ALLSTATE

The Allstate Corporation  |  NYSE: ALL  |  Northbrook, Illinois

STRATEGIC MARKETING REVIEW: TEN QUESTIONS, TWO HUNDRED AND TWENTY-SIX POINTS

Built from every document in this process and from the one hundred marketing suggestions produced by the reconvened focus group.

NO POSITION / NO COMPENSATION / NO MNPI

ANALYSES BASED ON PUBLIC INFORMATION. NOTHING HERE IS INVESTMENT ADVICE. 

How to Read This Document

Ten headings follow. Under the first there are one hundred numbered points, organized into eight subsections. Under each of the nine remaining headings there are between twelve and eighteen points. In every point, one to four consecutive words are set in bold, underlined and one point larger than the surrounding text. Those words are the main idea of that point, and a reader in a hurry can read only those words and still come away with the argument.

Where a recommendation takes real time to carry out, the approximate elapsed time appears in parentheses at the end of the point. Those are calendar estimates for an organization of this size and regulatory complexity, not effort estimates, and they assume the compliance review cycle described under Hurdles is not first repaired.

Every underlying fact comes from public disclosure: second-quarter 2026 results, the Form 10-Q, the proxy, trade and financial press through August 2026, and the public web presence. Nothing here is investment advice, and no valuation, price target or transaction recommendation appears anywhere in it.

1.  Marketing and Website Ideas

One hundred points, in eight subsections. Numbering is continuous across the whole section.

A.1  Answer Engines and Machine Visibility

1. Appoint a single named owner of answer-engine visibility at director level or above, reporting into marketing rather than into information technology, with a budget line, three to five headcount, and a standing seat in the marketing leadership meeting. Nobody owns this today, and it was the most unanimous finding across every focus group convened in this engagement. (Hiring and onboarding: 90 to 120 days.)

2. Build and publish a free state-by-state coverage reference covering what each state requires, what coverage is typical, what a claim actually looks like there, and what the filing and complaint process is. Build it to be cited, quoted and linked rather than to convert, because the citation is the conversion in an answer-engine world. (First ten states: 60 days. All fifty: 6 to 9 months.)

3. Publish machine-readable coverage definitions, eligibility rules and discount criteria in structured form so an assistant quotes the carrier directly instead of quoting an aggregator’s summary of the carrier. This is the single cheapest structural defense available against disintermediation. (Engineering and legal: 4 to 6 months.)

4. Run a monthly citation-share audit across the major AI assistants on the top one hundred real buyer questions, and report the result beside share of voice in the same deck so the two numbers are argued about together. (Standing up the process: 30 days. Then ongoing, roughly one week per month.)

5. Publish an honest comparison page that states plainly where Allstate wins and where it loses against each major competitor. It will outrank and outlive every page that does not, and specific, checkable, falsifiable content is exactly what retrieval systems preferentially cite. (Draft, legal review, publication: 90 days.)

6. Consolidate the templated duplicate pages across state and city inventory. Several hundred genuinely differentiated pages will outperform several thousand near-identical ones, and mass templating now reads as machine-generated to classifiers that did not exist when the pages were built. (Audit: 45 days. Consolidation: 6 months.)

7. Rewrite the top two hundred commercial pages so the answer appears first, inside the first hundred and fifty words, above the hero image and before the compliance preamble. The current pattern assumes a visitor who already decided to visit. (200 pages at current review speed: 4 to 6 months.)

8. Instrument every commercially important page with current structured data – product, FAQ, organization, and local business markup on every one of the agent pages. Consistency matters more than sophistication here. (Technical implementation: 60 to 90 days.)

9. Treat Bing as strategic rather than residual, with a properly configured Webmaster Tools presence and a named person watching it, because of its role in generating AI answers rather than because of its search share. (Initial configuration and audit: 3 weeks.)

10. Build topical authority clusters around claims, coverage and pricing questions, with tight internal linking and specific descriptive anchor text instead of the generic anchors currently in use. (First three clusters: 90 days.)

11. Take editorial ownership of the negative search real estate by publishing the company’s own honest answers to the complaint and dispute queries that presently outrank company content. This is a marketing problem being treated as a legal one. (First tranche: 60 days.)

12. Strengthen the Wikipedia article indirectly by generating better third-party sourcing on product and distribution structure. Never edit it directly and put that prohibition in written policy so a well-meaning employee does not create a story. (Sourcing program: 6 months, ongoing.)

13. Publish a public coverage-terms API so an autonomous agent can verify Allstate’s product without scraping a marketing page. The cheapest carrier to verify wins the machine’s recommendation before the cheapest carrier on price does. (Specification through launch: 9 to 12 months.)

