Jeff Bullas

Here’s the repeatable, 90‑minute “triangulation” to get a defensible market size with public data. AI does the grunt work; you verify the 1–2 inputs that move the needle.

What you’ll bring

• A precise market statement: who buys, what they buy, where, and the period (annual).
• Three public anchors you can quickly find: population or company counts, prices, and at least one comparable company’s revenue.
• Tools: browser, spreadsheet, AI chat, and a short list of assumptions to track.

The 3-anchor method (simple, credible)

1. Demand bottom-up (PPP ladder): Population → Participation → Pay.
   • Customers = Addressable population × Interest/eligibility × Paid conversion.
   • ARPU = Average price × Purchase frequency (per year).
   • Revenue = Customers × ARPU.
2. Top-down sanity check: Start from a related category spend or population and apply realistic penetration and share-of-wallet.
3. Comparable-company triangulation: Use 2–3 public companies: if one player has $X revenue and ~Y% share, implied market ≈ X / Y%.

Spreadsheet layout (copy into a sheet)

• Inputs: B2 Addressable_population, B3 Interest_rate, B4 Paid_conversion, B5 Avg_price, B6 Purchase_frequency.
• Formulas:
  • B7 ARPU = =B5*B6
  • B8 Customers = =B2*B3*B4
  • B9 Revenue = =B8*B7
  • Scenario cells: set C2:E6 with conservative/base/optimistic inputs and repeat the above formulas in each column.
• Quick sensitivity: create three rows that change Paid_conversion by −20%, base, +20% and show the resulting Revenue values.

Insider tricks to find public numbers fast

• Use search patterns: “keyword + market size + report”, “keyword + 10-K”, “keyword + investor presentation”, “keyword + pricing”, “number of locations + keyword”, “trade association + keyword”.
• Anchor participation with surveys or platform stats; anchor price with visible pricing pages or average order values mentioned in filings.
• Cross-check with per-capita spend: Revenue / population. If it implies an unrealistic spend per person, revisit assumptions.
• Supply-side proxy (if two-sided): Providers × Capacity × Utilization × Price → a second triangulation against your demand model.

Robust, copy‑paste prompts (use as-is)

Master prompt — full market size with checks

“Act as a market sizing analyst. Estimate the annual market size for [product/service] in [region] for [year]. Deliver:
1) Two approaches: top-down (category revenue or population) and bottom-up (Customers = Addressable population × Interest rate × Paid conversion; ARPU = Average price × Purchase frequency; Revenue = Customers × ARPU). Show the math with named variables.
2) Three scenarios (conservative/base/optimistic) with explicit percentages and a short justification for each.
3) Sensitivity: show how Revenue changes when [Paid_conversion] and [ARPU] vary by ±20% (small table).
4) Comparable-company triangulation: list 2–3 comparable public companies or well-covered private ones, their last reported revenue, and the implied market size if they hold [assumed]% share.
5) Spreadsheet-ready output: provide an input table and Excel formulas using cell references and a version using named variables.
6) Sources: list likely public sources and exact search queries I can run. If you are unsure or cannot access a source, label the figure as ‘assumption’ and explain how to verify it.
7) Call out the top 3 assumptions to verify and how to verify each.”

Rapid variant — I’m pasting excerpts

“I will paste quotes from public sources. Extract numeric data and map to these variables: Addressable_population, Interest_rate, Paid_conversion, Avg_price, Purchase_frequency, Comparable_revenue, Comparable_share. Build bottom-up and top-down estimates, three scenarios, and a short sensitivity on Paid_conversion (±20%). Flag any missing variables as ‘assumptions’ and suggest how to find them. Then output the spreadsheet input table and formulas.”

Supply-side triangulation prompt

“Using a supply-side view for [market], estimate: Providers × Capacity per provider × Utilization × Price to cross-check demand estimates. Show conservative/base/optimistic, and highlight any mismatch with the demand model greater than 30%, plus likely reasons (seasonality, informal supply, channel mix).”

Short worked example (structure you can mirror)

• Market: UK language-learning subscriptions, annual.
• Assume adults = 54M; Interest = 6%; Paid conversion = 4%; Avg price = £8/month; Frequency = 12.
• ARPU = £8 × 12 = £96; Customers = 54,000,000 × 0.06 × 0.04 = 129,600; Revenue (base) = 129,600 × £96 ≈ £12.4M.
• Conservative: Interest 4%, Conversion 3% → ≈ £6.2M. Optimistic: Interest 8%, Conversion 5% → ≈ £20.7M.
• Sanity check: If a known player reports ~£4M UK revenue and holds ~25% share, implied market ≈ £16M — within range. Good sign.

