NoticeKeeper

1. One-liner

Remembers which payee already got a first B-notice, so the second one goes out right and on time.

2. Trend signal — why now?

Three things moved in the last twelve months, and they compound.

The IRS explicitly refuses to carry the state for you. Publication 1281 asks the question directly and answers it flatly. Question 21: “Does a CP2100 or CP2100A Notice indicate whether it is the first or second notification of an incorrect TIN for a specific account?” Answer: “No. The backup withholding regulations require that payers be responsible for tracking the status of the notices they receive.”

That single sentence is the entire product thesis. The notice you get in September looks identical whether it is the payee’s first appearance or their second. But the required response diverges completely: a first notice means you send the First “B” Notice with a Form W-9; a second notice within three calendar years means you send the Second “B” Notice and, per Pub 1281, “Do not include a Form W-9” — because only an SSA or IRS validation letter clears it. Send the wrong one and your solicitation doesn’t count.

The filing system underneath all of this is being ripped out mid-cycle. The IRS states that “Due to the planned retirement of the Filing Information Returns Electronically (FIRE) System, the IRS is no longer accepting new Information Returns (IR) Applications for Transmitter Control Codes (TCCs).” FIRE closes permanently on 19 November 2026 at 3 p.m. ET, and “IRIS will be the only information returns electronic filing system, including current year, prior year, or corrections, after Jan. 1, 2027.” Existing FIRE credentials do not carry over — filers must obtain a separate IRIS TCC through a new application, and the IRS advises allowing 45 days for processing. So the three-year lookback window that determines first-versus-second straddles a platform migration that resets everyone’s tooling.

IRIS is quietly telling filers they succeeded when they partly failed. The IRS’s own February 2026 IRIS Working Group deck lists, under Resolved Issues, that “The system is returning a Transmission Status of Accepted with Errors instead of Partially Accepted when one or more submission is rejected.” The same deck confirms the top business rules — recipient name/TIN mismatch against the IRS database — resolve to “Accepted with Errors,” not a reject. And per the filing-software guidance, when IRIS marks a batch “Accepted with Errors,” the IRS “does not proactively notify filers”; you must log into the dashboard, open the acknowledgement page, or parse a JSON file the vendor’s own help doc calls “difficult to read.” The mismatch you never saw in January becomes the CP2100 that lands in September.

Provenance:

• Signal 1 (demand): IRS Publication 1281 Q21 — the IRS states payers are responsible for tracking first-vs-second notice status themselves, and the two responses differ materially — https://www.irs.gov/pub/irs-pdf/p1281.pdf — observed 2026-09-05
• Signal 2 (feasibility/urgency): FIRE retires 19 Nov 2026; IRIS becomes the sole system after 1 Jan 2027; FIRE TCCs do not transfer — https://www.irs.gov/node/880 — observed 2026-09-05
• Signal 3 (economic): CP2100/CP2100A notices mail “twice a year, in September and October and again in April”; uncorrected TIN errors carry $340/form after Aug 1 and 24% backup withholding liability — https://www.irs.gov/newsroom/irs-sends-cp2100-and-2100a-notices-when-payers-need-to-correct-backup-withholding-errors — observed 2026-09-05 Category: Regulatory arbitrage (a stateful duty the regulator explicitly refuses to track for you) + Underserved niche (the managed-service incumbent starts at 200 payees)

3. The opportunity

Every vendor in this category sells the same thing: verification before you file. TINCheck’s own marketing line is “Verify before you file, not after the notice arrives.” Tax1099, Track1099 and 1099FIRE all sell per-lookup TIN matching at $0.37–$0.50 a record. That is the capture side, and it is thoroughly productised.

The defence side is not. Once the notice arrives, the payer is running a stateful, multi-year rules engine by hand:

• Is this payee’s first or second appearance within three calendar years? The notice won’t say.
• If it’s a second CP2100 in the same calendar year as the first, Pub 1281 Q17 says you may disregard it entirely — a rule that saves work but only if you know you’re allowed to use it.
• Send the “B” notice within 15 business days of the notice date or the date received, whichever is later.
• Begin backup withholding at 24% “no later than 30 business days after you have received” the notice; stop within 30 calendar days of receiving the W-9 or, for a second notice, the SSA/IRS validation.
• If a notice comes back undeliverable, you begin withholding and keep the undelivered envelope on file for three years specifically to track the two-in-three-year rule.

