Marketing analytics agency or in-house analyst? How to decide

The board wants a number every quarter, the marketing team is guessing at it from four tools that disagree, and someone has suggested either hiring an analyst or bringing in a marketing analytics agency. The useful answer is that the choice depends on which of three jobs you need done, and the job most firms are short of is the one neither route does by default.

This piece sets out the three jobs, which an in-house analyst does well and which an agency does well, what the money looks like in UK terms, the hybrid most mid-market B2B firms settle on, what breaks when no one owns tracking, and how to brief whichever route you take.

What are the three jobs an analytics function does?

Job one is measurement setup and integrity. Tags fire, consent is handled, GA4 and the CRM agree on what a lead is, the ad platforms import the conversions that matter. Plain English: the numbers can be trusted.

Job two is reporting that answers a business question. Not "how many sessions did we get" but "did the June campaign pay back yet" or "which service line is cheapest to win a qualified lead in". The report is built backwards from the question.

Job three is analysis that changes a decision. Somebody looks at the reconciled data, notices that paid search is buying clicks on terms the site already ranks for, or that the "warm" leads from a free download never close, and a budget moves as a result. This is the job most firms are actually short of, and it's the one that gets squeezed when the same person is also fixing tags.

Put the three in priority order and the logic follows. Job one has to exist before job two means anything, and job two has to exist before job three has anything to work from. Most of the disappointment I see with analytics, in-house or agency, comes from starting at job three with data that never passed job one.

Which of those does an in-house analyst do well?

Jobs two and three, given time and access. An in-house analyst sits in the Monday meeting, knows that the sales director counts a lead differently from the marketing lead, and can chase the CRM admin across the desk. Context is the advantage. No external team gets that much of it.

Job one is where in-house often struggles, and not through any lack of skill. Measurement setup is lumpy work: a fortnight of GTM, consent mode and server-side tagging when the site changes, then very little for six months. Keeping one person current on that when they touch it twice a year is hard. It's also where an agency has seen the same failure on forty sites and recognises it in an hour.

The other in-house limit is cover. One analyst is one person. When they're on leave in the week the board pack is due, the pack is late, and when they leave the company the logic tends to leave with them unless it's been written down somewhere other than their head.

Agency versus salary: what do the numbers look like?

A mid-level marketing analyst in the UK outside London is advertised at £38,000 to £55,000 depending on region and how much SQL and GA4 depth the role needs. Add employer National Insurance, pension contributions, tooling and a share of management time and the real cost is roughly 20% to 25% on top, so £46,000 to £69,000 a year, or £3,800 to £5,700 a month. That buys around 220 working days of one person's attention, most of it on jobs two and three.

A marketing analytics retainer for a mid-market B2B firm, in our experience, sits between £1,500 and £4,000 a month for measurement upkeep and monthly reporting, with a one-off setup or rebuild at £3,000 to £8,000 in front of it. That buys three to eight senior days a month across two or three people, most of it on job one and job two. It does not buy someone in your Monday meeting.

So the comparison isn't like for like, and I'd be wary of any version of it that pretends it is. The salary buys presence and context. The retainer buys specialist depth and cover. If you have fewer than about 50 leads a month and one or two channels, the retainer alone usually does the job. Above that, with sales and marketing arguing over attribution, the in-house seat starts to earn its cost.

The hybrid most mid-market B2B firms end up with

An in-house owner of the questions and an external owner of the plumbing.

In practice that looks like one marketing operations person, often not a pure analyst, who decides which five numbers the board sees and chairs the monthly review. An agency builds and maintains the tracking, reconciles GA4 with the CRM and the ad platforms, and produces the report that person presents. The agency's hours are the lumpy ones. The in-house person's are the constant ones.

The break-even example above ended up in exactly this shape. We treated the three courses as one campaign, segmented by location, set a tiered goal for each city, and pulled it into one Looker Studio view. The client's marketing manager owned the goals. We owned the pipes. Neither side could have done it alone in the time.

A candid point about our side of that arrangement. The most common output of both routes, agency and in-house, is a dashboard that gets opened twice. We've built a couple of those ourselves early on. The cause was always the same: we started from the data we had rather than the decisions the client needed to make. The fix is to write the decisions down first (which budget moves, which channel gets cut, which service line gets a page) and build only the views that feed them. A dashboard with four numbers that a director reads every Monday beats one with forty that stays closed.

