Segmenting past clients

Segmentation in this trade is not demographics. It is two questions your booking system has already answered: what did they book, and where did they come from. Everything useful follows from those, and the second one is enrichable for nothing, because the Census Bureau runs a free geocoder that matches addresses to the geographic entities containing them, and both the tax authority and the economic analysis bureau publish detailed data about those places. A guide who knows which counties their clients actually travel from is doing better segmentation than one who has sorted a list by imagined persona. The rest of the client-facing work sits under the getting-booked hub.

Why not segment by demographics?

Because you do not have them and they would not help.

Age, income and household composition are things a guiding operation neither collects nor should.

What you do hold is behaviour: what somebody booked, how many times, in which month, with how many people, and how far they travelled.

Every one of those changes what you should say to them, which is the only test that matters.

A client who has booked three times needs no explanation of what a day involves. One who booked once needs a reason to return. One who never booked needs to know what has changed.

Those three groups want different messages and are trivially separable from data you already have.

Anything more elaborate is a spreadsheet exercise that produces no different sentence.

Where that data lives is described in the collecting addresses piece.

Crescent City Fishing Charters: Lafitte, LA

What does distance actually change?

The planning horizon, and therefore the timing.

Somebody an hour away books a week out, sometimes a day out, and responds to conditions.

Somebody flying in books months ahead, around a holiday or a long weekend, and responds to dates rather than to fishing reports.

Sending both the same message means one of them always receives it at the wrong time.

So the single most useful split after repeat-versus-once is local versus travelling, and it costs nothing to make.

Draw the line wherever your own bookings suggest, which is usually somewhere between two and four hours.

Then send the conditions-driven messages to the near group and the calendar-driven ones to the far group.

The calendar those far-group messages hang on is set out in the winter campaign piece.

How do you turn addresses into places?

A free federal geocoder, in a browser or in bulk.

The Census Geocoder provides interactive and programmatic access for matching addresses to geographic locations and the entities containing those addresses.

It offers one-line address processing, parsed address processing, batch address processing, and lookups from geographic coordinates, with a services interface documented separately.

It covers addresses within the United States, Puerto Rico and the island areas, which is the whole of most guiding client lists.

Batch processing is the useful one: a file of client addresses in, a file with counties and tracts out.

Do that once with a season of bookings and you will know which counties your clients actually come from, rather than which ones you assume.

Most operations are surprised, usually by one town that produces far more than its size suggests.

The geocoder is published by the Census Bureau.

Kona Sea Adventures, HI

What can you learn about those places?

More than you would expect, all of it free.

The economic analysis bureau publishes personal income by county, and its release for 2024 records that personal income in current dollars increased in 2,768 counties, decreased in 331 and was unchanged in 7.

It notes county-level changes ranging from a 22.6 percent increase in Harding County, South Dakota, to a 23.3 percent decline in Issaquena County, Mississippi.

That release was published on 5 February 2026, with the next scheduled for 2 December 2026.

None of that tells you anything about an individual client, and it does tell you something about a market you are considering advertising into.

Use it as context rather than as targeting, since inferring a person's circumstances from their county is both unreliable and unpleasant.

The honest use is deciding where to spend, not deciding what to charge somebody.

The release is published by the Bureau of Economic Analysis.

Can you see where people are moving?

Yes, county by county, going back decades.

The tax authority publishes migration data based on year-to-year address changes reported on individual income tax returns.

It presents patterns by state or by county for the entire country, covering inflows, meaning new residents and where they came from, and outflows, meaning residents leaving and where they went.

The series runs for filing years 1991 through 2023, and includes the number of returns filed, which approximates the number of households that migrated, and the number of personal exemptions claimed, which approximates individuals.

Total adjusted gross income is included from filing year 1995, and state-level flows by income size and age of the primary taxpayer from filing year 2011.

The publisher notes that beginning with data for 2022 to 2023 it enhanced the matching process, presented as a new series.

For a guide the use is narrow and real: if a nearby county is gaining households from somewhere your fishery is unfamiliar, that is a market that needs explaining rather than selling.

The data is published by the Internal Revenue Service.

How many segments should you run?

Three, and a fourth only if you will use it.

Repeat clients, one-trip clients, and people who enquired without booking is a complete working set for almost every operation.

Add local versus travelling as a cut across those three when a message is genuinely timing-sensitive.

Beyond that, every additional segment costs you a version of every message, which is the hidden price nobody counts.

An operation with nine segments writes nine versions or, more commonly, writes one and sends it to everybody anyway.

So the honest test for a new segment is whether you would actually write a different message for it this season.

If the answer is no, the segment is decoration.

What those messages should contain is set out in the newsletter ideas piece.

What about trip type?

A better cut than most, where you offer genuinely different days.

An operation running half days, full days and offshore trips has three populations with different budgets, different expectations and different reasons to return.

Somebody who booked a half day for their family is not the same prospect as somebody who booked a full day alone, and a message treating them identically will suit neither.

Tag the trip type at booking rather than trying to reconstruct it later, since that is a field rather than a project.

Then write the return message in the language of the trip they actually took.

Where you offer one kind of day, this cut does not exist and the effort belongs elsewhere.

Resist inventing trip types to create segments, which is the tail wagging the boat.

How the welcome message differs by type is covered in the welcome sequence piece.

