How we build an audience-centric content strategy

Most content strategies start with a list. Sometimes it's the list of services, sometimes it's the people who've said they'd be willing to write something, and sometimes it's a set of topics that came up in a planning meeting and felt current at the time.

The list isn't the problem so much as where it starts, because it begins with what the organisation does and works outwards to who might want it.

When you approach things from this angle, you tend to miss the one thing that decides whether a page gets read or not: what the target audience is actually trying to get done.

We always start with who, work out what they're trying to get done, and only then decide what to write, and that reversal is our approach in a nutshell.

This article is about how it works in practice, where it gets difficult, and which parts you can do yourselves.

 

Audience mapping is a way of working out who your content is for before you decide what to write. You list every audience the organisation actually serves, decide how much effort each one deserves, and then, for the ones that matter most, work out what they're trying to get done and the questions they ask along the way.

What you end up with is a map, and the content plan is built from that map rather than from the list of services.
 

It's worth saying that this isn't a new idea, or one we came up with ourselves. The thinking behind it is based on "jobs to be done", which Clayton Christensen developed at Harvard Business School and set out in The Innovator's Solution in 2003, building on work Tony Ulwick had been doing with companies since the early 1990s.

The principle is simple: people don't buy a product or a service for its own sake; they hire you to help them get something done. It's been used for the best part of thirty years, by organisations of every size, to understand what their customers are actually trying to achieve, and our approach applies the same thinking to content.

It matters because once you know what your audiences are trying to get done, you know what brings them to your site in the first place, and you can write for that rather than guessing.

A job also isn't a single moment. It's a journey that happens in stages, and at each stage a different person might be involved, asking a different question, so one page can't cover all of it. Mapping the job stage by stage means you can build a set of content that meets people at each point in the journey, from their very first question through to the point where they're ready to talk to someone.

Getting this right is more important than ever, because every piece of content now has three readers:

• The people it was written for.
• The search engines that decide whether those people find it.
• The AI assistants that increasingly answer their questions on your behalf.

All three reward the same thing: content written for one reader at one stage, built around one question, with detail only your organisation could provide. A map tells you exactly what that looks like for each piece.

 

Audience mapping starts by listing every audience the organisation actually serves. This doesn't always align with the sectors and services listed on the website; often those lists tend to mix the work it does with the work it would like to be doing.

The evidence for who it really serves nearly always exists already, just not in the obvious places. Most expert-led websites are organised by service line, with little or no audience layer, so the information about who the organisation works for is scattered: in the people pages, the case studies and named clients, the topics it writes articles about, and the industry bodies it belongs to, which are often the clearest sign of which audiences it really serves. A lot of it also never makes it onto the website at all, and sits in pitch documents, proposals and credentials that were written for one client and then filed away.

So rather than reading the website page by page, we've built a process that takes in everything the organisation has written, the website and, where clients are able to share them, those offline documents too, and works through it locally, on our own systems, so that every mention of who it works for can be brought together in one place. 

The next step is to group those audiences into three tiers. The tier grouping is a decision about how much content and marketing effort the organisation will put into each one over the next two to three years:

 

To make that concrete, here's how it played out for an advisory organisation we worked with, described in general terms. Its website listed twelve services and no audiences at all.

Bringing together the biographies, the case studies and a set of past proposals produced eleven audiences, and a conversation with the marketing team saw them grouped like this:

• Grow: owner-managed businesses preparing for a sale or succession, and technology companies raising their second or third round of funding. Both were where the organisation had won its best recent work, both had senior people with visible track records, and both were where it wanted to be known in three years' time.
• Defend: private landlords with sizeable portfolios, and family businesses with complex ownership. Long-standing, profitable, well served by experienced people, but not where the marketing team wanted to put new effort, because the relationships were mostly inherited rather than won.
• Maintain: charities, sole traders, international clients with a UK footprint, and four others. Real capability, real revenue in some cases, but served on referral and through the services listing rather than led with.

The tier is about focus, not how valuable the work is. In that example, the family-business audience was among the most profitable the organisation had, but it stayed in Defend because the relationships were already there and the team was at capacity.

There are two things to watch out for here, and both are about people rather than analysis. The first is that a senior person whose area lands in Maintain won't enjoy reading it, so the tiers need to be presented as a decision about focus, not a ranking of value.