Common mistakes and fast fixes

• Double counting audiences (free users vs paying). Fix: model only payers; keep free users as a separate stat.
• Mixing monthly and annual. Fix: standardize on annual in your sheet; convert everything before you calculate.
• Fantasy penetration. Fix: anchor participation and conversion to a public survey or a comparable company’s disclosed metrics.
• Single-source bias. Fix: apply the 2‑source rule for every critical input.
• No sanity check. Fix: compare to per-capita spend and to at least one comparable company’s revenue.

Action plan (60–90 minutes)

• 15 min: Define the market; list variables and initial assumptions.
• 25 min: Run the master prompt; capture the inputs, scenarios, and sensitivity.
• 20 min: Build the spreadsheet with your variables and formulas; create three scenarios.
• 20–30 min: Verify the most sensitive input (usually conversion or price) with 1–2 public sources; update the range.

Bottom line — You’re aiming for a tight range, not a magic number. Let AI assemble the math and options; your job is to choose sensible assumptions, verify the few that matter, and state them plainly. That’s what makes your estimate defensible.

Nice call-out: you nailed it — AI plus hypothesis-driven A/B testing is the combo that turns ideas into measurable wins. I’ll add a few practical tweaks to speed results and reduce noise.

Quick context

If you follow your plan but tighten the statistical and experiment design bits, you’ll avoid false positives and get repeatable lifts. Below is a simple playbook you can copy this week.

What you’ll need

• Landing page URL or HTML, baseline conversion rate, and weekly traffic by source
• Primary KPI (e.g., leads per visitor) and one customer persona
• A/B tool (CMS split, VWO, Optimizely) and analytics access
• Basic tracking events for CTA clicks and form completions

Step-by-step — do this

1. Collect baseline: 7–14 days of traffic + conversions by source.
2. Pick one clear test: headline or CTA wording. Only one variable.
3. Use an AI prompt (examples below) to generate 3 focused variants: clarity, urgency, social proof.
4. Write simple hypotheses: “Changing headline to outcome-first will lift CR by 10% vs control.”
5. Define stop rules: run until 95% confidence or at least 100 conversions per variant (minimum).
6. Launch, monitor early signals (CTR, bounce), but don’t stop early.

Copy-paste AI prompt — landing copy

“You are a senior B2B copywriter. Write three distinct landing page variants for a SaaS product that automates marketing reports for small teams. Each variant: 6–9 word headline, 12–18 word subhead with main benefit, two 8–12 word bullets tying features to outcomes, one 2–3 word CTA, and one short social-proof sentence. Tone: clear, confident, non-technical. Audience: marketing managers at 10–50 person companies.”

Copy-paste AI prompt — A/B test ideas

“Suggest five A/B test hypotheses for a landing page with current headline X and CTA Y. For each: hypothesis statement, expected impact (high/medium/low), required minimum sample, and one quick implementation note. Keep answers short and practical.”

Example outputs (sample headlines)

• “Stop Manual Reports — See Metrics Automatically Today”
• “Automate Your Marketing Reports in Minutes, Not Hours”
• “Reports That Tell You What To Do Next”

Common mistakes & fixes

• Testing many changes at once — fix: isolate one variable.
• Stopping early on a lucky day — fix: set min conversions or time window.
• Ignoring segments — fix: check results by source and device.

1-week action plan

1. Day 1: Pull baseline and choose KPI.
2. Day 2: Run the landing copy prompt and pick 1 variant.
3. Day 3: Build variant, set tracking.
4. Day 4: QA and launch.
5. Day 5–7: Monitor signals; commit to the stop rules.

Small, fast experiments win. Focus on clarity, test one thing, and let the numbers teach you.

Yes to the 50‑row quick win — that fast test tells you if the AI is worth your time. Let’s now turn it into a reliable, audit‑ready system you can run in an hour a month and expand to a full year without drown‑in‑paper pain.

Big idea: pair AI with a few simple “rules” (vendor map + mixed‑use allocations + evidence notes). That’s what turns clever categorization into cleaner deductions and fewer questions later.

What you’ll need

• CSV exports for the period (date, description, amount, merchant).
• Receipts or scans for higher‑value/ambiguous items.
• Your mileage log (or at least odometer start/end and trips) if you drive for work.
• Home office details (square footage dedicated to work vs. total home space).
• A simple spreadsheet and an AI chat that accepts pasted rows.