A.2  Paid Media, Measurement and Incrementality

14. Institute a permanent geographic holdout program across five to eight percent of designated market areas, rotated quarterly, against the roughly two billion dollar spend base. Sized correctly the revenue sacrifice is a rounding error and the learning is not. (Design and launch: 90 days. Then permanent.)

15. Separate measurement from media buying on the organization chart. The team that buys must never be the team that grades, and no amount of analytical sophistication substitutes for that one structural fact. (Reorganization: 30 to 60 days.)

16. Publish marginal not average cost per incremental policy by channel and by state on a quarterly internal cadence, and let the board interrogate it. Average cost per acquisition is the number that hides the answer. (First quarterly publication: 120 days.)

17. Incrementality-test branded search first, because it is reliably the largest wasteful line in a large advertiser’s budget and the easiest line to test cleanly. (Test design through readout: 90 days.)

18. Replace advertising spend as the reported metric with cost per retained customer-year by channel. A shopper acquired at the bottom of the funnel is by construction a shopper who shops again at renewal. (Metric definition and instrumentation: 4 months.)

19. Grade lower-funnel acquisition on three-year retention rather than on bind rate, or the funnel will efficiently and measurably buy the least loyal customers in the market. (Cohort tracking build: 60 days. Meaningful readout: 3 years.)

20. Run a media mix model and an experimentation program in parallel and treat disagreement between them as a research finding rather than as an embarrassment for whoever built the model. (Parallel operation: ongoing, first reconciliation at 6 months.)

21. Set an explicit acquisition-to-retention spend ratio and defend it at board level annually. Acquisition dollars poured into a leaky retention bucket are this category’s structural error. (Analysis and board approval: one planning cycle.)

22. Cap the single-property budget share any one creative franchise, including Mayhem, may consume. A character running since 2010 accretes budget by habit rather than by evidence. (Policy decision: immediate. Enforcement: next planning cycle.)

23. Stop advertising price where uncompetitive. Net-neutral national rate means price went up in roughly sixteen states and down in roughly twenty-three, and price messaging in the wrong half buys quotes that never convert and teaches those markets to think of Allstate as expensive. (Creative and media reallocation: 60 days.)

24. Move a defined share of national brand television into state-level and agent-level media in the states where the growth target and the catastrophe appetite actually agree. (Reallocation and test: one media planning cycle, roughly 6 months.)

25. Build a catastrophe-moment readiness buy with pre-cleared creative, pre-negotiated inventory and a standing trigger, rather than scrambling after landfall when everything costs triple. (Program build: 90 days.)

26. Create a pre-cleared language library with legal, compliance and actuarial in the room from the first meeting, so campaigns can move at auction speed inside approved boundaries instead of at legal-review speed. (Build: 4 to 6 months. Pays for itself immediately after.)

A.3  Website, Local Presence and Conversion

27. Rebuild the agent locator as conversion, not as a directory: real photographs, real reviews, real availability, real local content, and one unmistakable call to action per page. This is the largest single missed opportunity on the site. (Design through launch: 6 to 9 months.)

28. Standardize every Google Business Profile across the agency footprint – categories, hours, photography, review response and posting cadence – managed centrally with local input. Unglamorous, high return, and entirely within the company’s control. (First pass across all listings: 6 months.)

29. Build genuinely differentiated local pages per agency, with local claim examples, local weather risk, local staff names and local phone numbers, rather than templated fills with a variable swapped in. (Platform build: 4 months. Rollout: 12 months.)

30. Replace stock lifestyle photography with photographs of actual agents, actual adjusters and actual customers on every conversion surface. Faces convert; stock faces do not. (Shoot and swap on top pages: 90 days.)

31. Strip third-party tags off the quote path – marketing tags, personalization scripts, chat widgets – until it loads instantly rather than acceptably. Every one of them is a tax paid on the most valuable page in the company. (Audit and removal: 45 days.)

32. Flatten the site hierarchy so the answer to any common question sits two clicks from a search landing. Genuinely useful content currently lives several levels below pages that exist for organizational reasons. (Information architecture project: 6 months.)

33. Simplify top-level navigation by resolving the internal politics of who owns menu real estate. The customer is presently paying for the organization chart. (Decision: 30 days. Implementation: 90 days.)

34. Put proof on decision pages – testimonials, claim outcomes, resolution times – not only on brand pages where nobody is deciding anything. (Content and placement: 90 days.)

35. Link outward to authorities: state departments of insurance, safety organizations, mitigation standards bodies. Outbound links to credible third parties are a trust signal and an authority signal, and the site almost never does it. (Editorial policy and first pass: 45 days.)

36. Make human contact as prominent as the quote button. A buyer who cannot reach a person before purchase is a buyer who does not trust the purchase enough to complete it. (Design and rollout: 60 days.)

37. Test a radically simplified quote landing against the current homepage-led path – one purpose, no navigation, no hero video, no upsell. (Build and test: 60 days.)