This step also often shows a gap between how the organisation sees itself and how it actually presents itself online. In the example, the technology audience had two strong people profiles and some good work examples behind it, but nothing on the website a technology founder would have found, so it needed to become visible before it could grow. This is an easier one to walk through with stakeholders, as it's objective and can be backed with data. 

With the tiers agreed and a first audience chosen, the next step is to decide what, specifically, to map.

 

A persona tells you who someone is, such as a finance director at a mid-sized business, but a finance director likely has fifty problems at once and they would all require a different solution.

A job, by contrast, has a beginning, a middle and an end, and different people get involved along the way, so it gives you something you can map stage by stage, and something an expert will recognise as the work they actually do.

The difference is easiest to see with a good job and a bad one side by side.

"Advising owner-managed businesses" is not a job, because it has no start, no finish and no particular question in it; it's a description of a service.

"A founder preparing to sell their business" is a job, because it begins with a decision about timing, ends with the proceeds being dealt with, involves the founder, a finance director, a management team and several advisers at different points, and has a different question at each of those points.

A good job has three things going for it:

• A clear start and finish: a business sale, a funding round or a product launch all work, because you know where the content begins and where it ends. "Growing the business" doesn't, because it never ends.
• Context: Launching a product in pharmaceuticals is different from launching one in any other sector, and that difference is where an organisation's specialist knowledge comes through. A job without context produces content anyone could have written.
• Commercial value: the simplest way to find the first job is to ask the leadership which project, for which client, they would most like to win next year, and to start there.

Within a Grow audience there are usually two or three jobs worth mapping. For the owner-managed-business audience in the example, they were the sale itself, a succession to the next generation, and a management buyout, and the sale went first because it was the one the leadership most wanted to win more of. It's worth taking them one at a time rather than trying to map the whole audience at once.

With a job chosen, the next step is to break it into the stages the client actually goes through.

 

Once you have a job, you break it into its stages and work out who's involved at each one and what they're trying to find out, and that understanding of the journey is what the content plan is built on. It's also the part of the method that produces something the organisation has never seen before, because nobody inside it has ever laid the whole journey out from the client's side.

A job usually breaks into five or so stages. Here's a high-level version of the 'founder preparing for sale' example mentioned earlier:

 

Reading across, stage one is the founder alone, before anyone else knows, asking a private question about value and tax. Nothing an adviser publishes about its transaction experience will reach that person, because they aren't looking for an adviser yet; they're looking for a number.

By stage two the finance director is involved and the question has become practical, about what a buyer will look at, and that's a question an adviser can answer with real detail from real deals.

Stage three is the only stage at which the organisation's credentials matter, and even then the question is specific: our sector, our size.

Stage four belongs to the management team and is about running the business through the disruption, which is a question most advisers never think to answer because it isn't legal or financial work as they'd define it.

And stage five is the founder again, with a new question about what to do with the money, which is often where the next relationship begins.

Reading it this way shows two things, and they're the reason the map is worth building:

• The person changes as the job moves: the founder at stage one wants to know whether it's worth doing, the finance director at stage two wants to know what's involved, and the management team at stage four wants to know what it means for them, so it's the same job seen by different people, and each of them needs different content.
• The questions are specific enough to write for: "What will a buyer's due diligence look at first?" is a page that a finance director will search for and an AI assistant will reach for, whereas "corporate transactions" is a service line, and only one of them can be found by someone who's actually asking.

For each stage we also record which of the organisation's services serve it, which of its people are already named on the relevant pages and which should be, based on their recent work, the best recent examples of that work, and, where it differs from stage to stage, who the organisation is competing against.

The result is a single view, read from the first stage to the last, that shows the whole job and the organisation's current content and proposition at every point in it.

How the first draft gets built: 

This is the part of the method that has changed most in the last couple of years. When we started doing this, we built the first map in a room with the relevant experts, a blank whiteboard, and a lot of sticky notes, working through the job stage by stage until the questions were up on the wall. It worked, but it took a long time, and a good share of that time went on things the experts didn't really need to be there for, like agreeing what the stages should be called or listing services that were already on the website.

Now we create the first draft before anyone enters the room. The same process we use in step 1 takes in everything relevant that's already been written about the job: the organisation's own website and, where we have them, its pitch documents and other material that never goes online, alongside what its competitors publish, the industry's own publications and the research reports its audience reads.

We use AI to help us bring every data point together, from a line in an old proposal to a question that keeps coming up in a trade body's survey, and build a map from it that gets us about 80% of the way there without taking any of the experts' time.