Do this step‑by‑step

1. Create the category set that matches tax forms. Keep it simple: advertising/marketing, software, supplies, professional fees, education, utilities, phone/internet, meals, travel, home office, mileage/vehicle, other. The goal is to mirror what shows on your filing forms so the handoff is smooth.
2. Build a “vendor map” once, reuse forever. List common merchants and the default category. Add a short “business‑purpose question” for edge cases (e.g., meals). You’ll fix 80% of classification with this one move.
3. First AI pass: categorize with your vendor map. Run your CSV through the prompt below. Mark anything uncertain as REVIEW. Add one‑line notes (who/why/date) right in a spreadsheet column.
4. Handle mixed‑use items with simple percentages. Phone/internet, software bundles, and shared services often require a business‑use %. Apply a conservative split (e.g., 60% business) and keep a one‑line rationale.
5. Choose one car method and stick with it. If you use the standard mileage rate, don’t also deduct fuel/repairs separately; if you use actual expenses, then track those and don’t claim mileage rate. Ask AI to total the right bucket and exclude the rest.
6. Home office: capture the basics. Dedicated square feet and total home square feet. AI can compute a simplified approach vs. actual expense approach so you can decide which draft to take to your tax pro.
7. Receipts: extract key fields and auto‑create an audit memo. For any big/ambiguous item, paste the text or image OCR. Store the AI‑generated memo with the receipt.
8. Variance check. Ask AI to compare month‑to‑month category totals and flag anomalies, duplicates, or personal‑looking items. Clean before you total.
9. Export deliverables. Totals by category, a REVIEW list with your notes, and a folder holding CSV + receipts + memos. That’s your “audit binder.”

Copy‑paste prompt: build your vendor map and categorize

“I’m a freelancer reviewing expenses. Here are CSV rows (date, description, amount, merchant). First, produce a VENDOR MAP from these rows: for each recurring merchant, output: merchant name, common patterns/aliases, default category (marketing, software, supplies, professional fees, education, utilities, phone/internet, meals, travel, home office, mileage/vehicle, other), and a short ‘business‑purpose question’ if meals/travel/ambiguous. Then, use that map to: 1) Categorize each transaction, 2) Mark uncertain items as REVIEW with the reason, 3) Suggest a conservative business‑use % for likely mixed‑use items (phone/internet/software bundles), and 4) Output totals by category plus a list of REVIEW items with the exact follow‑up question I should answer. Do not give tax advice; just categorize, flag, and total.”

Bonus prompts (use when needed)

• Receipts to audit memo: “Extract from this receipt text: date, vendor, amount, item(s), payment method, and create a 1‑line business purpose memo. If it looks personal, label REVIEW and ask me what’s missing.”
• Mileage summary: “Here are my trip notes (date, start/end, purpose, miles). Total business miles by month and year. If any trip lacks a business purpose, mark REVIEW and ask me the missing detail.”
• Home office calculator: “I have X sq ft dedicated office and Y sq ft total home. List the two common ways to estimate a home office deduction in plain English and show a simple calculation for both with the info I provide. Label as a draft to discuss with my tax preparer.”

Example: side‑gig photographer

• Vendor map sets “Adobe, Capture One” to software; “B&H, Best Buy” to equipment/supplies; “Instagram ads” to marketing; “Zoom” to communication.
• AI flags “Starbucks” as meals REVIEW — you add: “Client consult 2025‑03‑11.”
• Phone/internet split at 60% business with a memo: “Client calls, uploads, galleries.”
• Mileage chosen over fuel — AI excludes gas/oil transactions to avoid double counting.

Mistakes to avoid (and quick fixes)

• Mixing vehicle methods. Pick mileage or actual expenses, not both. Fix: tell AI which method you’re using and have it exclude the other bucket.
• Forgetting mixed‑use splits. Phone/internet/software bundles are rarely 100% business. Fix: apply a conservative % and keep a one‑line memo.
• Double counting reimbursements or transfers. Fix: tag as NON‑DEDUCTIBLE and remove before totals.
• Loose receipts. Fix: run the receipts prompt, save memo + receipt in the same folder named YYYY‑MM_Category_Vendor_Amount.
• Personal creep in business accounts. Fix: mark as personal; do not include in totals. Consider a separate card for business going forward.

What to expect

• 80–90% of transactions auto‑categorized after your first vendor map.
• A short REVIEW list each month that you can clear in minutes with one‑line notes.
• A tidy package your tax preparer can use directly, with fewer follow‑up emails.

Action plan (90 minutes this week)

1. Export last 2–3 months of CSVs; gather big/ambiguous receipts.
2. Run the vendor‑map prompt on 100–200 rows; accept categories; mark REVIEW.
3. Add one‑line business‑purpose notes; apply mixed‑use % where needed.
4. Run the receipts and mileage prompts for gaps.
5. Produce category totals, a REVIEW list, and save everything in one folder. Repeat monthly.

