What Does a Marketing Agency Do?

Updated September 2026 · Written and maintained by the Progression Agency strategy team

Not what they say they do — what the work actually consists of, task by task, discipline by discipline, hour by hour. What happens in a week and in a month on a real account, who inside the agency does which part, what a $4,000 retainer buys in actual human time, which activities are genuine work and which are theater, and the twelve-point monthly check that tells you which kind you are paying for.

The short answer

A marketing agency is a way of renting six or seven specialists you could not justify hiring individually. Everything else follows from that — including why the fee looks large relative to a salary, and why the hours look small relative to the fee. Roughly 48% of what you pay is somebody’s wage, 13–15% is the agency’s net margin at a healthy firm, and the rest is overhead, tools, non-billable time and the cost of winning your business in the first place.

What that buys is about thirty senior-equivalent hours a month at $4,000. The whole question of whether an agency is worth it comes down to whether those thirty hours, applied by people who have seen your problem before, produce more than the alternatives. The pricing arithmetic is on marketing agency pricing; this page is about what happens inside the hours.

The four jobs, and the actual work inside each

Every task an agency performs sits inside one of those four boxes. It is a useful frame because it immediately exposes what is missing: a program with no first box is guessing, one with no second box is busy, one with no fourth box has launched something and walked away — and that last failure is the most common of the three.

Finding out what is true

Audits, analytics implementation, customer research, competitive analysis, keyword and demand research, and your own unit economics. It is unglamorous, it always comes first, and it is routinely skipped because it produces no visible output in month one. An agency that starts producing content before it knows your close rate and average customer value is producing content, not marketing.

Deciding what to do about it

Strategy, positioning, channel selection, budget allocation, prioritization and forecasting. This is the part that is hardest for a client to evaluate and therefore easiest to fake, which is why so much agency language accumulates around it. A real strategy is a small number of specific decisions with reasons attached and things explicitly excluded. A fake one is a deck.

Making the things

Content, creative, code, campaigns, emails, landing pages, video. The visible half of the work and usually the majority of the hours. Most of what you will see month to month lives here.

Running and improving them

Bid management, testing, optimization, publishing, monitoring, responding and reporting. This is what separates a working program from a launched one, and it is the box that quietly empties in month seven when the initial build is done and nobody has decided what happens next.

Where the hours actually go

That is a healthy allocation on a multi-channel engagement. Two things about it are worth noticing. First, account management is genuine time — your calls, your emails, and chasing your approvals — and it is time you influence directly. Engagements stall on client-side approvals far more often than on agency capability. Second, technical and site work is only useful if the agency can actually ship the change rather than write a recommendation and wait.

The recommend-versus-ship gap

A great many agencies can find a technical problem and cannot fix it. The audit identifies thirty-four issues, six get fixed, and the rest sit in a document for two years. Ask directly, before signing: do you ship code, or do you write recommendations for somebody else to ship? Both are legitimate answers, but if it is the second, somebody has to be the somebody, and it should be named in the scope.

What each discipline actually involves

Below is the real task list, by discipline. If you are evaluating a proposal, this is what should be underneath the headings — and if you are doing it yourself, this is the job.

Strategy

• Interview you properly: revenue by service line, margin, close rate, capacity, seasonality, what you have already tried and what it cost
• Establish unit economics — what a customer is worth and therefore what a lead can cost
• Define the single number the engagement will be judged by at twelve months
• Select channels on the basis of where your demand actually exists, not on what the agency sells
• Allocate budget across channels with a stated rationale
• Build a rolling 90-day roadmap with an expected outcome against each item
• Decide explicitly what you will not do, and say so
• Forecast, then compare the forecast against reality each month and explain the gap
• Reprioritise monthly on the basis of what the data actually showed

Research and discovery

• Technical audit of the site, naming specific URLs and specific problems
• Analytics audit: is anything being measured, and is it measuring the right thing
• Competitive analysis of the businesses actually ranking, not the ones you named
• Keyword and demand research in your real geography, not nationally
• Search intent verification against the live results page for every target term
• Customer research: sales call transcripts, support tickets, the words customers actually use
• Review mining — yours and your competitors’ — for objections and language
• Content gap analysis against the pages that rank
• Backlink profile audit for both opportunity and risk

Analytics and measurement

• Install or repair Google Analytics 4 and verify it fires correctly
• Define conversion events that match real business events, not page views
• Test every conversion end to end, from a real phone
• Install call tracking with a duration threshold so short hangups do not count
• Deduplicate: one lead that fills a form and then calls should not count twice
• Connect the CRM so closed revenue can be attributed back to landing pages
• Import offline conversions into ad platforms so bidding optimizes toward revenue
• Verify Search Console and Bing Webmaster Tools
• Build reporting that reconciles to your accounting system within a tolerable margin
• Document the baseline before any work begins

Search engine optimization

• Fix indexing problems on commercial pages before anything else
• Enforce one canonical hostname; resolve trailing-slash and parameter duplicates
• Build a keyword map with exactly one URL per intent, and resolve cannibalization
• Ship technical fixes: redirect chains, orphan pages, crawl depth, structured data
• Measure and improve Core Web Vitals from field data on mobile
• Write and publish substantial pages against verified demand
• Build contextual internal links between related pages
• Rewrite titles and descriptions on pages with impressions and no clicks
• Refresh and consolidate existing pages rather than only publishing new ones
• Acquire links through associations, suppliers, local press and genuinely linkable assets
• Monitor Search Console weekly and act on what it shows