There's nothing secret about the idea, and you could try it yourselves. The difference is how long it takes, because reading and cross-referencing that much material by hand is weeks of work, and the process we've built to do it quickly and consistently is a large part of the value we add.
 

The content plan comes after the map, and it's built from the questions on it, checked against the evidence, before any expert spends a minute writing. That order matters because it's the difference between a plan that produces content people are looking for and a plan that produces content people would like to have written.

The check is straightforward to describe, if not to do. Take every question from every stage and test it against four things: how many people are searching for it, who already answers it well, where the organisation currently ranks for it, and which of your pages on the subject AI assistants are already reading, which we covered in how we read AI crawler data. Then look at what the organisation already has, so that each question ends up in one of three states. Using the sale example again:

• Answered well on the site: "What is my business worth?" turned out to have a good valuation guide behind it already, written two years earlier and still accurate, so it was left alone and linked from the stages that needed it.
• Answered, but thin or hard to find: "What will a buyer's due diligence look at first?" was covered in two paragraphs halfway down a service page that nobody searching for the question would ever land on, so the answer was reworked as a page of its own rather than written again.
• Not answered at all: "When do we tell staff and customers, and what do we say?" wasn't answered anywhere on the site, or on most competitors' sites either, and became the first piece written.

[Diagram placeholder] The map graphic would work well here.

That sorting is the plan. Gaps first, weak pages second, Grow audiences before Defend, and one job at a time, with each piece written for one reader at one stage, around one question, and carrying something only the organisation knows, whether that's an example, an opinion or a detail from real work.

That last part is increasingly important, because it's what makes a reader trust the page and it's what search engines and AI assistants look for when they choose one source over all the others covering the same topic.

The thing to watch for at this step is usually the temptation to skip the evidence check, because an expert has a topic they want to write about, it feels current, and it goes straight on the plan.

Sometimes that's right, but more often it's a question nobody is asking, or one that three larger organisations already answer better, and the expert's time would have been worth more on a stage-four question with no good answer anywhere. 

Once the plan exists, the way you run the content calendar changes too, which is the next step.

 

A content calendar built this way stops asking "what shall we publish this month?" and starts asking "which job are we finishing this quarter?", and that's a bigger shift than it sounds, because you plan around jobs rather than months. A sprint takes one job from map to published content, with new pages where there are gaps, reworked pages where what exists is hard to find, and nothing at all for the stages that are already well served.

For the sale example, the sprint ran to about six weeks and produced eight new pages, two reworked ones, and three questions deliberately left alone because the existing content already answered them. The experts' time went to six pieces of writing rather than the twelve on the original topic list, and every one of the six could be traced to a stage, a person and a question.

That's usually how it goes: less writing than people expected, and considerably more useful, and because every piece is tied to a stage of a job for a Grow audience, you can say exactly why it exists.

It also gives you something to measure, because when a piece exists to answer a specific question at a specific stage, you can look at whether that question now finds you, in search, in AI assistants' answers, and in the enquiries that mention it, and that's a different conversation from page views.

[Placeholder] A paragraph on measurement? Build, measure, learn. An opportunity to say Second Site does this for you.

There are three things worth keeping in mind, because they'll trip you up if you're not expecting them.

The politics: 

The tiers are a decision about where content effort goes, and they will be read as a ranking of people's areas however carefully they're framed. Build the first version from the evidence before anyone senior sees it, present it as what the evidence says rather than what you think, and expect to be corrected, because the corrections are where the useful information is.

The validation: 

A map built from what's already been written is a first draft, however much material went into it, and it will feel finished long before it is. The session where your own experts argue with it is the step that turns it into something you can plan against, and it's the step most organisations skip because it costs senior time, so if there's only one part of this you can do properly, that's the one to choose.

The evidence check: 

An expert with a topic they want to write about is the most persuasive thing in a content meeting and the least reliable guide to what will be found. Check every question against search demand, existing answers and what AI assistants are already reading before anyone writes a word, and be prepared for the check to remove half the list.

We run all of this as a service, one job per sprint, with your experts adding what only they know. The process we've built for the first draft is what keeps that quick, because your experts' time goes on correcting a map rather than drawing one from scratch. The structure isn't a secret, though, and an organisation that builds its own tiers from its own site, picks one job, and gets its experts to pull the map apart will already have a better plan than the one built from the list.