Remember: AI is the tireless sorter; you’re the final reviewer. Small, consistent habits — vendor map, one‑line memos, and method discipline — turn AI speed into real savings and cleaner records.

Learning to manage this is a necessary skill for any YouTube creator.

Short Answer: The best strategy is a firm “no engagement” policy combined with proactive use of YouTube’s moderation tools. You should hide, block, and filter users and keywords rather than replying to them.

Your primary goal is to control the text format of your comments section to protect your community’s health.

The most effective approach involves managing the text format of your YouTube comments section without directly engaging. Firstly, do not reply to trolls; this only boosts their comment’s engagement and encourages them. Secondly, make aggressive use of the moderation tools YouTube provides by creating a “blocked words” list, which is a custom text format filter that automatically hides comments containing common insults or slurs. Thirdly, the most powerful tool is to “hide user from channel,” which makes all of their past and future comments invisible to everyone except them, effectively removing them from your community without notifying them. By focusing on filtering and hiding these negative text formats rather than responding to them, you maintain control and protect your own mental energy.

Cheers,

Jeff

These are powerful tools when integrated properly into your stream.

Short Answer: The most effective method is to actively integrate them into your live content. This involves acknowledging every contribution and creating specific calls-to-action that give viewers a reason to use them beyond just a simple donation.

Treating these features as interactive content formats, rather than just as donation buttons, is the key to encouraging their use.

You need a clear strategy for how you will handle these formats during your YouTube live stream. Firstly, you must acknowledge every Super Chat text format and Super Sticker image format with a verbal thank you and a direct response to their message; this validation is the most powerful incentive for others to contribute. Secondly, you should build segments of your live video format around them, such as a Q&A where questions sent via Super Chat are prioritised, or allowing a Super Chat to influence the stream’s direction. Thirdly, you can use a goal, a simple data format which you can mention or display on screen, to gamify the experience by setting a Super Sticker target that unlocks a special event when reached. This transforms them from a tip jar into a core part of the interactive YouTube experience.

Cheers,

Jeff

That’s a smart piece of data to focus on.

Quick Answer: This graph is your best guide for timing your content. The standard strategy is to publish your videos one to two hours before the peak viewing times indicated on the chart.

Effectively using this data format is all about giving your various content formats the strongest possible launch velocity.

This chart is a powerful data format that shows you when your specific audience is most active, and you should use it to schedule all of your content. Firstly, for your primary video format, publishing an hour or two before the darkest purple bars appear gives YouTube’s systems time to index your video and prepare it for distribution just as your viewers are starting to arrive on the platform. Secondly, this data is even more critical for a live stream format, where you should aim to go live precisely as those peak hours begin to maximise your potential for concurrent viewership. Finally, don’t neglect your text formats; timing your Community Tab posts to coincide with these high-traffic periods will ensure they get the most immediate engagement from your most active subscribers, boosting their visibility.

Cheers,

Jeff

An astute question—a poor approach can close a door before it even opens.

Short Answer: The standard etiquette is to build a genuine, public relationship first before sending a private, personalised pitch that clearly outlines the mutual benefit.

Let’s discuss how specific content formats play a critical role in each stage of this process.

Your approach should be a deliberate sequence of content interactions. First, your initial engagement must be through public content formats; spend weeks leaving thoughtful text replies on their threads or reposting their video content with your own valuable commentary to show you respect their work. Second, when you send a direct message, it must be a concise and personalised text-based format that gets straight to the point, and for complex ideas, you might link to a private video or a short document that visually outlines the proposed collaboration. Finally, the collaboration itself should be a strategic blend of your strongest content formats, whether that’s a co-hosted X Space, a joint video interview, or a co-authored educational thread. The most harmful practice is sending a cold, generic DM with no prior interaction; this is spam and will be ignored by any serious creator.

Cheers,

Jeff

This is an essential topic that is too often overlooked.

Short Answer: The core best practices are providing text alternatives for all non-text content, ensuring text is readable, and making video content usable for visually and hearing-impaired users.

Let’s detail the specific accessibility requirements for each primary content format on X.

To ensure your content is accessible, you must address each format’s specific needs. First, for every image format you post, from photos to infographics, you must write descriptive alt text that clearly communicates the meaning and context of the image for those using screen readers. Second, for all video formats, it is a professional standard to include accurate and synchronised closed captions for the hearing impaired, and for videos where visuals are critical to the message, you should also provide a separate audio description. Finally, even your text-based formats require attention; use clear language and for any multi-word hashtags, you must use CamelCase, which means capitalising the first letter of each word like #ThisExampleHere, to ensure screen readers can interpret them correctly. Neglecting these practices is a failure to communicate effectively and excludes a valuable part of your potential audience.

Cheers,

Jeff