Local search

• Audit the Google Business Profile primary category against those ranking above you
• Populate every field: services with descriptions, products, attributes, hours, holiday hours
• Upload real job photographs weekly; publish posts; seed and answer Q&A
• Point the profile’s website link at the most relevant page
• Fix NAP consistency across the fifteen citations that matter and remove duplicates
• Build and automate a review request process, by text, within a day of the job
• Respond to every review within seventy-two hours
• Write genuinely distinct location and service pages
• Track rank from a grid of coordinates, not a single point

Paid media

• Structure the account so budgets, geographies and intents are separable
• Set location targeting to presence rather than presence-or-interest
• Read the search terms report weekly for the first eight weeks and build negatives
• Write and test ad copy against a stated hypothesis, not at random
• Build landing pages that match the query rather than sending everything to the homepage
• Manage bids and budgets against a target cost per acquisition you agreed together
• Rotate creative before it fatigues, which on paid social is a matter of days
• Set up and monitor conversion tracking with duration thresholds
• Import offline conversions so bidding optimizes toward closed revenue
• Review the whole account against the target quarterly and cut what is not working
• Tell you when to spend less, which is the test of whether the relationship is honest

Content

• Brief writers properly, with the intent, the competing pages, and what must be covered
• Interview subject matter experts inside your business rather than paraphrasing competitors
• Write pages that contain something the ranking pages do not already contain
• Structure with question-shaped headings and direct answers underneath
• Include specific numbers, named places, real examples and cited sources
• Produce or source real photography rather than stock
• Edit — genuinely, not a spellcheck
• Publish, internally link, and submit
• Track performance per page and refresh what stalls at position eight to twenty
• Consolidate or remove what never performs

Creative and design

• Produce ad creative in volume, because platforms consume it faster than budget
• Design and build landing pages that match campaign intent
• Maintain a usable design system rather than reinventing each asset
• Produce and edit video, including cutdowns for each placement and aspect ratio
• Art-direct photography
• Design the assets nobody remembers to scope: email templates, social formats, one-pagers
• Version and archive files so you can actually find them later

Web development

• Ship technical SEO fixes rather than documenting them
• Build and maintain landing pages
• Implement tracking, tags and server-side measurement
• Improve performance: image formats, script auditing, caching, hosting
• Fix accessibility issues against WCAG 2.2 AA
• Maintain the CMS, plugins, security and backups
• Handle migrations, including the redirect map that decides whether you keep your traffic

Social media

• Decide honestly whether organic social is a channel for your business or a formality
• Produce content designed for the platform rather than repurposed from elsewhere
• Publish on a cadence the audience actually rewards
• Respond to comments and messages within a window that matters
• Run paid amplification, because organic reach on a business page is a fraction of the audience
• Report on the things that matter — traffic, leads, saves — rather than on follower counts

Email, SMS and lifecycle

• Build the flows that produce revenue without new traffic: post-purchase, abandonment, seasonal reminders, quote follow-up, reactivation, referral requests
• Segment on behavior rather than on demographics
• Maintain list hygiene and deliverability
• Write and test subject lines and offers
• Handle consent properly, particularly for SMS where the TCPA is not permissive
• Report on revenue per recipient rather than on open rates

PR and communications

• Build relationships with journalists who cover your sector, before you need them
• Find and pitch stories that are genuinely stories
• Produce original data worth citing
• Respond to journalist requests
• Handle reactive and crisis communications
• Reclaim unlinked brand mentions

Conversion rate optimization

• Test the contact form from a real phone and fix it when it is broken
• Audit the conversion path on the pages that actually receive traffic
• Form a hypothesis, test it properly, and accept the result
• Improve mobile performance, because most of the traffic is mobile
• Add trust signals, price transparency and response-time commitments
• Improve the offer itself, which is usually a larger lever than the page design
• Track and report the conversion rate as a first-class metric alongside traffic

Account management and reporting

• Maintain a change log listing every action taken, with dates
• Produce a monthly report that leads with qualified leads and blended cost per acquisition
• Explain what changed and why, including when it got worse
• Report what was tried that did not work
• Keep the roadmap current and move it when the data says to
• Chase your approvals, because that is where projects stall
• Be reachable, and be honest when the answer is that they do not know yet

A week inside an agency, on your account

Monthly calls give you a fortnightly-to-monthly view of something that is happening daily. Here is what the daily version actually looks like on a healthy engagement, and it is worth knowing because it lets you ask a much better question than ‘how is it going’.

Monday: check what changed

The first hour of the week goes on looking at what moved. Search Console queries compared week on week with brand terms filtered out. Ad account spend and conversions against the target. Grid rank positions if local matters. Any errors, alerts, or drops. Anything falling sharply gets investigated the same day rather than appearing as a line in a report three weeks later.

This is the routine that catches almost everything before it reaches revenue: a broken tag, a plugin update that removed a noindex or added one, a competitor who published something better, a form that stopped delivering, a Google update. Five minutes of attention on a Monday is worth more than five hours of investigation in month three.

Tuesday: paid media maintenance

Read the search terms report and add negatives. In a new account this is genuinely weekly work for the first two months, because the volume of irrelevant matched queries is high and each one you exclude is money returned. Pause what is not converting, shift budget toward what is, check that creative has not fatigued, and confirm conversion tracking is still firing correctly.

An account left unattended for a month does not fail dramatically. It degrades: match types drift wider, irrelevant queries accumulate, creative fatigues, and the cost per acquisition rises ten or fifteen per cent without anything obviously breaking. That gradual drift is what management fees actually buy.

Wednesday: production

The largest block of the week. Content written, edited or briefed. Pages built. Creative produced. Technical fixes shipped to the live site rather than added to a document. This is the visible half of the engagement and it is where most of the hours go, which is appropriate — a program where production is a minority of the time is producing meetings.