38. Build agency-level review generation and response, since local reviews are simultaneously a local ranking factor, a conversion asset and a source AI assistants cite. (Platform and training: 4 months.)

39. Make the app telematics experience a marketing asset with its own acquisition funnel rather than a post-purchase discount mechanism nobody sees until after they buy. (Repositioning and funnel build: 6 months.)

A.4  Proof, Claims and the Attribute Worth Owning

40. Publish an annual claims outcomes report: days from first notice of loss to payment, percentage of claims paid without dispute, percentage of estimates revised upward on reinspection. This is the most durable marketing asset available in a commodity category. (First edition: 9 to 12 months.)

41. Release the methodology before the number, publish on the same date every year, and never move the date. A standard everyone else has to be measured against is worth more than any campaign. (Methodology publication: 4 months ahead of the first result.)

42. Pick one attribute to own and resource it obsessively. The only candidate competitors cannot buy their way into is resolution speed and proof of payment, supported by an owned national claims organization. (Decision: one quarter. Consequences: permanent.)

43. Stop claiming five attributes at once – protection, reliability, price, service and technology. A company owns one attribute in the public mind, and claiming five is the reliable way to own none. (Messaging architecture rewrite: 90 days.)

44. Put real adjusters on camera. Documentary evidence that the promise gets kept is the most persuasive advertising available in this category and no competitor is running it. (Production of first campaign: 4 to 6 months.)

45. Show the after – the roof rebuilt, the car returned, the check cleared, the family back in the house – with consent and legal built in from the first frame rather than bolted on. (Documentary program: 6 months to first release.)

46. Build a claims-stories content franchise across video, podcast and written form that includes outcomes that were not perfect, because uniformly perfect testimonials read as fiction and are treated as such. (Franchise development: 6 months.)

47. Create a claims-quality function independent of the claims organization. A quality metric owned by the group being measured is an unfalsifiable number and everyone in the building knows it. (Organizational build: 6 months.)

48. Have senior executives read raw claim files – ten randomly selected disputed or denied files a week, personally, not summaries – then publish that the practice exists and publish the count. It costs approximately nothing and cannot be gamed by the people being sampled. (Immediate. One hour a week, permanently.)

49. Reconcile the legal voice and friendly voice. The company writes like a legal department and sells like a friend, and the gap between the two is exactly where the customer decides whether to believe any of it. (Voice guidelines and rewrite: 6 months.)

50. Make the slogan operationally true. A covenant phrase running since 1950 either gets backed with published evidence or it slowly becomes a punchline, and the second process is already underway in public. (Ongoing, anchored to the claims report.)

51. Publish a plain-language rate explanation before a journalist or a regulator publishes the version that ranks instead. Explaining your own pricing is cheaper than being explained. (Draft through publication: 90 days.)

52. Build a first-notice-of-loss experience good enough to advertise on its own, then advertise it with a specific published number rather than an adjective. (Operational improvement: 12 months. Campaign: 3 months after.)

A.5  Distribution, Agents and Channel Strategy

53. Name the three-channel position publicly, in those words, as the corporate strategy: exclusive agents, independent agents, direct, plus workplace and retail. No other national carrier can say it and the company has never said it out loud. (Positioning and announcement: one quarter.)

54. Finish the Custom360 state rollout and the Affordable, Simple, Connected rollout, and treat the remaining states as a marketing deadline rather than an engineering backlog item. The product moved from 36 states to 41 in two quarters; the last nine are the binding constraint on the independent channel. (Remaining states: 9 to 15 months.)

55. Publish appetite in machine-readable form so an independent agent’s management system can tell in one second whether a risk fits. Appetite clarity is a top purchase criterion in that channel and it is currently a PDF. (Specification and publication: 6 months.)

56. Market to independent agents deliberately: quoting speed, appetite clarity, a claims desk that answers on the first call, and co-marketing support. Brand advertising is nearly irrelevant to their placement decision and they will say so if asked. (Channel marketing program: 6 months.)

57. Build the agent local marketing system the field has been asking for: pre-approved creative, same-day co-op approval, automated local pages, reviews management, and a budget the agent can actually spend inside the buying cycle. (Platform build and rollout: 9 to 12 months.)

58. Resolve channel conflict in writing. Publish internally which channel owns which household and price so the channels stop bidding against each other for the same customer at different prices. (Policy work: 6 months.)

59. Tell the exclusive force what the target number is. The force has gone from roughly ten thousand agencies to roughly six thousand and agents are inferring the plan from a production statistic, which is the most demoralizing possible form of communication. (Communication plan: 60 days.)

60. Build an agent creator program for the highest-performing agencies: training, equipment, pre-approved content, compliance support, and distribution. Six thousand agencies are six thousand local publishers. (Pilot: 4 months. Scale: 12 months.)

