How to prove marketing is working to your CFO, CEO or leadership.

Cost per acquisition tells you what you paid, not whether you’re winning.

Paid advertising is one of the more expensive lines in the budget, so sooner or later you end up working with your CFO on numbers they can take to the CEO, the board or investors. We all know how tricky that gets once you try to connect spend to performance. Sometimes you can connect the dots, but often it isn’t as clean as anyone would like, and someone outside marketing has a harder time accepting indirect impact and numbers that aren’t revenue.

So the story ends up looking familiar. It explains what you spent, which is easy to quantify, how many leads you tracked, and what each one cost, maybe with a trend line showing where that’s heading. People nod, because there’s a metric and a trend and a story that goes with it.

But the question everyone in that room is actually asking is whether the money is working and whether you’re winning. Cost per lead answers a narrower question, which is what the leads cost. It’s a necessary number and I’m not saying to remove it, but it’s an efficiency number. It tells you nothing about your position in the market, or what you’re paying to hold it.

There are plenty of ways to show marketing is working, from revenue attribution to incrementality tests to marketing mix models. Most of them depend on your own tracking, your own data or your own models, and that’s exactly where the arguments start. Attribution in particular is more art than science, and anyone who wants to question the number can. The measure I’d add is share of demand. It tells leadership whether you’re winning, and the core of it comes from public search data your CFO could pull and check for themselves, without having to trust your tracking, your attribution model or you.

Cost per lead can improve while you’re losing.

Imagine a company with four competitors.

In the first quarter the company spends $63,000 a month and gets 300 leads, so a cost per lead of $210. Over the next two quarters the team trims some expensive category terms from its paid search campaign. Ad spend drops to $54,000 a month, leads hold at 300, and cost per lead falls to $180.

Everyone’s pleased, and reasonably so.

Meanwhile, here’s what’s happening to share of search:

Those expensive category terms were where competitor 2 kept showing up. When the team paused them, visibility on category searches fell and competitor 2’s rose, and six months later more buyers are searching for competitor 2 by name. The company has given ground to a competitor while its efficiency numbers improved.

Nobody made a mistake here. Trimming expensive terms is a reasonable call, and a reduction in cost per lead seemed to confirm it. The trouble is that cost per lead can only see the buyers who reached you. The ones who went to competitor 2 are invisible in the reporting. A buyer who chose someone else looks exactly like a buyer who never existed.

The trouble with tracking.

Cost per acquisition and cost per lead only mean something if you believe what’s behind them. You have to believe the tracking is accurate, that the attribution model is working properly, and that the click, the lead, the CRM record and the ad that got credit all line up.

That’s getting harder to believe. GDPR, consent banners and cookie blocking have all weakened attribution, and the platforms fill the gaps with modeled conversions. Your ad platform says one number and your CRM says another. So when a CFO questions your cost per acquisition, you can find yourself in a debate about attribution models, and nobody wins that debate.

I think this gets worse before it gets better (if it gets better). Privacy rules are diverging by country, and in the United States increasingly by state, so a company selling in California, the rest of the US and Europe is effectively working with different levels of visibility in each. Modeling is the platforms’ attempt to fill those gaps statistically. What it gives you is directional cues rather than reality, and when you’re making significant decisions about where the money goes, you want to be working from the best numbers you have, not an estimate from a black box you can’t see into.

So what if part of the reporting didn’t depend on any of that, and leadership could check it themselves?

What share of demand measures.

By demand I mean everyone who’s looking to buy what you sell, including the people who haven’t decided who to buy from yet. Share of demand is how much of that you’re capturing.

That’s hard to measure directly, because these are people you aren’t in contact with and whose clicks you may never see. The traditional approach was surveys: sample the market, build a statistical model of who’s in market, and estimate what percentage was looking at brand A, brand B and brand C. It was slow and expensive, and search data now gets you a usable version of the same picture. It comes in two parts.

Category visibility. When a buyer searches for what you sell rather than a specific brand, how often do you show up compared with your competitors? These are the undecided buyers, and this is the input. Your ad account may already report this as the share of eligible category searches where your ads appeared, along with which competitors appeared alongside you and how often. It’s built from auction data rather than from following anyone to a sale. The limit is that it only covers the searches you’re bidding on, so it describes your visibility on the terms you chose, which is one more reason to keep that list stable from quarter to quarter. Leadership can’t pull this one themselves, but it comes from the ad platform’s auction data, not your tracking.

Share of search. Searches for your brand divided by searches for all the brands in your named set of competitors. This is the outcome. It’s where category demand ends up once buyers have a shortlist, whatever they researched along the way, and it shows which brands the market has settled on and how they compare. This is the part leadership can check for themselves. Google Trends is free and public, so anyone can enter the same brand names and get the same lines you’re showing them.