Thursday: local, reviews and community

Google Business Profile updates, weekly photograph uploads, posts, Q&A, and responses to every review received since last week. Citation checks. Duplicate hunting. For businesses where local is the primary channel, this is not an afterthought — it is a recurring commitment, and the recurrence is the point, because velocity and recency are ongoing signals rather than one-time deposits.

Friday: publish, record and plan

Push what is ready. Update the change log with everything done that week and the date it happened. Note what did not work. Plan next week against the roadmap, and move the roadmap if what happened this week changed the picture.

If your agency cannot describe a week like this for your specific account, ask what they do instead. The answer will be more informative than any report they have sent you.

A month inside an agency

The weekly rhythm produces activity. The monthly rhythm is what turns activity into compounding, and it hinges almost entirely on one week that most engagements skip.

Week one: review against the forecast

Not a review of what happened — a review of what happened versus what was expected, and an explanation of the gap. This distinction matters enormously. Reporting what happened is bookkeeping. Reporting what happened against what you predicted, and being wrong out loud, is the only mechanism by which anybody learns anything about your specific market.

Week two: production

The bulk of the delivery. Content, pages, campaigns, fixes, creative. Nothing surprising here, and it should be the largest block of hours in the month.

Week three: analysis and reprioritisation

This is the week that separates programs that compound from programs that repeat. What is working gets more resource; what is not gets cut or changed. The roadmap moves. New hypotheses are formed on the basis of what the previous month actually demonstrated rather than on what was assumed at the start.

If your roadmap has not changed in three months, nobody is doing week three. That is the single most reliable indicator of an engagement running on autopilot, and it is easy to check: ask what changed in the plan since the kickoff, and why.

Week four: report and plan

A report leading with qualified leads and blended cost per acquisition including the fee. What changed and why. What was done, specifically, with dates. What will be done next month and what it is expected to produce — stated before the work happens, so that next month’s week one has something to compare against. And an honest account of what was tried that did not work.

All month: monitoring

Paid accounts checked at least twice a week. Reviews answered within seventy-two hours. Alerts acted on the same day. Search Console glanced at weekly. None of this is billable-looking work and all of it is the difference between a program that is running and one that is merely live.

Who inside the agency actually does the work

The gap between who sells and who delivers is the single most common source of disappointment in agency relationships, and it is entirely avoidable by asking one question before signing: who specifically will work on this account, and how many other accounts do they carry?

The strategy director or partner

Roughly five per cent of the delivery hours and a much larger share of the outcome. They set direction, prioritize, make the difficult calls, and — in a good agency — are the person willing to tell you not to spend money on something. They are also the person most likely to be present at the pitch and absent from the account, which is why the question above matters.

Specialist practitioners

The SEO, the paid media manager, the email specialist, the analytics person. Around thirty-eight per cent of the hours and the largest share of the actual doing. Their depth is most of what you are buying that you could not easily hire, because no single in-house marketer covers all of these disciplines well — nobody does.

Content writers and editors

Around a fifth of hours and frequently the largest single production cost. The quality gap between writers is enormous and it is the gap that most determines whether content ranks: a writer who interviews your operations manager and includes information nobody else has produces something no competitor can copy. A writer who paraphrases the pages already ranking produces something with no reason to outrank them.

The account manager

Around fourteen per cent of hours: coordination, your emails, chasing approvals, running the calls, assembling the report. Necessary and frequently undervalued — a good account manager is the reason things ship on time — but if the account manager is the only person you ever speak to, you have no way of judging the specialist work underneath.

Designers, developers and analysts

Around twenty per cent between them. Designers produce the assets, developers ship the changes, and the analyst makes the numbers trustworthy. The analyst is typically the smallest allocation and one of the highest-leverage, because everything downstream depends on the measurement being right.

What this means for you

Ask for the names. Ask how many accounts each carries. Ask who stays if the person who pitched leaves. An agency comfortable answering those questions is describing a real team; one that deflects is describing an org chart it would rather you did not examine.

Real work versus agency theater

Two agencies, the same fee, the same industry, the same client. One spends thirty-one per cent of its hours shipping changes and eight per cent in meetings. The other spends nine per cent shipping and twenty-nine per cent in meetings, with another thirty-eight on reporting. Both send a monthly invoice for the same amount. Only one of them is doing marketing.

What theater looks like

• A report that grows longer as it grows emptier. Forty pages of charts in month nine when month two’s report was eight pages and said more.
• Reporting that leads with impressions, reach or followers. These are inputs. If they open the report, they are there because the outputs are not good.
• Recurring meetings with no decisions in them. A weekly call that reviews a dashboard everybody could read alone is a subscription to reassurance.
• Strategy documents that never become tasks. A beautiful deck in month one, referenced never again.
• Recommendations without shipping. Thirty-four issues found, six fixed, the rest still in the document two years later.
• Activity metrics as deliverables. ‘Twelve social posts, four blog articles, thirty citations’ tells you what was produced, not whether anything worked.
• A roadmap that has not changed since kickoff.
• No record of anything that failed. Over six months this means nothing was tried, or nothing is being reported honestly.
• Proprietary dashboards that repackage Search Console and Google Ads data you already own, presented as a platform.
• Jargon density rising as results fall. A reliable inverse correlation.

Why theater happens, without anybody being dishonest

Mostly it is structural rather than cynical. An agency overservices a new account to prove value, then corrects quietly as the economics bite. The genuinely high-value work — fixing the indexing, consolidating the cannibalizing pages, rewriting ten titles — gets done in the first quarter and cannot be repeated, but the retainer continues at the same level. Something has to fill the hours, and reporting and meetings expand to fill them because they are the easiest things to produce.

