split and the outcome split rarely line up, how to calculate a cost per channel honestly, and how much your choice of attribution model changes the ranking. After that we put shift thresholds into a table and work through saturation signals, organic channels, cadence, and the mistakes that show up most often.

What does a channel mix analysis actually measure?

A channel mix analysis is not an answer to "which channel is better". It measures how much share each channel takes relative to the others, and how well that share matches the results it produces. It looks at the balance of a portfolio, not the performance of a single line item. A channel that looks strong on its own may still have a ceiling on the extra load it can absorb.

The output is three numbers: the channel's share of spend, its share of outcomes, and the gap between them. "Outcome" can be qualified records, opportunities created, or won revenue, depending on what you sell. The critical part is measuring every channel with the same definition of outcome. If you judge one channel on record count and another on revenue, the comparison is broken before it starts.

The Channel Mix Analysis in Marketing Hub brings that trio onto a single funnel: paid and organic sit on one screen, with the spend split and cost per channel side by side. That turns "paid social brings in a lot of records" into "paid social takes 30 per cent of spend and produces 18 per cent of qualified records". The second sentence is a decision; the first is an impression.

Why do the spend split and the outcome split rarely match?

In most teams the budget split is the product of habit rather than arithmetic. Last year's numbers get a small increase, and a new experiment gets funded by trimming something else. So the spend split is a photograph of the past, while the outcome split is a reading of the present. The gap between the two tells you where movable budget is sitting.

Before you read the size of that gap it helps to know how the budget was built in the first place; if you have never set the allocation from scratch, the order in which a channel-level budget gets assembled is the input this analysis audits. Mix analysis reviews an existing split, it does not invent one.

The gap is informative in both directions. When the share of spend clearly exceeds the share of outcomes, the channel is expensive for what it returns. When the outcome share exceeds the spend share, the channel may deserve more fuel. Be slower with the second case: some channels look efficient precisely because they are small, and whether they hold that efficiency at three times the budget is a separate question.

How do you calculate cost per channel honestly?

Cost per channel is usually calculated by dividing ad spend by record count. That is the easiest version of the sum and the most misleading. An honest calculation adds three items: creative and content production cost, the staff hours needed to run the channel, and the load that channel's records put on the sales team. A channel that produces cheap records while eating a rep's day is more expensive than it looks.

The second correction is on quality. Use qualified records rather than raw records, and keep the definition of qualified identical across channels. The third correction brings revenue in: if two equally priced records convert into deals of very different size, cost alone points you the wrong way. This is where working out which customers actually pay you back feeds directly into the channel decision.

UTM & Source Tracking holds up the denominator of that sum: because each record carries the campaign, ad and channel it came from, the cost calculation is not left to guesswork. Campaign Breakdown then walks the same number from campaign to ad set to a single ad. That is how you find the one badly performing ad inside a channel that looked weak at the top level.

How much does the attribution model change your decision?

Run the same data through two attribution models and the channel ranking can flip. First-touch favours the channels that create demand; social, content and display usually look strong there. Last-touch rewards the channels that collect demand, so brand search and direct traffic swell. Neither model is wrong and neither is complete.

The practical approach is to pick the model that matches the decision. If you are spending to create new demand, weight first-touch. If you want to harvest existing demand more cheaply, weight last-touch. Writing down which model a decision rested on matters as much as the choice itself, because three months later everyone argues from the same definition. Setting up the attribution model behind source analysis is what makes that definition concrete.

The Source Attribution Report in Marketing Hub distributes records, conversions and won revenue by first-touch source. So the report shows you the revenue a source produced, not just the records it delivered. When the mix conversation is held in revenue terms, budget-shift proposals meet noticeably less resistance in the room.

At what threshold do you actually move budget?

The way to take emotion out of a shift decision is to write the thresholds before you need them. The table below is a starting template; tune the numbers to your own sales cycle and volume. What matters is that the figure existed before the meeting did.

Signal	Observation window	Decision
Outcome share exceeds spend share two periods running	2 months	Increase in steps, no more than 20% per step
Spend share is more than 1.5x the outcome share	1 month	Trim, do not cut; change creative and targeting first
Cost rising while qualified records stay flat	3 weeks	Suspect saturation; freeze budget and test
Low volume, thin data	No threshold	Do not decide; extend the measurement window
Broken connection or missing data	Immediately	Postpone the shift and repair measurement first

That last row is the one teams skip. A shift made on broken measurement tracks the measurement error, not the channel. It also pays to store the reasoning next to each decision: six months on, a structured review of the deals you won and lost lets you match those reasons against what actually happened.