Miss the 15-day window and you haven’t just missed a deadline — you’ve damaged the reasonable-cause defence that is your only shield when the 972CG proposed-penalty notice arrives in the autumn. Penalty relief here is unusually unforgiving: the IRS’s own guidance notes that first-time abatement is not available for information reporting penalties, and every reply to a 972CG must be in writing. Your documented solicitation history is the case.

This is the capture-vs-defence gap. The incumbents own the moment the data goes in. Nobody self-serve owns the eighteen months afterwards when the state has to survive.

Honest caveat, stated up front: there is a direct incumbent. TIN Comply sells exactly this remediation, and sells it well — mismatch triage, 1st/2nd classification, deadline calculation, W-9 collection, second-notice SSA/IRS validation tracking, plus a free browser-based triage tool as a funnel. But it is a managed service: humans, fixed fee quoted after an intake assessment, “1–4 weeks depending on payee count,” and explicitly “Best fit for 200 – 200K payees.” Their free tool runs entirely in-browser, stores nothing across sessions, doesn’t integrate with filing, and doesn’t produce mailable notices — it is a lead magnet that hands you a CSV and points at the paid service.

So the gap is not “nobody does this.” It is: the only real solution is a consultant engagement with a 200-payee floor, and the free alternative deliberately forgets everything the moment you close the tab. The payer with 40 mismatches and a three-year memory problem is holding a spreadsheet.

4. Target market

Primary customer: The controller, office manager, or outsourced bookkeeper at a US business filing roughly 20–500 information returns a year, in a trade with churny unincorporated payees — residential property management, specialty-trade construction, staffing and per-diem labour, medical practices paying locums, trucking outfits paying owner-operators, and the small CPA firms who file on behalf of a book of these clients.

Why they buy: They receive a CP2100A (the sub-50-error version), they have no idea whether any listed payee appeared on last year’s notice, and they have fifteen business days. The current workaround is opening last year’s PDF next to this year’s and eyeballing TINs. When a payee is misclassified as first-time, the wrong letter goes out, the solicitation doesn’t count, and the reasonable-cause argument quietly evaporates.

Rough TAM reasoning: I won’t invent a filer count — I couldn’t source a clean one, and the ground is actively moving. What I can say is bounded and honest: CP2100A is defined as the under-50-error notice, which means the IRS mails a purpose-built notice class at exactly this segment, twice a year. Property management alone is a large, fragmented category where filing 1099-MISC to owners and 1099-NEC to vendors is routine, and where Buildium and AppFolio generate the forms but neither tracks notice state afterwards. A few tens of thousands of reachable payers in the 20–500 band is a defensible working assumption and needs validating, not asserting.

Why now for them: Three reasons stack in the same twelve months. The notices mail this September and October. FIRE dies in November, so whatever ad-hoc tracking lived in their old filing tool is about to be orphaned. And the OBBBA threshold change means the 2027 filing season runs on a different rule than the three-year lookback window they’re being judged against — 2025 filings were $600-threshold, 2026 filings are $2,000-threshold, and the two-in-three-year test spans both.

5. Product sketch (MVP)

• Upload this year’s CP2100/CP2100A extract and last two years’ — get every payee classified first-notice, second-notice, or disregardable-under-Q17, with the Pub 1281 citation behind each call.
• A live deadline strip per payee: 15-business-day send-by date, 30-business-day withholding-start date, and the 30-calendar-day stop-withholding clock once a W-9 or validation letter lands.
• The right letter, generated and mailable — First “B” Notice with Form W-9 attached, or Second “B” Notice with the W-9 deliberately withheld and the SSA/IRS validation instructions in its place, envelope text pre-set to “IMPORTANT TAX RETURN DOCUMENT ENCLOSED.”
• W-9 chase with response capture — trackable link, reminders on the 30-day clock, and the returned form filed against the payee record.
• Undeliverable handling — mark returned mail, auto-start the withholding flag, and hold the artefact for the full three years the rule requires.
• IRIS acknowledgement watch — paste or connect the acknowledgement file, translate the raw business-rule codes (SHAREDIRFORM014/015, S1H001) into plain English, and flag “Accepted with Errors” batches the dashboard wouldn’t have told you about.
• The reasonable-cause export — a dated, per-payee solicitation history packaged as the written reply a 972CG demands.