The honest resolution is a conversation about scope rather than a change of agency. If the high-value work is finished, the retainer should shrink, or it should be redirected into something with a genuine expected outcome. An agency willing to have that conversation is worth keeping; one that fills the hours silently is not.

What agencies should do and rarely do

None of the twelve items above costs an agency much, and most are absent from the retainers we review. All of them are straightforwardly negotiable before signing and awkward to introduce afterwards, which is the argument for reading this list before the contract rather than during month six.

Conversion work on the pages receiving the traffic

Almost every agency sells traffic and almost none sells conversion, because traffic is easier to attribute and easier to sell. But a site converting at 0.8% against a 3% benchmark is discarding three quarters of everything every channel delivers, and fixing it is cheaper than tripling traffic. Ask for it explicitly in the scope.

CRM integration

The difference between reporting leads and reporting revenue. Without it, everybody is optimizing toward lead count, and lead count is the metric most easily inflated and least connected to whether your business grew. With it, automated bidding can optimize toward closed revenue, which is the single largest available upgrade to most paid accounts above roughly $8,000 monthly spend.

Proactive recommendations to spend less

The clearest test of whether the relationship is honest. Every agency will tell you when to spend more. An agency that has, at some point, told you to cut a campaign, pause a channel or reduce its own fee is demonstrating something no case study can. Ask directly what would make them tell you to stop spending.

An honest account of what did not work

Six months of reports with no failures in them means nothing was tried or nothing is being reported. Both are problems. The presence of a short, unglamorous ‘this did not work and here is what we learned’ section is among the most reliable indicators of an agency worth keeping.

What only an agency can give you

This is the honest case for using one, and it is narrower than agencies claim and broader than skeptics assume. Two of the items are what you are really paying for and neither appears in a scope document.

Pattern recognition across many accounts

An agency that has run forty accounts in your category has seen your problem before, including the version of it that looks like a different problem. That is genuinely valuable and genuinely difficult to replicate internally, where you see one account for years.

Somebody to disagree with you

An in-house marketer reports to you. An agency does not, which means it can afford to tell you that the rebrand is a distraction, that the channel you want to try will not work at your budget, or that the reason leads are not converting is the sales process rather than the marketing. Whether it actually does so is a question about the specific agency, but the structural possibility exists and it has real value.

Six disciplines at once

One in-house hire covers one or two disciplines well. A competent agency gives you technical SEO, paid media, content, creative, email and analytics simultaneously, each done by somebody who does only that. For businesses under roughly $25M in revenue this is usually the deciding factor, and it is why the hybrid model — an internal coordinator plus an agency — is where most mature businesses land.

Capacity that flexes

A launch, a seasonal peak, a crisis. An agency can put four people on something for three weeks and then stand down. An in-house team of two cannot, and hiring for a peak leaves you overstaffed in the trough.

Redundancy

Underrated until it matters. An in-house marketer who resigns takes the context, the passwords and the momentum. An agency has other people, and the handover is somebody else’s operational problem rather than a three-month gap in your marketing.

What an agency cannot do

Worth stating plainly, because a great many disappointing engagements begin with an expectation that was never achievable.

1. Fix a product nobody wants. Marketing amplifies; it does not create demand for something the market has evaluated and declined.
2. Fix a sales process that loses the leads. If your close rate on inbound is four per cent, doubling the leads doubles the waste. This is the most common cause of an engagement that produces results everybody is unhappy with.
3. Fix a capacity problem. More leads are not a benefit when you cannot service the ones you have. The correct response is higher prices, not more marketing.
4. Control Google, Meta or TikTok. Nobody can guarantee a ranking or a cost per acquisition, and anybody who does is either misinformed or misleading you.
5. Compress an eighteen-month compounding curve into a quarter.
6. Make decisions you will not make. An agency waiting six weeks for approval on a page is not the reason the page is late.
7. Substitute for your own knowledge of your business. The best content requires interviewing your people, and your people have to be available.

The first ninety days, in detail

Ask for this by name in month one. An agency still ‘getting up to speed’ at day sixty is charging you for a ramp that should have taken two weeks, and in our experience the pattern rarely improves — the engagements that start slowly tend to stay slow, because the underlying cause is usually capacity rather than complexity.

Weeks one and two: access and instrumentation

Every account handed over: analytics, Search Console, Google Business Profile, advertising accounts, hosting, DNS, CMS, CRM. This sounds administrative and it is the single largest source of wasted time in month one — engagements routinely lose two or three weeks to chasing a password from somebody who left the company. Gather it before the kickoff call.

Alongside that: the audit begins, the baseline is documented, and the change log is opened. By the end of week two you should know what your current position actually is, in numbers, written down. Without a documented baseline, every future claim about improvement is unfalsifiable.

Weeks three and four: the audit, the map and the backlog

A written technical audit naming specific URLs and specific problems — not a four-hundred-line export from a crawling tool, which is a screenshot of software rather than an audit. A keyword map with exactly one URL against each target intent. And a prioritized backlog with an expected impact and an effort estimate against each item, so that the sequencing is defensible rather than arbitrary.

Month two: shipping

Conversion fixes live on the site. Tracking verified end to end from a real phone. Paid campaigns live if they are in scope. This is the month where the engagement stops being preparation, and if nothing has shipped by day sixty, that is the conversation to have.

Month three: production at cadence

The content program running at whatever cadence the retainer supports. Local work visible in the profile. First measurable results, which for most engagements means impressions moving and early long-tail rankings appearing rather than lead volume changing.