61. Run referrals through agents rather than through the brand, because the trust that produces a referral is local and always has been. (Program design and launch: 6 months.)

62. Convert Protection Plans into acquisition. A customer with device coverage is an existing, trusted, transactional relationship that costs nothing to acquire and can be walked to renters, then auto, then home. It is currently a segment disclosure. (Cross-sell program: 9 months.)

63. Sell renters through retail partners who already sell protection plans at the same checkout. The shelf space is already bought and it is not stocked. (Partner negotiation and pilot: 9 to 12 months.)

64. Attack embedded vehicle point-of-sale distribution through dealerships, lenders and manufacturers before captive alternatives close the channel. Root built a real auto book inside somebody else’s checkout, at a fraction of this company’s scale. (Channel build: 12 to 18 months.)

65. Build a landlord and small-commercial cross-sell motion aimed at existing homeowners customers, who have already trusted the carrier with the larger asset. (Product and go-to-market: 12 months.)

A.6  Retention, Lifecycle and Existing Customers

66. Tell customers their price fell. Nearly eight million people received an average seventeen percent reduction under the SAVE program and most of them do not know it happened, which is a rare and expensive kind of marketing failure: a good deed performed silently. (Campaign: 60 days.)

67. Replace the annual renewal notice as the primary relationship touch. It is the weakest contact cadence available in any consumer category and it is currently the whole relationship for most households. (Lifecycle program design: 6 months.)

68. Build proactive rate-review outreach that contacts customers before they shop, and grade it against the cost of reacquiring them afterward. (Program build: 4 months.)

69. Thank long-tenured customers explicitly and materially before a competitor’s quote reaches their phone. A forty-year claim-free policyholder has never once been thanked. (Program: 90 days.)

70. Build life-event trigger marketing – new home, new driver, new baby, new job, moving state – because the life event and the purchase decision are the same moment and the company already knows when most of them happen. (Data and trigger build: 6 to 9 months.)

71. Reposition telematics from coupon to product: a family safety and driving-improvement product that happens to lower the premium. The parent of two teen drivers buys the first framing and ignores the second. (Repositioning: 6 months.)

72. Run a teen driver safety program as a marketing property with genuinely useful content for parents, because that household buys three products at once and stays for a decade. (Program build: 6 months.)

73. Build a ninety-day onboarding sequence that explains the policy in plain language, because a customer who understands the coverage disputes the claim less and renews more. (Build: 4 months.)

74. Give customers a reason to open the app between renewals: maintenance reminders, severe weather alerts, mitigation tips, roadside status, claim tracking. (Feature work: 6 months.)

75. Engineer one delightful moment yearly per relationship. In a hundred million relationships there is currently only the absence of catastrophe, which is not an experience anyone tells a friend about. (Design and pilot: 6 months.)

76. Build a bereavement and estate path handled by a named person rather than by a form. It is the single highest word-of-mouth moment in the business and it is currently administrative. (Process redesign: 4 months.)

77. Publish retention by channel internally and treat a retention point as worth more than an equivalent acquisition point, because it costs a fraction as much to buy. (Reporting build: 90 days.)

A.7  Creative, Brand and Content

78. Keep Mayhem for reach and build a second serious creative track underneath him that never attempts humor and carries price, claims speed, bundling and telematics. One character cannot carry the whole funnel and has been asked to for sixteen years. (Development and production: 6 to 9 months.)

79. Set an explicit creative review date for every long-running property rather than letting it drift on the marketing organization’s own nostalgia. (Policy: immediate.)

80. Lead with home not auto. Home carries higher premium, longer tenure, less shopping behavior, and drags the auto policy behind it. Homeowners premium grew 8.3 percent at a 94.6 combined ratio in a year peers were exiting the line, and it is marketed as a bundling discount. (Repositioning: 6 months.)

81. Treat YouTube as a search engine rather than as a television buy: an indexed library of short answer videos matched to the questions people actually type – deductibles, total loss, rental coverage, roof claims. (First hundred videos: 6 months.)

82. Develop post-catastrophe creative for the days when category attention is highest and competitive noise is lowest, and have it cleared and sitting on the shelf. (Production and clearance: 4 months.)

83. Rebuild direct mail as price delivery – state-accurate, filed, specific to the household – because it is one of the few remaining channels where a carrier can put a real number in front of a real person, and deliverability now favors mail over email. (Program rebuild: 6 months.)

84. Build a modest trade print and event presence aimed at independent agents and small commercial prospects, where credibility still transfers through print in a way it does not online. (Program: 6 months.)

85. Sponsor life-event podcasts aimed at new homeowners, new parents and new drivers, where the purchase decision and the content consumption coincide. (Buy and creative: 3 months.)