What breaks when no one owns the tracking?

Four things, and they compound.

GA4 and the ad platforms disagree. Google Ads reports a conversion on the date of the click and gives itself full credit; GA4 reports it on the date it happened and shares credit across channels. A gap of 20% to 40% between the two is normal and explainable. Without an owner, the gap becomes an argument in the board meeting instead of a footnote. Our piece on why GA4 and ad platform numbers differ goes through the reconciliation.

Consent mode quietly changes what's counted. Since Google's consent requirements tightened, a site with a cookie banner and no consent mode configured can lose a third or more of its measurable conversions. Modelled conversions fill some of that back in, but only if the setup is right, and the setup is job one.

The CRM sync does something you didn't expect. Just so you know, for reference: a HubSpot sync we looked into recently only synced live leads, so anything already closed never updated in the destination. Nothing had been done wrong. The setting was documented, but not by anyone who'd read it in a year. That's what an unowned setup costs: not a crash, a slow drift between two systems that both look fine.

Sampling misleads at scale. GA4 samples when an exploration passes 10 million events, and it estimates the rest from the first batch. For a small site that never happens. For a large one, assuming the sample is right can be wrong by enough to move a budget decision. Server-side tagging and BigQuery export are the usual answer, and both are cheaper than the wrong decision.

The reassuring part: all of this is fixable, usually in a fortnight of focused work. The previous setup on the accounts we inherit is normally largely sound. What's missing is the logic that says what a lead is and when it's counted, and that's a decision, not a rebuild.

How do you brief either option?

Start from the five questions the board asks every quarter, and give them to whoever you're hiring as the specification. In most B2B firms they are:

An in-house candidate who can explain how they'd answer question four from your current systems, and what they'd need to change first, is the one to hire. An agency whose proposal opens with those five questions, and states which of the three jobs it will own, is the one to shortlist. A proposal that opens with a list of tools is describing job one and hoping you don't ask about the rest.

One more thing on the brief. Put a revenue figure next to it. "We spend £40,000 a month across paid and organic and can't say which half works" is a brief. It tells both the analyst and the agency what a good year looks like, and it makes the fee, or the salary, easy to justify against it. Our guide to marketing analytics and commercial impact covers how to put that number together.

Questions we get asked

What does a marketing analytics agency do? Three things, in varying proportions: sets up and maintains measurement (GA4, GTM, consent, server-side, CRM sync), builds reporting that answers business questions, and provides analysis that changes budget decisions. Ask which of the three a given agency actually spends its hours on.

Do we need a data analyst or a marketing analyst? A data analyst is stronger on SQL, warehousing and modelling; a marketing analyst knows what a lead source is and why Google Ads and GA4 disagree. For a mid-market B2B firm the marketing analyst is usually the right first hire. Bring in data engineering skills once you're joining three or more systems.

How much does a marketing analytics consultant cost in the UK? Independent consultants run £400 to £900 a day. Agency retainers for measurement and reporting sit around £1,500 to £4,000 a month, with setup work at £3,000 to £8,000. Full-time analysts cost £46,000 to £69,000 a year once on-costs are included.

Can our SEO or PPC agency also do our analytics? Often, and it's usually the cheapest way to get job one done properly, because they need the tracking to be right to do their own work. The thing to check is whether they'll report against a commercial target that includes channels they don't run. Our piece on marketing attribution explains what to look for.

Is GA4 enough? For a small site with one or two channels and a simple lead form, GA4 plus a properly configured CRM is enough. Once you're running paid, organic and events into a sales pipeline, GA4 on its own can't tell you what closed, and you'll need the CRM joined to it before the reporting means anything.

Where we fit

We're a small search and CRM consultancy that builds the measurement and reporting side for B2B firms, usually alongside an in-house marketing lead who owns the questions. For Hubject, a budget modelling exercise built on reconciled data saved £40k, and for EWA we centralised 100% of their customer data into one CRM so the reporting could start from something true. Our data and analytics work is described in more detail on that page.

If you're weighing the hire against the retainer, send us the five questions your board actually asks. We'll tell you which of the three jobs we'd take and which we think belong in-house.