Month four: the roadmap moves

The plan updates on the basis of what month three actually demonstrated. If the roadmap in month four is identical to the roadmap in month one, nothing has been learned and nobody is analyzing anything.

How to work with an agency well

A meaningful share of failed engagements fail on the client side, and almost all of that is avoidable. These are the things that make the difference, in rough order of impact.

Name one decision-maker

Approval by committee turns a four-week project into a fourteen-week one and it is the single largest cause of overrun in this industry. One person who can say yes, with the authority to overrule colleagues, is worth more to the engagement than an extra thousand a month.

Give them your real numbers

Revenue by service line, margin, close rate on inbound leads, average customer value, capacity, seasonality from your own books. Without these, an agency optimizes toward the leads that are easiest to generate rather than the ones worth having, and every budget recommendation is a guess. Businesses withhold these out of habit and it costs them directly.

Budget four to eight hours a month of your own senior time

Approvals, interviews, asset gathering, feedback, decisions. The best content requires interviewing somebody who actually does the work, and that person has to be available. Engagements that stall almost always stall here: the draft sits waiting for a review nobody has time for and three weeks disappear.

Say what you will not do

If nobody will appear on camera, if you will not publish prices, if you will not change the CRM, if there is a competitor you will not name — say so at the start. Otherwise somebody builds a strategy around it and both sides lose a month.

Read the change log

Two minutes a month. It is the fastest available check on whether the fee is buying work, and simply being read changes what gets written in it.

Ask ‘why’ rather than ‘what’

‘What did you do this month’ produces a list. ‘Why did you do that rather than the other thing’ produces the reasoning, which is the part you are actually paying for and the part that tells you whether anybody is thinking.

Give it long enough, and not longer

Compounding channels need nine to twelve months to demonstrate anything. But an engagement with no shipped changes by day sixty, no roadmap movement by month four and no honest failure reported by month six is not going to improve by month twelve. Patience and passivity are different things.

Disagree out loud

If a recommendation does not make sense to you, say so. Either you learn something or they do, and both outcomes are better than silent skepticism that surfaces as a cancellation in month eight.

The kinds of agency, and what each actually does

Why the same allocation cannot fit everybody

The chart above shows four businesses of similar size with completely different needs. A local service business needs a third of its hours on local search and almost none on creative. A restaurant group needs nearly forty per cent on creative and almost nothing on organic content. A B2B supplier needs a third on content and a fraction on creative. An agency proposing the same allocation to all four is selling a package rather than making a plan, and the tell is that the proposal would barely change if you swapped the company name.

What the work looks like in different industries

Local service businesses

The majority of the hours go to local search, review generation and service and location pages. Paid search carries the early months. The distinctive work is operational content nobody else writes: permit requirements, building types, seasonal failure modes, what a job actually involves in your specific area. Success is measured in booked jobs and cost per booked job, and the reporting should reconcile to your job management software.

Ecommerce

Paid social and shopping campaigns dominate, creative production is a permanent cost rather than a project, and category-page SEO matters more than blogging. Lifecycle email is the highest-margin work available and is chronically under-resourced. The distinctive discipline is creative volume: platforms consume assets far faster than they consume budget, and an agency that treats creative as an occasional project will see performance decay predictably every eight weeks.

B2B and industrial

Content depth carries the program. The buyers search extremely specific technical language that consumer-oriented content never uses, and the competition for those terms is frequently negligible. Paid search is narrow and expensive. The distinctive work is subject-matter extraction — interviewing your engineers and turning what they know into pages — and the measurement is qualified opportunities and pipeline value rather than lead count, because three deals a year can make the number.

Professional services

Practice-area pages, individual professional pages, and procedural content answering the questions nobody else answers. E-E-A-T is not optional because these are Money-or-Your-Life categories. The distinctive constraint is compliance: professional advertising rules restrict claims and require disclaimers, and an agency that has not asked about them has not thought about your category.

Healthcare and dental

Reviews carry more weight than in any other category. Procedure pages written to answer anxiety, insurance acceptance pages, provider bios with verifiable credentials. The distinctive constraint is HIPAA: advertising pixels on pages containing protected health information have produced real enforcement actions, and any agency running your tracking needs to have raised this before you did.

Hospitality and restaurants

Creative and social carry discovery; the website carries conversion. Menu content as indexed HTML rather than a PDF is one of the highest-return single changes available. Review response is part of the product rather than a marketing task. Long-form content is close to worthless here and any agency proposing a blog program has not understood the category.

Nonprofits

Program and eligibility pages, event content, and Google Ad Grants — which provides substantial free search advertising and is chronically unclaimed and, where claimed, chronically mismanaged. The distinctive work is writing eligibility and application content clearly, in every language the community speaks, which almost nobody does properly.

How to tell whether yours is doing the work

Run that twelve-point check every month. Three or more reds for two consecutive months is a conversation rather than a cancellation — most of these items are fixable simply by asking for them, and an agency that responds well to being asked is usually worth keeping. An agency that responds badly has told you something useful.

The three questions that reveal most

1. ‘Which of our pages are indexed but not ranking, and why?’ A competent answer names specific URLs, specific queries and specific reasons — intent mismatch, insufficient depth, cannibalization, no internal links. A script talks about needing more content and more backlinks.
2. ‘What did we try in the last quarter that did not work?’ If the answer is nothing, either nothing was tried or nothing is being reported honestly.
3. ‘What would make you tell us to spend less?’ The answer tells you whether you are buying advice or execution, and both are legitimate purchases as long as you know which one you made.