6. AI angle — what’s load-bearing

Two places, and I’ll be straight that this is not an AI-first product — it’s a state machine with AI at the edges.

The genuinely load-bearing use is document ingestion. CP2100 extracts arrive as pipe-delimited files in one shop and as scanned PDFs in the next; W-9s come back as phone photos, faxes, and half-signed PDFs. Extracting name, TIN, TIN type and signature status reliably from that mess is what makes the 15-day clock survivable, and it is exactly what vision models became cheap and accurate enough to do in the last two years. Without it, the customer is retyping under deadline, which is the failure mode we’re selling against.

The second use is explaining the verdict. Turning “SHAREDIRFORM014, Record ID 4471” into “the individual’s name and SSN don’t match IRS records — this is a first notice, send the W-9 by 26 September” is what the incumbent’s JSON acknowledgement, by its own vendor’s admission, is too hard to read for.

Strip the AI out and a determined developer still ships a rules engine — the deadline maths and the two-in-three-year test are deterministic logic, and they should be, because a hallucinated compliance deadline is worse than no product. That is a real limit on the AI story and I’ve scored it accordingly. AI makes this usable inside fifteen business days; it does not make it possible.

7. Localization angle

N/A — this is a US-only play, and deliberately so. The entire product is the shape of one country’s backup-withholding regime: Publication 1281’s two-in-three-year rule, the 24% rate, the CP2100/CP2100A split at fifty errors, and the IRIS migration timetable. There is no version of this that ports; the equivalent duty elsewhere is a different statute with a different clock. Depth in one jurisdiction is the moat such as it is.

8. Business model — path to $1M–$5M ARR

Pricing: $39/mo for up to 50 payees on notice, $99/mo up to 250, $249/mo for firms managing multiple client EINs. Annual billing pushed hard, because the pain is seasonal and monthly churn after October is the obvious failure mode.

ACV: ~$950 blended, assuming most self-serve customers land on the $99 tier annually and CPA-firm accounts pull the average up.

Rough math to $1M ARR: ~1,050 accounts at $950. Realistically that’s ~800 direct payers plus ~150 accounting firms carrying several client EINs each.

Rough math to $5M ARR: requires either winning the CPA-firm channel properly — per-client-EIN pricing at a few hundred firms — or expanding into the adjacent notices that share the same evidence file: 972CG penalty responses (45-day written reply, reasonable-cause argument) and the annual-solicitation duty for missing TINs. That’s a credible expansion path, but $5M is not the base case and I won’t pretend otherwise.

Expansion path: EIN count, then payee count, then the 972CG response module as the natural autumn upsell to customers who already have the solicitation history sitting in the product.

Where the model is fragile: the pain is bi-annual, not daily. September/October and April are real; June is not. This is a deferred-pain product, and my own catalogue says those convert badly unless paired with something that bleeds continuously. The counter is that the evidence file accrues year-round and is worthless if you start it late — but that’s an argument I’d have to win with customers, not with logic.