86. Establish a monitored honest community presence in the insurance subreddits and local homeowner groups, accepting roughly a year of hostility as the entry cost, because absence is a decision to let the angriest customer write the answer. (Hire, policy, launch: 4 months. Credibility: 12 to 18 months.)

87. Use X for catastrophe utility rather than for brand posting: fast, plain, useful claims guidance at the exact hour people are searching for it and no competitor is being useful. (Playbook and staffing: 60 days.)

88. Publish a recurring original data study built from claims and telematics data, aggregated and privacy-safe, designed to be picked up by journalists and cited by models. (First study: 6 months. Then quarterly.)

89. Build a Spanish-language content program of equal quality rather than translated afterthoughts, in the states where the growth math and the demographics both work. (Program build: 9 months.)

A.8  Organization, Geography and Governance of Marketing

90. Run marketing on the pricing grain. The company already prices by state, files by state and is regulated by state; marketing is the one function that defaults to national and it is the only one that can afford not to. (Budget and org redesign: one planning cycle.)

91. Set state growth targets against capacity so acquisition does not buy homeowners policies that underwriting must immediately shed at full acquisition and reputational cost. (Planning integration: one cycle.)

92. Treat regulatory relationships as marketing with a named owner, especially in the states where the company has previously restricted or withdrawn. Goodwill is not automatically available on the way back in. (Ongoing, immediate start.)

93. Fund employer brand as marketing, with a line item and a target, because a company running an AI transformation must be a credible employer for AI talent and that is a marketing problem before it is a recruiting one. (Program: 6 months.)

94. Publish technical work externally – engineering, data science, catastrophe modeling – as a recruiting and authority instrument. It is the cheapest available fix for an invisible technical employer brand. (First publications: 4 months.)

95. Market the Belfast and Bengaluru centers externally. Roughly four thousand technologists in two genuine centers of excellence, and almost nobody outside the company knows they exist. (Content and campaign: 4 months.)

96. Protect service-role stability and training budgets, because frontline attrition silently determines whether the promise the advertising makes is actually delivered, and it is chronically the first cut and the last measured. (Budget decision: immediate.)

97. Establish a quarterly outside-in review deliberately staffed with skeptics, and document what it says even when it is unflattering. (Standing program: first session within 90 days.)

98. Build a marketing-to-underwriting feedback loop so campaign targeting is informed by which risks the company actually wants, rather than by which audiences are cheapest to reach. (Process build: 6 months.)

99. Create a single customer data view across auto, home, protection plans, identity protection and roadside, because the cross-sell recommendations in this document are impossible without it. (Data engineering: 12 to 18 months.)

100. Give the state general managers budget authority over a defined slice of media, with accountability for the result. They already own the rate filing; they should own the message. (Org change: one planning cycle.)

2.  Hurdles That May Be in Your Way

Fifteen points. These are the reasons the recommendations in Section 1 fail when they fail, and most of them are internal rather than competitive.

1. The compliance review cycle is the binding constraint on almost every recommendation in this document. A state-page rewrite sitting in review for nine months is not a legal problem, it is a throughput problem, and it will silently defeat any content strategy that does not design around it. (Fixing throughput: 6 to 12 months.)

2. Fifty separate regulators mean a national savings claim invites fifty separate readings, and the most common self-inflicted wound in this category is a marketing team writing a claim the actuarial and filing teams cannot substantiate everywhere it runs.

3. The margin environment is borrowed. Roughly $1.5 billion of prior-year auto reserve releases in the first half of 2026 flatters the auto combined ratio by several points. Every recommendation here assumes marketing is funded out of underwriting profit, and in a worse loss year marketing is cut first and the proof program dies first of all.

4. Channel conflict is unresolved and currently managed by not discussing it. Better marketing into an unresolved conflict simply makes the collision more expensive, because two channels bid against each other for the same household at different prices.

5. The exclusive force is demoralized. A reduction from roughly ten thousand agencies to roughly six thousand, communicated as attrition rather than as strategy, produces a distribution channel that is slow and expensive to re-energize even when the strategy is correct.

6. Catastrophe geography constrains growth. Not writing new homeowners business in Florida and writing California only through excess and surplus lines means national campaigns generate demand the company cannot serve, at full cost.

7. The privacy litigation is live. The Texas action against the company and Arity was the first enforcement ever brought under a state comprehensive privacy law, and a consolidated private case survived dismissal across twenty states’ claims. Any targeting program built on driving data has to assume the rules will move.

8. Claims scores lag the promise. J.D. Power scored the company 665 on 2025 property claims against a 682 average and 693 on auto against Nationwide’s 729. Advertising resolution speed before fixing variance would be worse than saying nothing.

9. Organizational politics own navigation. The site’s complexity is not a design failure, it is an org chart rendered in HTML, and no information architecture project survives without an executive willing to take menu real estate away from a business unit.

