Window Company Marketing: A Year of Local Campaigns Drafted With AI

Build the year around the moments homeowners notice their windows: autumn draughts, a spring refresh, summer survey slots and a pre-winter order deadline. Give an AI assistant one fact sheet of claims you can prove, ask it for each month's posts, emails and leaflet copy, then check every claim against your documents before anything goes out.

The risk is not weak copy. It is invented claims: left alone, an AI will happily promise heating-bill savings you cannot document and a "lifetime" guarantee you do not offer. So the first jobs are a fact sheet, a claims check and a way to measure which campaign paid. The drafting comes after that and takes far less time than most owners expect.

Follow me on Instagram@sagnikteaches

Start with your own enquiry log, not a generic calendar

Seasonal advice for window companies tends to say "autumn is busy". Your own log will say how busy, and when the enquiries start. Export twelve months of enquiries to a sheet with four columns: date, source, job type and outcome (survey booked, quote sent, sold, lost). Remove names and phone numbers before you paste anything into an AI tool; the dates, sources and outcomes are all it needs.

Connect on LinkedInSagnik Bhattacharya

Here is a year of enquiries for a window and door installer (dates, source, job type, outcome; no personal data).
1. Count enquiries by month and by source.
2. Tell me which three months bring the most, and what share of the year that is.
3. For each source, give enquiry-to-survey and survey-to-sale rates.
4. Separate what the data shows from anything you are guessing. Label guesses "guess".
Do not explain causes unless the data supports them.

Here is an illustrative result for a nine-person installer with two fitting crews and 289 enquiries in the year: January 14, February 17, March 26, April 31, May 27, June 18, July 13, August 15, September 29, October 52, November 36, December 11. The AI's summary would be that September to November brings 117 enquiries, about 40 per cent of the year, that October alone brings 18 per cent, and that June to August is the quiet stretch at 46 enquiries, about 16 per cent.

Subscribe on YouTube@sagnikteaches

The usual fix is on the "why". A first draft tends to add "enquiries rise as temperatures drop", which sounds right but is a guess unless your log holds weather or campaign data. Ask it to relabel that sentence as a guess, or delete it. What the numbers do support is a planning rule: with about 40 per cent of enquiries arriving in autumn, the work to capture them starts in summer, particularly if your suppliers need weeks between order and fitting.

A fact sheet that keeps the AI to claims you can prove

Every prompt in this tutorial starts with the same page of facts. Keep it in a ChatGPT Project or a Claude Project (Claude's free plan allows up to five projects) so it is always the first thing the assistant reads, and update it on the first of each month. Everything in square brackets is yours to fill in; nothing here is a real figure.

BUSINESS FACT SHEET (paste at the top of every prompt)
Company: [name], a window and door installer, nine staff, two fitting crews.
We fit: uPVC and aluminium windows, composite and uPVC doors, sealed-unit replacements for misted glass.
We do not do: roofline, conservatory roofs, commercial glazing.
Areas we cover: [Town A], [Town B], [Town C]. Nothing outside [distance] of [Town A].
Survey: free, about [X] minutes, written quote within [X] working days.
Lead time: [X] weeks from order to fitting, confirmed with suppliers on [date].
Guarantee: [X] years on frames and fitting, details in our written terms. Never say "lifetime".
Ratings: figures come from [supplier certificate], stored in [folder]. Quote only those, naming the product range.
Finance: [yes with provider / no]. Do not mention finance if "no".
Accreditations: [only those we hold, with numbers].
Never claim: heating-bill savings, "best in the area", "cheapest", urgency we cannot honour.
Voice: plain, friendly, specific. No exclamation marks, no emojis.
Photos: only jobs listed in [permissions sheet].

Two lines earn their place. The "never claim" line gives the assistant a rule to follow instead of a mood to guess, and the lead-time line stops the most common invention: a confident "fitted within two weeks" that your supplier cannot meet. If either is blank, the AI fills the gap with something plausible.

Twelve months, one job per month

Each month needs one main asset and one check. Resist the urge to run everything at once; a two-crew company cannot fit a flood of orders, and a long queue produces the kind of review you do not want. This is the plan, with what the AI drafts and what stays with you.

Autumn: the draught campaign without a savings promise

When the heating comes on, homeowners notice cold spots, rattling sashes and handles that no longer pull tight. That is the opening for the autumn campaign, and it can be strong without a single energy-saving figure. Sell what the customer can feel and what you will do at the survey: look at the seals, test the handles, check the sealed units, and give a written quote.

The trouble is that the first draft of almost any heating-related message from an AI includes a saving. Savings depend on the house, the old windows, the heating system, the weather and how people live, so no installer can promise one in advance. Before anything goes out, run every claim through a table like this.