Where should the segments live?

In one file, as columns rather than as separate lists.

The common failure is keeping four lists, which reliably produces somebody on two of them and nobody on the right one.

One row per person with columns for trip type, booking count, last trip date and distance band gives you every segment as a filter.

It also survives changing tools, which separate lists do not.

Add the county from the geocoding pass as another column, since it will be useful for advertising decisions and costs nothing to carry.

Keep the do-not-contact flag in the same file rather than in a separate one, because that is the flag you cannot afford to lose.

Review it once a year, which takes an hour and prevents most of the mistakes in this subject.

Why one file matters is explained further in the automations piece.

What about the people who never booked?

A real segment, and the one handled worst.

An enquiry that went nowhere is not a failed client, it is somebody whose date, budget or nerve did not line up that year.

Most operations either delete them or fold them into the main list, and both are mistakes for the same reason: the right message to that group is different from the right message to anybody else.

What they need to hear is what has changed, meaning new availability, a new trip type, a different price, or simply that the season they asked about has come round again.

What they do not need is a newsletter from a business they never used, which is the fastest route to a complaint.

Tag them at the point of enquiry with the date and what they asked about, since that is the only context you will have later.

One message a year, written as a person, is the entire correct treatment.

And where somebody has ignored two of those, stop, because their silence has answered the question.

How the spam consequences of getting this wrong work is covered in the deliverability piece.

Does the segment change what you charge?

No, and this is worth stating plainly.

Everything above is about what you say and when you say it, never about what you charge whom.

Quoting a different price to somebody because of where they live, what their county's income looks like, or how far they have travelled is a route to a problem that no amount of segmentation sophistication justifies.

It is also commercially poor, because the client who discovers it tells everybody, and in a small market everybody is a short list.

Published prices, applied to everybody, remove the entire question, and they make every other part of this easier.

Where you genuinely price differently, price the trip rather than the person: a longer day, a second boat, a peak weekend.

Those are differences in what is being sold rather than in who is buying, which is the line worth holding.

County-level economic data belongs in decisions about where to advertise and nowhere near a quotation.

How pricing should be presented is covered in the content calendar piece, alongside everything else a page has to say.

How often should the file be rebuilt?

Once a year, in the off-season, in an afternoon.

Segments decay because people move, book again, or stop opening anything, and none of those updates itself.

The annual pass is short: export the season's bookings, update the counts and dates, re-run the geocoder on new addresses, and move anybody who has crossed a line.

While you are there, send one message to anybody silent for two seasons asking whether they still want to hear from you, and remove whoever does not answer.

Pruning feels wasteful and is the single most effective habit in this whole subject.

Do it in the same week each winter so it becomes a fixture rather than a decision.

Write down what you did, because next year you will not remember which columns you added.

An operation doing this three years running has something no competitor has, which is an accurate picture of where its work comes from.

What is the smallest useful version?

Two columns, added to what you already have.

An operation with no appetite for any of this can get most of the benefit from two additions to whatever holds its bookings.

The first is a count of how many times each person has fished with you, which takes one pass and never needs recalculating from scratch.

The second is the date of their last trip, which the booking system already knows and rarely surfaces.

Those two columns alone give you repeat versus once, and recent versus lapsed, which is three quarters of the value here.

Everything else described above is refinement on top of that, and refinement is optional.

So if the full pass will not happen this winter, do the two columns and stop, rather than doing nothing because the whole thing looked large.

Half of this discipline is choosing a version small enough to actually finish.

The same reasoning applies to the sending side, covered in the tools comparison piece.

Which habits waste the effort?

Seven, and the first is segmenting without writing differently.

Building segments and then sending everybody the same message anyway.

Splitting by imagined persona rather than by what somebody actually booked.

Keeping separate lists instead of columns, so people appear twice or not at all.

Guessing where clients come from rather than geocoding a season of addresses.

Inferring something about an individual from a county statistic.

Creating a segment you will never write a separate message for.

And letting the do-not-contact flag live somewhere it can be lost in a tool migration.

The deliverability consequences of getting this wrong are covered in the spam piece.

What surprises operators here?

That the geocoding is free and takes an afternoon.

A federal geocoder will batch-match a file of client addresses to counties and tracts at no cost, and almost nobody in this trade has run one.

The second surprise is the concentration that falls out, with one or two counties usually producing far more than their share.

The third is that migration between counties is published annually, back to filing year 1991, from tax return address changes.

The fourth is that county personal income moved in almost every county in a single year, which is a reminder that a market is not static.

The fifth is that three segments is enough, and that most operations would do better with fewer and more distinct messages.

Taken together, this is an afternoon of data work that changes where money gets spent.

Where that spend should point is covered in the local search piece.

The segmentation, in order

Three groups, one file, one geocoding pass.

Split into repeat clients, one-trip clients, and enquiries that never booked.

Add a distance band as a cut across all three, drawn where your own bookings suggest.

Tag trip type at booking if you genuinely run different days.

Keep all of it as columns in one file rather than as separate lists.

Run a season of client addresses through the federal geocoder once and add the county as a column.

Use county-level published data to decide where to advertise, never to infer anything about a person.

Write a genuinely different message for each segment, or delete the segment.

And review the whole file once a year, keeping the do-not-contact flag with everything else.