10. Answer-engine behavior is volatile. Retrieval and citation behavior changes faster than an enterprise content calendar. The ownership recommendation is durable; any specific tactic may be stale within a year, and the program has to be built to be rewritten. (Re-baseline quarterly.)

11. The engineering talent pipeline is the constraint on the AI plan. A Northbrook headquarters competing for machine-learning engineers against technology employers needs exceptional remote flexibility, exceptional problem interest or exceptional pay, and preferably two of the three.

12. Measurement threatens careers. A permanent holdout program will eventually produce a number that embarrasses somebody senior, and the program dies at that moment unless the reporting line was separated before the first result.

13. The category advertising arms race means a three-fold spend increase raises lead costs for every carrier including the one doing it, and GEICO has turned its spending back on. Winning by outspending is the one strategy structurally unavailable to the fourth-largest carrier.

14. Succession is unaddressed publicly. A 68-year-old chair and chief executive with nineteen years in the seat, alongside a finance chair who arrived in August 2026, is a question the market will eventually ask loudly, and unanswered leadership questions freeze long-horizon marketing investment.

15. The data is not unified. Cross-sell from protection plans to renters to auto to home is the cheapest acquisition available and it is currently impossible because the customer record does not exist in one place. (12 to 18 months to fix.)

3.  Things You Could Be Doing Right Now

Fifteen points. Each of these can be started this quarter without a new budget line, a new platform, or a reorganization.

1. Name the answer-engine owner this week. It does not require a new headcount to start – it requires a named person with the authority to convene search, content, product and legal. (One week to name. 90 days to staff properly.)

2. Run the first citation audit on the top hundred buyer questions and set a baseline nobody has ever measured. Any competent analyst can do this manually in two weeks with no tooling budget. (2 weeks.)

3. Tell eight million customers their premium went down. The SAVE reductions already happened; the marketing did not. This is free goodwill sitting on the table. (60 days.)

4. Audit the Google Business Profiles and fix categories, hours and photos on the worst thousand listings first. No new technology, no new budget, immediate local ranking effect. (90 days for the first tranche.)

5. Strip the quote-path tags. Remove every non-essential script from the highest-value page in the company and measure the conversion change. (45 days.)

6. Publish the comparison page that says where Allstate wins and where it loses. It is the single most differentiating page available and it costs a writer, a lawyer and a month of arguing. (90 days.)

7. Start the executive claim reading immediately. Ten random disputed files a week, read personally, no summaries. It costs an hour and it cannot be gamed. (This week.)

8. Set the holdout design in motion. Even a single-market pilot on branded search produces a defensible number within a quarter and builds the political case for the permanent program. (90 days.)

9. Separate measurement from buying on the org chart before the first uncomfortable result arrives, because doing it afterward looks like retaliation. (30 to 60 days.)

10. Ship fifty answer videos against the fifty highest-volume coverage questions, filmed simply with real employees. This does not require a production company. (90 days.)

11. Photograph real agents in the top hundred markets and swap them into the local pages. (90 days.)

12. Publish the methodology for the future claims outcomes report now, before the first number, so the standard is on record and the company is committed. (4 months.)

13. Open the pre-cleared library with the twenty most-used claims and phrases, so the content team stops re-litigating the same sentences every quarter. (90 days.)

14. Brief the field on what the exclusive agency target number is, whatever it is. Silence is producing a worse story than any real number would. (60 days.)

15. Start the X catastrophe playbook before hurricane season peaks, with pre-cleared claims guidance and a named person on call. (30 days.)

4.  Things Worth Avoiding

Fourteen points. Each of these is a plausible, well-intentioned move that would make the position worse.

1. Do not advertise resolution speed early. Claiming the attribute before the variance is fixed converts an operational problem into a credibility problem, and the J.D. Power gap is public.

2. Do not chase the spending war. GEICO has turned its advertising back on and Progressive compounds on price. The fourth-largest carrier cannot win an escalation whose main effect is raising everyone’s lead costs including its own.

3. Do not plan against 83.3. The recorded auto combined ratio is flattered by non-repeating reserve releases; the adjusted underlying figure near 90.0 is the number a durable plan is built on.

4. Do not let measurement report to buying. Every analytical improvement is cosmetic if the team producing the grade reports to the team being graded.

5. Do not build targeting on fragile data. A program that depends on driving data being usable a particular way is a program with active litigation attached to its foundation.

6. Do not edit the Wikipedia article directly, or permit any employee or agency to. The story about the edit is always worse than the article.

7. Do not generate more templated pages. More near-duplicate state and city pages dilute rather than compound, and classifiers now read the pattern as machine-generated.

8. Do not market price into raised states. Net-neutral national rate hides an increase in roughly sixteen states, and price messaging there buys unconverting quotes and a reputation for being expensive.