9. Go-to-market wedge — first 100 customers

• Ride the notice calendar. The IRS says these mail “twice a year, in September and October and again in April.” That’s a known, dated demand spike. Run tightly-targeted search and LinkedIn against “CP2100A,” “first B notice vs second,” and “backup withholding 15 business days” in the two weeks either side. Intent doesn’t get more explicit than someone typing a notice number into a search bar.
• Beat the free triage tool at its own funnel. TIN Comply’s free classifier is browser-only and forgets everything on close. Ship a free classifier that persists — upload your notice, get the 1st/2nd verdict, and we hold it so next year’s comparison is automatic. The stored state is the upgrade prompt, and it is the one thing their tool structurally cannot offer without cannibalising the managed service.
• Sell to the bookkeeper, not the business. Outsourced accounting and property-management back-office firms (APM Help and its peers publish for exactly this audience) each carry dozens of client EINs and eat the notice pain repeatedly. Twenty such firms is a few hundred EINs. Cold outreach with a specific offer: send us last year’s and this year’s extract, we’ll return the classified list free, and you’ll see immediately how many you’d have got wrong.
• The FIRE-migration hook. Between now and 19 November, every FIRE filer must get a new IRIS TCC and the IRS advises allowing 45 days. Publish the genuinely useful migration checklist, capture the email, and be present in their inbox when the September notice lands.
• Buildium/AppFolio adjacency. Both generate 1099s for property managers; neither tracks notice state afterward. Content and integration aimed squarely at that handoff — “your PM software filed it, now what?” — reaches a concentrated, well-defined audience.

10. Build complexity — justification

Low. A deterministic rules engine over Pub 1281 (the two-in-three-year test, three date clocks, the Q17 disregard exception), document extraction on off-the-shelf vision models, letter generation, and a mail integration like Lob. No IRS system integration is required for v1 — the customer uploads their own extract, which sidesteps e-Services enrolment entirely and is why this stays Low rather than Medium. A capable solo builder ships this in six to eight weeks. The hard part is not engineering; it is encoding the publication correctly and keeping it current, which is domain work.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helping payers comply with published IRS procedure. No filing authority required for v1.
Ethical — no harm / dark patterns	✅	Straightforwardly helps small payers avoid penalties and helps payees get correctly classified.
Market exists (evidence above)	✅	The IRS mails a dedicated notice class at this segment twice yearly; a managed-service incumbent already monetises it above 200 payees.
1–5 person team can build this	✅	Solo builder, 6–8 weeks.
Launchable with <$50K / ₹40L	✅	Well under. Main costs are mail postage pass-through and a domain expert’s review of the rules encoding.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	14/20	Real, dated, and penalty-backed — $340/form after 1 Aug, 24% withholding, damaged reasonable-cause defence. But it bites twice a year, not weekly, and most payers get a small notice and muddle through. Not hair-on-fire.
Demand evidence	15	10/15	Strong institutional evidence: IRS publishes the duty, mails the notice on a schedule, and a managed-service vendor monetises the remediation. Weak on direct customer voice — I could not source verbatim practitioner complaints, and I’m not inventing them. That gap is the score.
Build feasibility	15	13/15	Deterministic rules + document extraction + mail. No IRS integration in v1. Six to eight weeks solo.
Distribution clarity	15	9/15	Notice-number search intent is genuinely high-quality and calendar-predictable, and the bookkeeper channel is concentrated. But the addressable spike is two windows a year, and the free-tool wedge is a fight with an incumbent who already owns that funnel.
Revenue mechanics	15	10/15	Pricing is defensible against $0.37–$0.50 per-lookup TIN matching and against an opaque fixed-fee consultancy. $1M is reachable at ~1,050 accounts. $5M needs the CPA channel to work, which is unproven. Seasonal churn is a live risk.
Time to first revenue	10	7/10	The September/October notice window is imminent, so a fast build can charge inside this season. Miss it and the next real window is April.
Defensibility	10	3/10	This is the weak axis and I won’t dress it up. The rules are public, the incumbent already knows them, and a competent competitor rebuilds the engine in a quarter. The only accruing moat is the stored multi-year notice history per customer, which genuinely does get stickier every September — but that takes two cycles to bite.
Total	100	66/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

You need someone who will read Publication 1281 line by line and encode it exactly. Getting the two-in-three-year rule or the Q17 disregard exception subtly wrong produces confidently incorrect compliance advice, which is worse than no product.