Now the email itself. This prompt goes to past enquirers who never ordered, which is often your cheapest audience (more on that in the cost table below). If you use Mailchimp, its free plan stops at 250 contacts and 500 sends a month, so a bigger past-enquiry list needs a paid plan; Essentials starts at $13 a month on the pricing page.

[paste the fact sheet]
Write an email of about 150 words to people who asked us about windows in the last 18 months and did not order.
Month: October. Angle: draughts and cold spots people can feel. Offer: a free survey including a seal-and-handle check.
Include the current lead time and the booking link [link]. Do not mention heating bills, savings or ratings.
One clear call to action. Subject line: three options under 45 characters.

A realistic first draft comes back warm, readable and wrong in three places.

Illustrative AI first draft: Feeling the chill? As temperatures drop, old windows can cost you hundreds in wasted heat. Our premium units are among the most energy efficient on the market, and with our lifetime guarantee you can relax for good. Book this month for fastest fitting!

The fixes are the claims table in action. "Cost you hundreds" is a saving claim, "among the most energy efficient on the market" is a comparative claim with nothing behind it, and "lifetime guarantee" contradicts the fact sheet. After editing:

After editing: Where do you feel the cold? Many of the calls we get in October are about draughty frames, handles that no longer pull tight and glass that has misted between the panes. Book a free survey and we'll check the seals, test every handle and look at the sealed units, then send a written quote within [X] working days. Orders placed by [date] are fitted before [date]. Book here: [link].

One caution on that edit: "many of the calls we get in October" must be true for your firm, and your log will tell you. The AI cannot know it. For how to answer the enquiries this email brings in, see replying fast and booking the survey.

Spring: a refresh campaign built on before-and-after jobs

Spring is the appearance season. Homeowners look at the front of the house, think about the door and the old frames, and some start to plan work before summer. The refresh campaign trades on what people can see, so its raw material is your own photos: the same front elevation before and after, a door in a new colour, a bay finished in a different frame.

Give the AI the captions job, not the creative job. Describe each photo pair in one line (what was there, what you fitted, how long it took) and ask for three caption options at different lengths. Keep the descriptions factual, because the AI will embellish a thin one.

[paste the fact sheet]
Photo pair 1: before, three original timber windows with peeling paint on a 1970s semi; after, three uPVC casement windows in anthracite grey, same openings. Fitted in one day by two people.
Write three captions for a profile post: one under 20 words, one about 40, one about 70. Mention the one-day fit. No claims about energy or price. End each with "Free survey: [link]".

The typical output adds "transformed the whole look of the home" and sometimes "saving the family a fortune". Strike the second at once; keep the first only if the photos support it. A visual aid helps here: the idea in a shutters visualiser that shows customers the look before you quote carries over to frame colours and door finishes, and is worth testing for the spring push.

Spring is also when a short before-and-after run does more for you than a long ad: five jobs, five posts, one a week, each pointing to the survey. If you have fewer than five jobs with written permission, start collecting the permission on the next five jobs rather than posting without it.

Summer: sell the survey slot while the lead time is still short

June to August is quiet in the illustration (46 enquiries), which makes it the cheapest time to fill the survey diary. A customer who books a survey in July and orders in August is fitted before the autumn rush; one who calls in October joins the queue. So the summer message is practical, not seasonal: book a survey now, order early, be fitted before the cold.

[paste the fact sheet]
Write a 120-word email to people who had a quote from us in the last 12 months and have not ordered.
Angle: our survey diary has space in the next [X] weeks and the current lead time is [X] weeks, so an order placed now is fitted before [month].
Tone: helpful, no pressure, no discount, no countdown. Include one line asking if anything about the quote put them off.

That last line matters more than the pitch. Replies to "did anything put you off?" tell you whether the problem is price, timing, trust or the quote itself, and an AI can sort a dozen replies into themes in a minute. Paste the replies without names, and ask it to group them and give a count for each group. The count is the point: it turns a vague sense of "price objections" into "nine of fourteen mentioned the deposit".

February and July: misted units, reviews and referrals

Two of the quieter months do real work if each gets one small job. February suits an offer that costs the customer less than full replacement: swapping sealed units where the glass has misted between the panes. A household can fix a problem it can see without buying new frames, and it puts a fitter in the house, where the visit may turn up a bigger job. The AI drafts the post; you supply the price from your supplier and the conditions, such as frames in sound condition and size limits.

[paste the fact sheet]
Write a profile post of 80 words for February about replacing misted sealed units without replacing the frame.
Facts: price from $[X] per unit fitted, supplier lead time [X] weeks, frames must be sound (we check at the visit).
Say we will tell the customer honestly if the frame needs more than the glass. No energy claims. End with the booking link [link].