9. Do not promise what claims cannot deliver in the worst week of the worst catastrophe year. Every proof-based recommendation here is an operational obligation before it is a campaign.

10. Do not position AI as headcount reduction, internally or externally. Frame it as variance reduction, because variance is the actual defect and headcount framing poisons the frontline cooperation the program requires.

11. Do not enter communities defensively. A carrier that appears in a subreddit only to do damage control confirms the community’s existing view and wastes the entry cost.

12. Do not let the character carry everything. Mayhem has been running since 2010 and dramatizes a risk every carrier covers identically, which makes it excellent advertising for the category rather than for the company.

13. Do not optimize what measures easily. Lower-funnel spend measures well precisely because those buyers had already decided to shop, which is the same reason its incrementality is weakest.

14. Do not claim quantum pricing benefits the program cannot yet support. Overclaiming an early research effort is the cheapest way to lose credibility with exactly the technical audience the company is trying to recruit.

5.  How Specialized Venture Capital Might View Your Plans

Thirteen points. This is how a specialist insurtech or fintech investor would read the same public facts, which is a useful lens precisely because it is not the lens a personal-lines carrier uses on itself. Nothing here is a valuation or a recommendation about any security.

1. A specialized insurtech investor would immediately note that the seams are already proven. Nobody in the venture cohort replaced an incumbent; every survivor found a seam – renters and pet, builder-distributed home, auto inside a car purchase, coastal homeowners at a 49 percent operating margin. Every one of those is something this company could do at a hundred times the scale.

2. They would value the embedded book most highly. Protection Services carrying roughly 177 million policies and $3.4 billion of revenue, fee-based, short-tail and uncorrelated to catastrophes, is the asset a venture investor would call a platform and the company calls a segment.

3. They would treat distribution as the moat, not underwriting. Three channels at once is a position no other national carrier holds, and in venture language that is owned demand rather than rented demand.

4. They would read the advertising line as rent. Two billion dollars a year buying attention with no published marginal return curve is, to a growth investor, a customer acquisition cost problem dressed as a brand budget.

5. They would ask about payback and cohort retention within the first ten minutes and would be genuinely surprised that the answer is reported as a single spend line rather than as cost per retained customer-year by channel.

6. They would price the answer-engine risk as an existential distribution question rather than as a marketing subtopic, because they watched the same disintermediation happen to categories that thought they owned their customer.

7. They would find the telematics asset underexploited. More than two trillion miles of driving data positioned as a discount mechanism would strike a data-focused investor as a mispriced asset inside a well-run company.

8. They would discount for regulatory and consent risk on that same asset, given live litigation and the first-ever enforcement action under a state comprehensive privacy law.

9. They would like the expense ratio compounding loop far more than the advertising. A 6.6-point reduction in the adjusted expense ratio since 2018 funding a 17 percent price cut for nearly eight million customers is the kind of superlinear mechanism venture investors actually reward.

10. They would be skeptical of ALLIE without an evaluation set. A named platform with eight components and no public statement of what it is measured against reads, to a technical investor, as a program rather than a product.

11. They would flag growth rate as the verdict. Auto policies growing 2.8 percent while a direct competitor compounds above 11 percent is, in their framework, the whole story, and a $4 billion buyback authorization is management agreeing with them.

12. They would see the claims variance as the opportunity, not the liability – a measurable defect in a business with the data and scale to fix it is precisely the thesis a specialist fund underwrites.

13. They would ask who owns the customer question in three years, and would judge every marketing recommendation here by whether it moves the company toward being the answer rather than toward buying placement next to it.

6.  Competitor Analysis: What Is Working, What Is Not, and What It Means for You

Eighteen points in three subsections.

What is working for the competition

1. Progressive prices rather than persuades. It passed State Farm on a trailing-twelve-month basis through March 2026 – the first time State Farm has not been number one since 1942 – while growing private auto premium 11.6 percent and posting a 2025 combined ratio of 87.4. Price compounds; advertising does not.

2. Progressive’s segmentation is the engine, not its brand. Better segmentation lets a carrier take the good risks and leave the rest, and the brand character is a delivery vehicle for a price message rather than the message itself.

3. GEICO owns a single attribute – cheap and fast – with the most fully understood advertising economics in the category, and it has now turned spending back on after several years of deliberate retrenchment.

4. State Farm owns the neighbor, with the deepest local agent footprint in the category and a brand built on proximity rather than price, which is the one thing a national campaign cannot buy.

5. Independent agents keep gaining, placing 39.5 percent of all personal lines premium in 2025, a fourth consecutive annual gain, which validates the National General strategy the company has been quiet about.

6. Insurtechs found real seams – renters, pet, builder-distributed home, embedded auto, and coastal homeowners underwriting at a 49 percent operating margin on the risk the nationals are exiting.