How does a channel tell you it has saturated?

Saturation means that adding budget no longer adds proportional results. The symptoms arrive in stages. Frequency climbs first, so the same people see more of you. Then cost per click drifts upward. In the last stage volume still grows but quality falls, and the sales team starts saying the incoming records are not what they used to be. That is a measurement signal, not a complaint.

The first response to saturation is not a budget cut. Refresh creative and targeting: a new message to the same audience often reopens part of the channel's capacity. Screening ad creative before it goes live lets you do that refresh on evidence rather than instinct.

Automated Budget Allocation mechanises the layer underneath: the winning ad's share grows, the one with falling ROAS gets throttled. It does not replace the channel-level decision, but it clears in-channel waste day by day, so the monthly mix discussion starts from a tidier base.

Where do organic channels sit in the mix?

Leaving organic out of the mix means ignoring your cheapest channel. Organic content has no click cost, but it has a production and maintenance cost. Total that cost per period rather than per article and divide it by organic qualified records. The result is usually below paid, but it lags: content written today is measured a quarter from now.

Maintaining the organic side is its own discipline. Refreshing pages that are losing traffic often pays back faster than writing new ones, and an audit that decides which articles need updating puts that maintenance into priority order. The Organic & SEO Traffic section in Marketing Hub reads clicks, impressions, CTR and average position from Search Console for your top queries and pages, so the organic share can be discussed in the same table as paid.

Email usually lands on the organic side too, and usually takes less share than it earns. A send to an existing list converts far more cheaply than an ad to a cold audience. When you are choosing between messaging routes, the cost and reach trade-off between SMS and email settles the decision at that end of the mix.

How often, and in what size steps, should you move money?

The right cadence depends on how long your deals take. Where a deal closes in two weeks, a monthly shift makes sense. Where it takes three months, a monthly shift means deciding before you can see the result, and a quarterly rhythm serves better. The general rule: the measurement window cannot be shorter than the average conversion time.

On step size, small and repeatable wins. Doubling a channel's budget in one move resets its learning phase and destroys the comparison you were relying on. Moving in steps of 15 to 20 per cent lets you read each step separately. If you are trialling a new channel, ring-fence a small fixed percentage of budget as test money and keep that slice out of the shift conversation entirely.

When you add messaging channels to the mix, local habit enters the calculation; the market-by-market split between WhatsApp, Telegram and Viber shows why the same budget has to be allocated differently in different geographies. One mix does not travel.

What goes wrong most often in mix analysis?

The first failure is comparing channels whose measurement quality is not equal. A channel with a leaking data feed always looks weak, and gets trimmed a little more every month. A dashboard that catches channel outages early makes that silent decay visible. The Connection Health Centre serves the same purpose for ad accounts, showing last sync time and data freshness so you can confirm the table is complete before you read it.

The second failure is judging a channel on a single metric. A channel praised for cheap clicks is expensive if those clicks never qualify. The third is ignoring how channels feed each other: a brand seen on social arrives two weeks later through search, and search takes the whole credit. That is why switching a channel off deserves far more caution than turning it down.

The fourth failure is leaving mobile out of the picture. If you have an app, reaching your own installed base is almost always cheaper than buying traffic; treating push and in-app messaging as a channel in its own right makes that share visible. The fifth is the most human: deciding the shift on whoever argues loudest for their own channel. Writing thresholds in advance exists to defuse exactly that.

Performance Alerts tell you when ROAS drops or spend jumps unexpectedly, so the mix conversation does not have to wait for month end. If you want to see your own channel mix on a single funnel, open a free Rocketly account and connect your ad accounts to start.

FAQ

How often should I run a channel mix analysis?

Keep the measurement running continuously, but make decisions on intervals no shorter than your average conversion time. Monthly for short cycles and quarterly for long ones works for most teams.

Should I ever switch a channel off completely?

Usually not. Refresh creative and targeting, then trim the budget. When a channel goes dark you lose the indirect contribution it made to the others, and that loss shows up late in reporting.

Is it fair to compare organic and paid in the same table?

Yes, provided you apply the same arithmetic. Total production and maintenance cost per period, divide by qualified records, and reflect the delay in organic results in your comparison window.

What if my volume is too low to be conclusive?

Do not decide; extend the window. A cost per channel built on a handful of records can reverse next month. Holding budget steady until volume is adequate is cheaper than shifting it the wrong way.