Signals that things are going well

• The change log is specific and dated, and things on it actually appear on your site
• The roadmap has moved at least once since the start
• Somebody has disagreed with you about something
• A failure has been reported without you having to ask
• The report’s first page contains a number your accountant recognizes
• You have spoken to the person who does the work, not only the account manager
• Impressions moved before clicks, clicks moved before leads — the expected sequence
• The reported numbers reconcile roughly to your CRM

Signals that things are not

• The report is getting longer and less informative
• The roadmap is unchanged since kickoff
• Nothing has shipped to your live site in two months
• Every conversation is with an account manager who cannot answer specifics
• You cannot name three things done last month
• The numbers in the report do not reconcile to anything you can verify
• Requests for owner access to your own accounts meet resistance
• The word ‘optimization’ appears more often than any specific noun

Common failure modes, and whose fault each one is

Roughly half of these are client-side and the other half are agency-side, which is worth stating plainly. The engagements that work are ones where both parties recognize their half.

Agency, freelancer, in-house or hybrid

The pattern that works most reliably for businesses between roughly $3M and $25M in revenue is the last row: one internal marketing coordinator who owns the calendar, the CRM and the agency relationship, plus an agency for the specialist channels. The coordinator does not need to be a specialist. They need to be organized, inside the business, and empowered to say ‘that report does not answer my question’.

For the full cost comparison see marketing agency pricing, and for the senior-direction-without-headcount option see what is a fractional CMO and fractional CMO services.

Glossary

What agencies do that clients never see

A substantial share of the hours goes on work that produces no visible artefact, which is why agency value is so hard to judge from the outside and why so much of the industry’s reporting is defensive.

Reading things

Search terms reports, Search Console queries, analytics segments, competitor pages, review text, support tickets, platform release notes. A specialist reads far more than they write, and the reading is where the decisions come from. It appears in no deliverable list and it is most of the difference between an account that improves and one that runs.

Deciding what not to do

A large part of strategy is exclusion, and exclusion is invisible. The channel that was considered and rejected, the keyword cluster that was researched and abandoned because the intent was wrong, the campaign that was not launched because the maths did not work. None of it appears in a report and all of it saved you money.

Waiting correctly

A test that has not reached significance, a page that has not been crawled yet, a campaign that has not accumulated enough conversions to optimize against. Knowing when not to act is genuine expertise and it looks identical to inactivity from outside, which is why agencies feel pressure to change things that should be left alone.

Absorbing platform change

Google, Meta and the rest change constantly: interface changes, policy changes, deprecations, new formats, ranking updates. Somebody has to track it, work out what it means for your account and act. You never see this and it is one of the more defensible reasons to use specialists.

Chasing

Your approvals, your assets, your subject-matter experts, your IT department, your web host, your previous agency’s passwords. An uncomfortable share of account management hours goes here, and reducing it is almost entirely within your control.

Fixing things quietly

A tracking tag that broke after a plugin update. A form that stopped delivering. A campaign that started spending in the wrong geography after a platform change. Caught on a Monday check and fixed by lunchtime, and it never appears in the report because nothing bad happened — which is exactly the problem with judging an agency by its report.

What a good report actually contains

Reports are where the relationship is either maintained or quietly hollowed out. The structure below is what we consider minimally honest, in this order, and it takes about ninety minutes a month to produce properly.

1. The headline number, first. Qualified leads or booked revenue this month, versus last month, versus the same month last year. Not sessions, not impressions, not rankings.
2. Blended cost per acquisition. Total cost including the agency fee divided by customers won. Agencies reporting cost per lead on media spend alone are flattering themselves and you should say so.
3. What changed, and why. If leads fell eighteen per cent, that belongs on page one with an explanation, not on page eleven under a chart.
4. What was done, specifically, with dates. The change log, summarized.
5. What will be done next month, and what it is expected to produce. Stated before the work, so next month’s report has something to compare against.
6. What was tried that did not work.
7. Supporting metrics, last. Rankings, sessions, impressions, engagement. Context, never conclusions.

What should not be in it

• Screenshots of dashboards you already have access to
• Impressions and reach on page one
• Follower counts, unless you are genuinely running an audience-building strategy
• Charts with no interpretation attached
• The same commentary as last month with the numbers changed
• Anything included purely to make the document longer
• Percentages without absolute numbers underneath them
• Comparisons chosen because they flatter, rather than because they are the right comparison

The reconciliation question

Ask once a quarter: do these numbers reconcile to my CRM? They will never match exactly — attribution is approximate, some leads are duplicated, some are never entered — but if they are off by more than about twenty per cent, something is broken and it is worth an hour to find out what. Most agencies have never been asked this, and the answer tells you a great deal about how much the reporting is worth.

A worked example: one real month

A local service business, $4,200 monthly retainer, month seven of the engagement. Roughly thirty-one senior-equivalent hours. This is what the month actually contained.

Two things are worth noticing. The seventh row — analysis and reprioritisation — changed the plan: the discovery that two location pages were not indexed meant that publishing six more would have compounded a problem rather than solved one, so the content plan changed to improving four existing pages instead. That decision was worth more than the hours that produced it.

And the last row: nine days waiting for an approval on one page. That is not an agency failure and it is not free — it delayed the cost guide by a fortnight in a seasonal category. This is why naming one decision-maker is genuinely the highest-leverage thing a client controls.