Illustrative output: Misted between the panes? Often the glass unit alone can be replaced, not the whole frame. Units from $[X] fitted, ready in about [X] weeks. We check the frame when we visit and tell you straight if it needs more than glass. Free survey: [link].

Check two things before it goes out: "often" has to be true for your frames, and the "from" price must match your supplier's current list. The honesty line is the one to keep; it is the reason someone picks you over a firm that quotes for a full replacement by default.

July is for asking. Send the review request the day after fitting, while the job is fresh, and keep it short enough for one text segment with plain characters only. Send it to every customer, not only the ones who seemed pleased: Google's review policy does not allow selectively asking for positive reviews, and it bans incentives for leaving one.

Hi [first name], thanks for choosing [Business]. If you have a minute, a short review of how the job went helps us a lot: [link]. Reply STOP to opt out.

Keep any reward well away from the review request. For referrals, write the rule on the card itself and pay the thank-you after the referred job is fitted, not when they first enquire. That way you never pay for a lead that goes nowhere, and nobody argues about when the voucher is due.

Case studies from job notes, with written permission

A case study is the most persuasive page a window company can publish, and the cheapest to make if the notes are captured on the day. Ask the fitting crew to fill in a five-minute form at handover: what the customer's problem was, what was fitted, how long it took, anything awkward and how it was solved, and whether the customer has agreed in writing to photos and a quote. Then give the AI the notes, not an open brief.

JOB NOTES (illustrative)
Property: three-bed semi, 1970s. Customer first name: [first name]. Written permission for photos and quote: YES, form dated [date].
Problem: nine original timber windows, three with failed putty, one that would not close; front door draughty and hard to lock.
Fitted: nine uPVC casement windows in white, one composite front door in [colour]. Two-person crew plus one helper, two days.
Awkward: one bay needed a steel lintel check before fitting; our surveyor flagged it at the survey.
Customer words (approved in writing): [their own words].
Area to name: [town only], not the street.

[paste the fact sheet]
Write a 250-word case study from these job notes: problem, what we fitted, how the job went, and what the customer said (use their words only).
Use only facts in the notes. Do not add figures, savings, awards or the street. Finish with "Free survey: [link]".

The first draft usually does three things you must undo. It invents a number ("cut heat loss by a third"), because case studies in its training data have them; it names the street or house number; and it adds a quote that sounds like a customer but was never said. Delete all three. What remains, the awkward bay and how the surveyor caught it, is the part a generic competitor cannot copy, and it shows how you work. The how-to for turning satisfied customers into proof is covered in contractor reviews, case studies and local posts.

The consent checklist before anything is published

• Written permission that names what you may use: photos, a quote, the first name, the town.
• No house number, street name or anything else that identifies the address; check the background of every photo.
• No price unless the customer agreed to it in writing.
• The customer's right to ask you to take it down, and a promise that you will.
• A copy of the form stored with the job record, and the permission noted in your photos sheet.

Privacy rules differ by place, so have an adviser look over your consent wording once, then reuse it. The asking itself is easier at handover than a month later; and while you have the customer's attention, asking for a review after every job belongs in the same five minutes.

Google Business Profile posts scheduled a month ahead

A profile post costs nothing, and the profile can schedule posts in advance, including repeating ones, so one prompt in the first week of the month can cover the whole month. As last checked in October 2026, posts are archived after about six months unless they are dated, so a rolling plan beats a burst. Google's guidance also discourages auto-generated post content, which is a reason to treat the AI as a first draft: add a real photo and one real detail from a real job to every post.

[paste the fact sheet]
Write four profile posts for October, one per week, each under 100 words, for a window installer.
Week 1: free survey including a seal-and-handle check. Week 2: a sealed-unit replacement for misted glass. Week 3: case study [one line]. Week 4: orders placed by [date] are fitted before [date].
Each post ends with a plain call to action that points to the booking button. No heating-bill claims, no emojis.

Edit each post against the claims table, schedule all four in one sitting, and add the photo yourself. For a deeper routine on keeping the profile fresh, including review replies and questions, see writing profile posts with AI every week.

Which campaign paid? A cost-per-sale sheet

Without a record of where each enquiry came from, the year's plan is a guess dressed as a calendar. Add one field to your enquiry form and your phone script, "How did you hear about us?", with fixed options that match the campaigns (spring leaflet, autumn email, profile, referral, other) rather than free text. When people answer vaguely, an AI can tidy a month of free-text answers into your categories, but a drop-down is better than tidying.