Share of search comes from Les Binet’s work, presented at the IPA’s EffWorks conference in 2020. Looking at cars, energy and phones, he found it actually tracked market share fairly closely and tended to move first. When a brand’s share of search rose, its market share tended to follow, with a lead of up to a year for cars. He was also clear that it isn’t a perfect predictor, because things like price affect whether interest turns into a purchase. Someone searching for a brand isn’t necessarily buying from it.

His research came before the recent growth of AI search, and it covered big consumer categories, so for the rest of us it’s directional evidence rather than proof. I still think it’s the strongest footing available at present.

Category visibility and share of search tell you more than either does alone. In the scenario above, category visibility fell first and share of search followed. Visibility can also hold steady while share of search slips, which usually means someone is winning buyers later in their research, somewhere your ads never appear.

What makes both sturdier than cost per lead is that they’re built from searches in aggregate. You don’t need to follow anyone from a click to a sale, so you don’t lose resolution to consent banners or cookie loss. As attribution gets weaker, these don’t.

Present cost per acquisition alongside share of demand.

Show the numbers side by side. Present them quarterly as a trend. Cost per lead or cost per acquisition goes in one column, and category visibility and share of search go in the next two, measured against a fixed set of competitors you name in the report so that anyone could rebuild the same analysis and get roughly the same answer.

Then read them together. When costs are falling and both share numbers are rising, that’s a good quarter all around. When costs are rising but share is rising with them, it may be a fair price for winning. The one to worry about is costs falling while share falls, because that’s the quarter where you’re congratulating yourself on efficiency while losing the market.

Once you have a baseline, you can start relating spend to share over time. “This year we spent this much and moved share of search from 19% to 22%” is a more interesting way to justify a budget than cost per lead alone. Treat it as directional, though. Plenty of things move share besides your spend, including whatever your competitors are doing, so it’s a relationship you watch over several quarters rather than a price per point you can promise.

And keep your confidence attached to the number. “Directional, based on Google’s data, against these four competitors” is something a CFO can work with.

Efficiency numbers on their own tell people what you paid. Put them next to your share of the market and you’re showing them whether it worked.

The questions a CFO will ask:

Can we get this data from SEO tools? Tools like SEMrush, Ahrefs and SpyFu are useful for finding which keywords exist and roughly how they rank against each other. They get much less reliable once you use them to compare volumes between brands. I’ve checked their traffic and keyword estimates against the analytics of sites I have access to and seen them diverge significantly. I stopped quoting competitor volume or making traffic estimates from this data years ago. The problem is that they don’t have anyone’s actual data. They’re modeling from panels and browser plugins, and the modeling error isn’t the same for every site. It can vary with a site’s size and with how many categories it competes in, so if they’re 30% off for you and 60% off for a competitor, the shares are wrong too. I’d rather get as close to the source as possible, which for search means Google’s own Keyword Planner and Google Trends. Those aren’t perfect either, since Trends is sampled and indexed and Keyword Planner rounds, but the distortions apply to every brand in the comparison the same way, and that’s what a share calculation needs.

How did you pick the competitors? They’re the brands a buyer would realistically shortlist. Leave out the ones that barely register, say 0.5% against your 20% share. Look at them again every quarter in case one starts to climb. The list stays mostly fixed, because share only means something when you’re comparing against the same names, and the report should say “share of the named set” rather than claiming to describe the whole market.

One competitor sells ten different things. Doesn’t that make it lopsided? It would if you compared against their company name, because they get brand searches for everything they sell. Use the brand’s product name, or their brand plus the product if needed. That will undercount them a little, since some people search the company name and navigate from there, but a bias you’ve stated, and whose direction you know, keeps the number defensible.

Isn’t search moving to AI? Some of it is, and that weakens absolute search volume figures. Share of search holds up as long as the shift affects you and your competitors about equally. If a fifth of the market starts researching in AI assistants, the split among the people still searching on Google probably won’t change much. At the moment that is still an assumption, and there’s no reliable AI search reporting yet to test it against.

How does share of search connect to revenue? Not directly, and you shouldn’t claim it does. It’s a leading indicator, closer to pipeline than to revenue. Binet’s evidence is strongest in large consumer markets, so for considered purchases expect it to hold directionally and check it against your own history. You can go a step further by tying branded search traffic to the revenue it produces, as long as everyone understands that part depends on your attribution again.

What if we’re too small for this? Low volume makes share of search noisy, and very small brands may not register in Trends at all. Looking at quarterly or yearly data instead of weekly smooths out a lot of it. Keyword Planner can give you actual search numbers for your brand and your competitors’, though it may show only ranges unless the account is actively spending. And if nobody searches for you by name yet, a small brand campaign works like a fishing rod. It sits there doing nothing until people start looking for you, and then you see it move. A smaller company can often match spend to revenue directly anyway, which is a luxury bigger ones lost a long time ago.