Questions to ask before you hire anyone

1. Who specifically will work on this account, and how many other accounts do they carry?
2. Is the person in this meeting the person doing the work?
3. Do you ship code, or do you write recommendations for somebody else to ship?
4. What specifically changes on our site in the first sixty days?
5. How many hours a month does this fee represent, and at what seniority?
6. What is the one number we judge this by at twelve months?
7. What is explicitly out of scope?
8. Which of our pages are indexed but not ranking, and why?
9. What would make you tell us to spend less?
10. Show us an engagement that did not go well, and what you learned.
11. Can we speak to two current clients and one former client?
12. Who owns the ad accounts, the analytics, the content and the tracking numbers?
13. What does offboarding look like, and can we have it in writing?
14. What is your average client tenure in months?
15. What do you rank for organically, and can you show us Search Console?

Ask all fifteen, of everybody on your shortlist, including us. The pattern of what gets answered precisely and what gets deflected is more predictive than any case study, because case studies are selected and answers under pressure are not.

What changes as your business grows

The work an agency does for a $1M business and a $30M business share a vocabulary and very little else. Knowing where you sit prevents both over-buying and under-buying.

Under $1M in revenue

The majority of the value is in fixing conversion and claiming the free surfaces: the Google Business Profile, the review process, the contact form that does not work, the missing service pages. Most businesses at this stage do not need an agency at all yet — they need about forty hours of somebody competent doing the obvious things, after which they will know whether marketing is their constraint.

$1M to $5M

The stage where an agency starts genuinely paying for itself. Paid search for immediate volume, organic content for compounding, local for proximity capture, and enough conversion work that the traffic is not wasted. Typically $2,500 to $6,000 a month, and the internal requirement is one person who owns the relationship and can approve things.

$5M to $25M

Specialization begins to matter. Creative production becomes a permanent line rather than a project, analytics and CRM integration become necessary rather than nice, and the reporting has to reconcile to financial systems. This is where the hybrid model — an internal coordinator plus an agency — becomes the obvious answer, and where an agency without genuine specialists starts to be outgrown.

$25M to $100M

Brand and PR finally make sense, because there is now enough distribution for awareness to convert rather than evaporate. Multi-location or multi-market complexity appears. In-house capability starts to be worth building for the disciplines you use every day, with agency support for the ones you do not.

Above $100M

An in-house team with agency partners for specialist and surge capacity. The agency’s role shifts from doing the work to doing the work you cannot justify staffing — a specific discipline, a specific market, a specific campaign.

Working with an agency across a full year

The most common failure across a year is not a bad quarter — it is four identical quarters. If Q3 looks like Q1 with different dates, nothing has been learned, and the honest response is to change the scope or change the agency rather than to renew and hope.

Quick answers

What does a marketing agency actually do all day?

On a healthy account: checks what moved, maintains paid campaigns, produces content and creative, ships technical fixes, manages local profiles and reviews, tests conversion changes, analyzes what happened and changes the plan accordingly. Roughly thirty senior-equivalent hours a month at $4,000. See marketing agency pricing for how that breaks down financially.

How do I know if I need an agency?

If you have done the free high-return work — complete Google profile, review process, working contact form, a page per service, internal links — and growth is still constrained by marketing rather than by capacity or product, then yes. If you have not done those, do them first; they outperform a small retainer.

What is the difference between an agency and a freelancer?

Breadth and redundancy. A freelancer gives you one or two disciplines done deeply by the person you hired, with no cover if they are busy or ill. An agency gives you six disciplines at once and a bench, at the cost of your account being one of many.

Should the agency do my website too?

It can work well and it can create a hostage situation. It works when you own the domain, the hosting and the code, and the site is on an open CMS you can move. It fails when the site lives on a proprietary platform the agency owns. See website design and development.

How involved do I need to be?

Four to eight hours a month of somebody senior: approvals, interviews, asset gathering, decisions. Engagements that stall almost always stall waiting for these. Name one decision-maker with authority and the whole thing runs faster.

What should happen in the first month?

Access to every account, a written audit naming specific URLs and problems, a documented baseline, a keyword or channel map, a prioritized backlog, verified conversion tracking, and a change log. Not rankings — rankings in month one are noise.

Why is my agency’s reporting so long?

Frequently because the results are not good. Reports that grow in length while shrinking in substance are a reliable signal. Ask for a one-page version leading with qualified leads and blended cost per acquisition and see what happens.

Can an agency guarantee results?

No, and anybody who does is either misinformed or misleading you — nobody controls Google, Meta or TikTok. What an agency can commit to is a scope, a cadence, a named team, ownership terms, and honest reporting including of failures.

What is a fair amount of the fee to be management overhead?

Account management around fourteen per cent of delivery hours is normal. If more than a third of the hours are coordination, reporting and meetings, you are funding a relationship rather than a program.

How long before I should expect results?

Paid search in one to three weeks. Local and reviews in four to ten weeks. Organic in four to nine months. Conversion work in four to eight weeks and it multiplies everything else. See how long does SEO take.

What if I want to change agency?

Check the contract first: notice period, auto-renewal, and who owns the ad accounts, analytics, domain, content and tracking numbers. Ask for a written offboarding process. The time to negotiate this is at the start, when they want your business, not when you are leaving.

Should I hire a specialist or a full-service agency?

A specialist if one channel is clearly your growth engine and you can coordinate the rest. A full-service agency if you need several channels working together and do not have somebody internal to coordinate them. A specialist will recommend their channel for everything, which is worth knowing.

What does ‘strategy’ actually mean?

A small number of specific decisions with reasons attached, and things explicitly excluded. If a strategy document contains no exclusions, it is a wish list. If it never becomes a task list, it is a deck.

Why does my agency want a twelve-month contract?

Because compounding channels genuinely take that long to demonstrate anything, and because client acquisition is expensive for them. Both are legitimate. What is not legitimate is a twelve-month term with no termination provision — negotiate a 30 to 90 day exit.

What is the single biggest thing I can do to make this work?