Key assumptions to validate

1. Assumption: Payers in the 20–500 return band genuinely cannot tell first from second notice, and have got it wrong before. How to test: Offer 25 bookkeeping and property-management firms a free classification of their 2024–2026 extracts. Measure what fraction of payees they had classified incorrectly. If it’s under 10%, the pain is imaginary and the product is a vitamin.
2. Assumption: They’ll pay a subscription for a twice-yearly event. How to test: Price-test annual-only against monthly with the first 40 signups. If more than half try to cancel in November, the model is wrong and this becomes per-notice transactional pricing.
3. Assumption: The OBBBA $2,000 threshold doesn’t gut the segment. How to test: Ask the same 25 firms what share of their 1099 volume sat between $600 and $2,000. If it’s over half, the addressable band shrinks materially from the 2027 season onward.
4. Assumption: TIN Comply won’t simply ship a persistent self-serve tier. How to test: Watch their free tool for a login. If it gains accounts and storage, the wedge closes.

Risk flags

1. Direct incumbent, correctly aimed: TIN Comply already does the workflow and understands the domain. The entire bet is that a managed service with a 200-payee floor won’t chase the band below it, and that their free tool stays deliberately amnesiac to protect the consulting revenue. That’s a reasonable bet about incentives, not a moat.
2. Regulatory shrinkage: OBBBA raised the reporting threshold from $600 to $2,000 effective for payments after 31 December 2025, and the IRS’s own estimate is “tens of millions fewer” 1099-NEC and 1099-MISC forms annually. Fewer forms means fewer mismatches means a smaller market — concentrated exactly in the small-payer tail this product targets.
3. Seasonality: Two demand windows a year against a monthly subscription is a structurally awkward fit. Deferred-pain compliance products convert badly without a continuous bleed alongside them, and this one doesn’t have one.
4. Thin defensibility: Public rules, no proprietary data, no network effect. The stored notice history compounds, but only after the customer’s second season.
5. Advice liability: Telling someone to send a Second “B” Notice when it should have been a First is a compliance error with their name on it. Needs careful product framing and, honestly, an E&O policy.

14. Structured verdict

Score:                  66/100
Verdict:                VALIDATE
Confidence:             Medium
Best-fit builder:       Technical solo founder willing to become the Pub 1281 expert,
                        ideally with an existing line into bookkeeping or PM back-office firms
Time to revenue:        6–10 weeks if built into the Sept/Oct notice window; otherwise April
Capital to launch:      $5–8K (mail integration, vision API, domain review, E&O)
Top 3 assumptions to validate first:
  1. Payers in the 20–500 band actually misclassify first-vs-second notices — free
     classification of 25 firms' three-year extracts, measure the error rate
  2. A twice-yearly pain sustains an annual subscription — price-test annual vs monthly
     across the first 40 signups, watch November cancellations
  3. The OBBBA $2,000 threshold doesn't hollow out the segment — ask those same firms what
     share of their 1099 volume sat in the $600–$2,000 band
Kill criteria:
  - Abandon if fewer than 10% of payees in the free classification pilot were misclassified —
    the state-tracking pain is imagined and the incumbents' pre-file products are sufficient
  - Abandon if TIN Comply (or any e-filer) ships a persistent, self-serve 1st/2nd tracker
    below $50/mo before v1 launches
  - Abandon if more than half of the first 40 customers churn between November and March —
    the seasonality objection is fatal and the model has to become transactional

15. Next step — 1-week validation sprint

• Day 1–2: Encode the Pub 1281 classifier — two-in-three-year rule, Q17 same-year disregard, the three clocks — and validate the logic against the publication line by line. Build nothing else. This is the product’s only real asset and it either survives contact with the text or it doesn’t.
• Day 3–4: Recruit 25 bookkeeping, property-management back-office and small-CPA firms. Offer: send your 2024, 2025 and 2026 CP2100/CP2100A extracts, get a free classified list back within 24 hours. Ask each one, before showing results, how they classified those payees at the time.
• Day 5: Compare their historical classification against the engine’s. Go if ≥20% of payees were misclassified or unclassified, and ≥8 of the 25 firms say they’d pay $99/mo to never do it by hand again. Below 10% misclassification, kill it — the incumbents’ pre-file TIN matching is doing its job and this is a solution hunting a problem that the IRS’s own notice already solves well enough.

The falsifiable bit is the misclassification rate. Everything else in this proposal is downstream of whether these firms are actually getting it wrong.