Digital marketing planned around measurement, not guesswork

Every channel decision reduces to one number

Before a single campaign is built, there is a number that determines whether any of this can work: what you can afford to pay to acquire a customer. It comes from gross margin, repeat purchase behaviour and close rate, and it is remarkable how often a business spending substantially on acquisition has never written it down. Without it, a cost per lead is not information. It is a figure with no context, and every argument about channel performance becomes an argument about taste.

So that is where a digital marketing engagement starts here. Not with a channel recommendation, which is easy and mostly interchangeable between agencies, but with the arithmetic that tells us which channels are even available to you. A business with high gross margin and strong repeat purchase can afford to lose money on a first order and can therefore compete in auctions that a low-margin, single-purchase business must simply avoid. Two competitors in the same category can have entirely opposite correct strategies for this reason alone.

The second number is how long your buying cycle is, because it determines how quickly you can learn anything. An ecommerce brand converting daily accumulates enough evidence to make confident decisions relatively quickly. A business selling a small number of large contracts a year does not, and pretending otherwise leads to the most common failure in performance marketing: changing a campaign because of a fortnight of noise.

What an engagement actually contains

The word “digital marketing” covers a genuinely wide field, and agencies use its breadth to avoid being specific. Concretely, an engagement here is some combination of five things: a strategy layer that decides what to do and what to skip, an acquisition layer that runs the channels, a conversion layer that improves what happens after the click, a measurement layer that makes the first three arguable with evidence, and a reporting layer that turns all of it into a decision you can act on.

Very few clients need all five at full weight. A business with a strong in-house team and broken analytics needs the measurement layer and nothing else. A business with excellent measurement and no traffic needs acquisition. Scoping honestly means saying which of the five you actually need, and it is the part of the sales conversation where agencies most reliably overreach.

Strategy is mostly the decision not to do things

A strategy document that recommends being present on every channel is not a strategy; it is a menu. The useful version states which audience you are pursuing, what you are offering them that a competitor is not, which two or three channels can reach them at a viable cost, and what you are deliberately declining to do this year. The declining part is what makes it a plan. It is also what makes it uncomfortable to present, which is presumably why it is so often omitted.

Choosing channels by business model, not by fashion

The single most useful distinction in channel planning is between capturing demand that already exists and creating demand that does not.

Demand capture means intercepting people who are already looking. Search is the obvious form: someone types a query that describes their problem, and the market that answers it is an auction plus an organic result set. The economics are usually favourable because intent has already done the persuasion for you. If your category has meaningful search volume, this is nearly always the cheapest place to prove that your offer converts, which is why we generally recommend testing with paid search and shopping campaigns before committing to a long organic programme. Paid tells you within a manageable budget whether the offer works. Organic then holds the position at lower marginal cost.

Demand creation means reaching people who are not looking yet — paid social, content, partnerships, community, PR. The cost per acquisition is higher and the feedback loop is slower, because you are paying to interrupt rather than to answer. It is unavoidable when the category is new, when the need is latent, or when search volume is simply too small to build a business on. A great deal of wasted budget comes from applying demand-capture expectations to demand-creation channels and concluding after one quarter that “social does not work for us”.

Most businesses need one of each, funded to a level where the result is legible. What they usually have instead is a thin presence on five channels, none of which has accumulated enough data to justify its own existence.

Paid media, and its honest limits

Paid media is the fastest way to learn what your market thinks of your offer, and the fastest way to spend money confirming that nobody has told the sales team the leads are arriving. Modern platforms have moved most of the tactical levers into automated bidding, which means the work has shifted decisively towards the inputs: what conversion you optimise for, what value you assign it, how the account is structured, what creative you feed it, and what the landing page does.

That shift matters more than it sounds. When you use a value-based strategy such as target return on ad spend, you are telling the platform to predict the value of each conversion and bid accordingly — as Google documents, ROAS is simply conversion value divided by spend, expressed as a percentage. If the conversion value you are sending is wrong, or every lead is passed back with the same nominal value regardless of quality, the machine optimises confidently towards the wrong thing. Most underperforming paid accounts we look at are not badly managed in the traditional sense; they are well managed against a badly defined goal.

The limit worth stating plainly: paid media cannot fix a weak offer, a slow response time or a landing page that contradicts the ad. It will find you the ceiling of your current proposition faster than anything else, which is genuinely valuable, but it is a diagnostic as much as a growth channel.

The compounding half

Organic search, content and email behave in the opposite way. They cost more before they return anything and less afterwards, and their value accrues to an asset you own rather than to a media buy you rent.

Search engine optimisation is the largest of the three for most businesses, and the part of the programme most often started too late. Content is what makes it work: Google is explicit that its systems try to reward content demonstrating experience, expertise, authoritativeness and trustworthiness, which in practice means content written by someone who has actually done the thing. A content programme built on that basis is slower and more expensive per article than the alternative, and it is the only version that still works.

Email is the least fashionable and frequently the highest-return channel in the mix, because it addresses an audience that has already identified itself. It also has the useful property of being unaffected by an algorithm change. If you have an existing list you have not mailed properly, that is often the cheapest available improvement to the whole programme.