Name one decision-maker and give the agency your real numbers — close rate, customer value, capacity, seasonality. Those two things together remove most of what causes engagements to underperform.

Watch: what agencies actually do

Three from Google Search Central that bear directly on the material above.

Every guide we publish, grouped by what you are trying to do.

What separates good from adequate, discipline by discipline

Every agency lists the same services. The difference between an engagement that works and one that merely runs is visible at the level of individual tasks, and it is worth knowing what the distinction looks like in each discipline so you can recognize it.

Strategy: good versus adequate

Adequate strategy selects plausible channels and allocates budget across them. Good strategy starts from your unit economics, states what a lead can cost before choosing anything, names the channels it is deliberately not using and why, and commits to a forecast it will be judged against. The test is whether the document contains exclusions. A plan with no exclusions is a wish list, and a plan that would barely change if you swapped the company name is a template.

Research: good versus adequate

Adequate research runs a keyword tool and exports a list. Good research reads your sales call transcripts and support tickets for the words customers actually use, pulls your existing Google Ads search terms report because those are queries that already produced revenue, verifies search intent against the live results page for every target term rather than assuming it, and looks at the pages actually ranking rather than at the competitors you named.

Technical work: good versus adequate

Adequate technical work produces a four-hundred-line crawler export with color-coded severity. Good technical work names twelve specific URLs, explains what is wrong with each, prioritizes them by expected impact, ships the fixes, and confirms afterwards that the pages actually index. The distinction is whether anything changes on the live site, and it is startling how often the answer is no.

Content: good versus adequate

Adequate content covers the topic competently by reading what already ranks and reorganizing it. Good content contains something the ranking pages do not: a number you measured, a photograph you took, a process you actually follow, a constraint only somebody in your business would know about. That is the only property that gives Google a reason to prefer your page, and it cannot be produced without interviewing somebody who does the work.

Paid media: good versus adequate

Adequate paid management sets campaigns up correctly and reports on them. Good paid management reads the search terms report weekly for the first two months, tests ad copy against a stated hypothesis rather than at random, imports offline conversions so bidding optimizes toward closed revenue rather than raw leads, and proactively recommends cutting spend on campaigns that are not working — which reduces its own revenue under a percentage-of-spend model and is therefore the clearest available signal of honesty.

Local: good versus adequate

Adequate local work claims the profile and fills in the obvious fields. Good local work audits the primary category against the businesses actually ranking above you from inside your service area, populates every service with a real description, builds a review process that runs weekly by text within a day of the job and asks a question that produces service-and-place text, responds to everything within seventy-two hours, and tracks rank from a grid rather than a single point.

Analytics: good versus adequate

Adequate analytics installs the tag and builds a dashboard. Good analytics defines conversion events that correspond to real business events, tests them end to end from an actual phone, adds a call duration threshold so hangups stop counting, deduplicates leads that both call and submit a form, connects the CRM so closed revenue is attributable to landing pages, and reconciles the reported numbers against your accounting system once a quarter.

Creative: good versus adequate

Adequate creative produces attractive assets on brief. Good creative produces enough volume to survive platform fatigue — which on paid social means days rather than months — tests distinct concepts rather than color variations, and produces the unglamorous formats nobody scopes: the vertical cutdown, the email template, the one-pager the sales team actually uses.

Reporting: good versus adequate

Adequate reporting summarizes what happened accurately. Good reporting compares what happened against what was predicted, explains the gap, states what will be done next and what it is expected to produce, and includes an honest account of what was tried and failed. The second kind is the only one from which anybody learns anything about your specific market.

What the work looks like when it is going right

A composite, drawn from engagements that worked. Not a case study with a number attached — a description of the texture, because that is the thing you can actually recognize from inside a relationship.

Month one feels slow and is not

Very little visible happens. Access is gathered, the audit runs, the baseline is documented, the keyword map is built. You are paying for preparation and it feels like paying for nothing. The signal that it is going right is specificity: the audit names your URLs, not general principles, and the backlog has expected impacts against items rather than a severity color.

Month two produces the first argument

Somebody tells you that a page you like is not working, or that a term you want to rank for is not worth chasing, or that the reason inquiries are not converting is the form rather than the traffic. This is a good sign. An engagement with no disagreement in the first quarter is one where somebody has decided that agreeing is safer, and that decision costs you for the next two years.

Month three shows the wrong metric moving

Impressions rise while clicks and leads do not. This is correct and it is the point at which many businesses lose confidence. The sequence is always impressions, then clicks, then leads, then revenue, and each step lags the one before by weeks. An agency reporting lead growth in month two on a compounding channel is reporting noise or something else entirely.

Month four the plan changes

Something in month three did not behave as expected and the roadmap moves in response. This is the single most reliable indicator that anybody is thinking. Ask what changed and why; the answer tells you more about the agency than any report will.

Month six the reporting gets shorter

Counterintuitively, good engagements produce shorter reports over time, because the conversation narrows onto the two or three numbers that actually matter and everything else becomes an appendix. Reports that lengthen are usually compensating.

Month nine you stop checking

Not because you have disengaged, but because the pattern is established: things ship, the numbers move in the expected direction, failures get reported before you find them, and the monthly call takes twenty minutes because there is nothing to reconcile. This is what a working agency relationship feels like and it is quieter than the pitch suggested.

Month twelve the scope changes

Either upward, because the program is producing and there is more worth doing, or downward, because the high-value work is finished and the agency has said so. Both are healthy. What is not healthy is month twelve looking identical to month four with a renewal attached.

Where to go next

Everything we publish, organized by the question you are actually asking.

Frequently asked questions