Then keep one sheet per year. Below is an illustrative one for the same nine-person installer. Spend includes ad costs, printing and tools; the profile line counts two hours a month at $30 an hour, because "free" channels still cost time. Of 289 enquiries in the year, 226 carried a campaign code; the other 63 came from word of mouth and the website alone.

Read it in three passes. First, volume: autumn produces the most sales (12) and, at $300 each, is worth keeping at full size. Second, cost: spring is the dearest way to a sale at $343, so before spending the same again, change the offer or trim the ad budget. Third, relationship channels: the emails and postcards to past enquirers cost $52 a sale and referrals $120, which says the cheapest customers are the ones who already know you. Against an average job of, say, $6,200, none of these is alarming; the table is for deciding what to repeat, not for panic.

The pre-winter line has a quirk worth noticing. It has the fewest enquiries but the best survey-to-sale rate of the four seasonal campaigns (4 of 9, 44 per cent), which is what a real deadline can do. That only works if the deadline is real, so it must match the lead time on your fact sheet. Give the AI the sheet and ask for three things: the best and worst campaign by cost per sale, one change to try for each, and a warning where the sample is too small to trust. A line with four sales is a hint, not a result.

A 90-minute routine for the first week of each month

1. 10 minutes: update the fact sheet (lead time, offer, diary gaps, any new certificate).
2. 15 minutes: add last month's enquiries, surveys and sales to the cost sheet; ask the AI to flag anything odd.
3. 25 minutes: draft the month's assets from the plan table: email, posts, captions, leaflet text.
4. 25 minutes: run each draft through the claims table, fix it and add the real photo or detail.
5. 15 minutes: schedule the posts, load the email, and book the reply slots for the enquiries it will bring.

Over a year that is eighteen hours, and most of it is checking, which is where your knowledge of the business matters most. The AI saves the blank-page time. It cannot know your lead time, your guarantee or what your crews can fit, and that is exactly why the fact sheet sits at the top of every prompt.

Questions window companies ask about seasonal marketing

Can I tell customers new windows will cut their heating bills?

Only with documents behind it. Quote rated figures from the manufacturer's certificate for the exact product, say what the figure measures, and avoid promising a saving, because bills depend on the house, the old windows, the heating and how people live. Check the advertising rules where you trade before you publish any figure.

How far ahead should I plan a seasonal window campaign?

Work back from your lead time. If your suppliers need a given number of weeks between order and fitting, an autumn push has to start taking survey bookings that many weeks before the cold arrives. Plan a quarter at a time and refresh the lead time in your fact sheet every month.

Can I post AI-written text on my Google Business Profile?

Use it as a first draft, not a finished post. Google's guidance discourages auto-generated post content, so rewrite each draft with a real photo and one detail from a real job. The profile can schedule posts in advance, which makes a weekly rhythm easy to keep.

Do I need permission to use a customer's job photos in marketing?

Get it in writing before you publish. Cover the photos, any quote, how you will describe the location, and the customer's right to withdraw. Leave out house numbers, faces and street names. Privacy rules vary by place, so ask an adviser to check your consent wording.

Further reads

• Insulation Contractor Marketing: Autumn Campaigns and Local Pages With AI — Autumn campaigns and local pages for a neighbouring trade.
• AI Marketing for Plumbers: A Worked 12-Month Posts and Offers Calendar — A 12-month posts and offers calendar for another home trade.
• Painting Contractor Marketing: Seasonal Campaigns With AI Content — Spring and autumn content ideas for a visual trade.
• Flooring Contractor Marketing: Seasonal Offers and Posts Drafted With AI — Seasonal offers drafted with AI, with a different job mix.
• How to Plan a Month of Marketing Content With AI in One Hour — A one-hour method for planning each month's content.
• HVAC Google Business Profile: Service Areas, Q&A and Review Replies With AI — Service areas, questions and review replies on a trade profile.
• Glazier Software: Quoting and Scheduling Tools With AI Features — Three glass-specific tools compared on measure, quote, order and fit, with published prices, vendor AI claims to test and a one-week trial checklist.
• Window Measurement App: Measure and Quote on the Doorstep With AI Help — Windowmaker Measure, PrefMeasure, DoorWin, Aufmaß App and others compared, with a phone-against-laser test and a survey sign-off for installers.
• Gutter Cleaning Marketing: Eight Spring and Autumn Campaigns With AI — A Leaf Week calendar of eight campaigns, with AI prompts, sample copy and a tracking sheet so you know which gutter cleaning marketing booked jobs.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Mailchimp's pricing page (checked 1 November 2026) and Google Business Profile help pages on posts (as last checked in October 2026). All campaign spend, enquiry and sales figures are illustrative.