Where the product is a mobile app, the equivalent compounding channel is app store optimisation, which follows the same logic on a different surface: get the listing right once and it keeps earning installs without per-install cost.

Landing pages are where most budget quietly dies

An improvement in conversion rate applies to every visitor from every channel simultaneously, which makes conversion work the highest-return part of a programme and the most consistently under-resourced. The failures are unglamorous: the ad promises one thing and the page describes another, the page takes too long to render its main content on a mobile connection, the form asks for a phone number before it has earned one, or the primary action is below three paragraphs of company history.

A word on testing, because it is oversold. A/B testing requires enough traffic to reach statistical significance, and the Nielsen Norman Group is direct that low-traffic pages are simply not testable this way — you need thousands of users, a test long enough to cover the natural rhythm of your traffic including weekends, and the discipline not to stop the moment the result looks good. Below that threshold, the honest approach is to make well-reasoned changes based on qualitative evidence — session recordings, user testing, what your sales team is asked repeatedly — and accept that you are exercising judgement rather than measuring. Agencies that run A/B tests on pages with insufficient traffic are producing decoration, not evidence.

Measurement, and what attribution can and cannot tell you

This is where we spend more time than most, because everything else depends on it.

The mechanics are straightforward enough: a Google Analytics 4 property configured against a written measurement plan, a Google Tag Manager container implementing it, consistent campaign tagging using the standard utm_source, utm_medium, utm_campaign and utm_id parameters, and conversions reconciled against the CRM rather than assumed. Where data quality or control justifies it, server-side tagging moves collection to a server container, which improves reliability and gives you more control over what is sent to which vendor.

The judgement is harder. Two things are worth understanding before you read another marketing report.

Consent changes what your numbers are

Under GDPR and equivalent regimes, visitors decline tracking, and a meaningful proportion of them do. Consent Mode handles this by adjusting parameters such as analytics_storage and ad_user_data according to the visitor decision, and by modelling the behaviour of the visitors who declined. That is a reasonable design, but it has a consequence people rarely state: some of the conversions in your reports are estimates, not observations. We label them as such. Any agency presenting post-consent conversion counts as precisely measured is either not paying attention or hoping you are not.

Two dashboards will never agree, and that is not a bug

Google Ads and GA4 will report different conversion counts for the same campaign, permanently. They use different attribution models, different conversion windows and different definitions of what counts. GA4 offers data-driven attribution, paid and organic last click, and Google paid channels last click, and each will tell you a different story about the same month. The mature response is to pick one model as your decision-making basis, state it in the reporting, and stop trying to reconcile the others. Attribution is a lens, not a measurement, and treating it as a measurement is how a well-performing top-of-funnel channel gets cut for underperforming on a last-click view. If your setup needs more than a dashboard — modelled channel contribution, warehouse-level joins, incrementality testing — that is data and analytics work rather than campaign management, and we would rather scope it as such.

Reporting: what we send, and what we leave out

Every engagement gets a live dashboard and a monthly written analysis. The analysis contains four things: what we changed, what happened, what we are doing next, and what did not work.

What it does not lead with is impressions, reach, engagement rate or follower growth. Those numbers have their uses in diagnosis — an organic social programme genuinely does need audience metrics to steer by — but as headline reporting they mostly function to make an unproductive month look energetic. If the commercial numbers are flat, we would rather say so and explain why in a paragraph than distribute a slide of green arrows.

Budget allocation and the discipline of stopping

Budget should move towards evidence, on a defined cadence, in increments large enough to matter. Two rules do most of the work. First, do not reallocate on noise: if the volume of conversions is small, a bad fortnight is not a signal. Second, decide in advance what result would cause you to stop something. A test with no stated failure condition is not a test; it is a commitment with a review meeting attached.

The corollary is that some things should be killed while they still feel promising. This is the hardest conversation in the discipline, because by the time a channel has been running long enough to judge, someone internally has become attached to it. Writing the stopping condition down at the start is the only reliable protection against that.

When to bring marketing in-house

We will say this plainly, because it is the question a good client eventually asks. Bring a channel in-house when it is predictable, high-volume, and central enough to your business that the knowledge should not sit outside it. Day-to-day campaign management, email and organic social all reach that point for many businesses, and paying an agency retainer to perform a stable routine is poor value.

Keep external the things that are intermittent or that need a depth of specialism you cannot keep busy — technical SEO, analytics architecture, a major migration, creative production at variable volume. And be realistic that in-house is not cheaper by default: one marketer covering six channels is usually worse at all of them than a focused team plus a specialist supplier. We would rather help you build that team than quietly obstruct it, and it is worth asking any prospective agency how they answer this question before you sign anything.

How this connects to the rest of the work

Digital marketing is the layer that decides where effort goes; the specialisms underneath it do the work. Organic growth runs through search and the content that supports it. Paid acquisition runs through campaign management. Audience building sits with organic social. And all of it depends on measurement that holds up under a finance director asking where the money went — which is why we treat the analytics layer as the foundation rather than as a reporting afterthought.

The corollary is that not every client needs the whole programme, and the scoping conversation is where that gets decided honestly or dishonestly. If you are not sure which parts apply to you, that is a reasonable place to start. Tell us what you are trying to change and we will say which of this is relevant — including, where it applies, that the answer is fewer things than you expected.