What is not working for them

7. State Farm’s flat premium is not stability, it is erosion measured slowly, and its abrupt agent-contract overhaul is taking reputational damage of exactly the kind this company avoided by moving gradually.

8. GEICO remains weak in home and in bundling, which is precisely the ground where a multi-line structure has an argument, and homeowners drags the auto policy behind it.

9. Progressive’s home exposure has been managed by restricting it, which is a rational answer that cedes the household – and it has now named the bundled home-and-auto customer its 2026 priority, which is an admission of the gap.

10. The insurtech cohort did not scale. Public property-casualty insurtechs sit broadly 70 to 90 percent below their listing-era peaks. They proved the openings exist; they could not fund closing them.

11. Every national carrier retreated from catastrophe-exposed homeowners at once. Unanimous avoidance is exactly the condition under which returns go unclaimed, and one focused competitor is claiming them.

12. Nobody publishes claims outcomes. The entire category competes on a promise and not one competitor has published a verifiable number about whether the promise is kept.

What it means for you

13. You cannot win the price war against two larger carriers with better direct economics, so competing on their category definition caps you structurally at third or fourth.

14. You can win the proof war because you own a national claims organization at scale and nobody has claimed the ground. It is available, it is defensible, and it is expensive to copy.

15. Your loss ratio is the weapon nobody markets. A 55.64 direct loss ratio, best among the seven largest carriers and better than Progressive’s 59.07 and State Farm’s 65.44, achieved while cutting price for 7.88 million customers, is a fact that belongs in the market and currently lives in an investor deck.

16. Your ten-year record is the proof: an auto combined ratio averaging 95.2 over a decade in which the industry made no underwriting profit at all, and roughly ten points better than the industry in homeowners.

17. Home is the wedge against GEICO and Progressive both, and it is currently marketed as a discount attached to auto rather than as the lead product it should be.

18. The seams are yours to close. Everything the insurtechs proved – embedded, single-segment depth, seconds-not-days resolution – you can do at a hundred times their scale, and in protection plans you already do.

7.  Trends and Direction for Your Industry

Fourteen points. Direction of travel, not prediction. No probabilities are assigned and no financial outcomes are projected.

1. The buying moment is moving to a machine. When a meaningful share of shopping is mediated by an assistant, the carrier that is cheapest to verify wins the recommendation before the carrier that is cheapest on price does.

2. Claims resolution is collapsing toward seconds for simple losses, and roughly two thirds of insurers say they plan scaled AI claims agents this year. Resolution speed is a claim with a shelf life.

3. The market is softening. Fitch expects continued softening through 2026, personal auto premium growth has slowed to roughly 3.6 percent, and softening markets reward share discipline over rate.

4. Reinsurance has cheapened. January 2026 property catastrophe renewals fell as much as 20 percent for US loss-free business, and the largest domestic buyer of that cover benefits directly and quietly.

5. Severe convective storm frequency, not hurricanes, has changed the economics of homeowners marketing, and it lands in places the catastrophe conversation historically ignored.

6. Homeowners availability will be political. The Florida and California withdrawals are individually rational and collectively a market failure that will be resolved legislatively rather than actuarially, which makes regulatory affairs a marketing function.

7. Embedded distribution keeps advancing toward the point of sale – the dealership, the lender, the manufacturer, the retailer’s checkout – and captive alternatives are being built now.

8. Independent agents keep taking share, four consecutive years, which makes a competitive middle-market product in that channel more valuable each year it is delayed.

9. Telematics is converging across carriers, which means the differentiation moves from the score to the experience built around it, and the carrier pricing off the vehicle’s own sensors rather than a phone eventually wins the segment.

10. Privacy enforcement is arriving state by state, and the first comprehensive-law enforcement action in the country landed on this industry’s data practices rather than on a technology company’s.

11. Consolidation will be technology-driven rather than underwriting-scale-driven, which changes who the plausible acquirers and targets are and what marketing has to explain afterward.

12. Brand-age dynamics are intensifying. As substantive product differences disappear under regulation and shared vendors, spending escalates on brand and quality becomes a threshold rather than a differentiator – until the promise visibly breaks, at which point it is a write-down of the only remaining asset.

13. Trust is migrating local even as buying migrates digital, which is why six thousand local businesses with addresses, phone numbers and reviews is a more durable asset in 2026 than it looked in 2016.

14. Quantum is a recruiting story for now. Pricing precision is a real long-horizon application, and the honest near-term value of a small expanding team is credibility with technical talent.

8.  Where AI Might Fit Into Your Business

Fourteen points, in rough order of how much value they would create against how much risk they carry.

1. Variance reduction in claims is the highest-value application in the building. The public gap is not capability, it is consistency – reviewers say the experience depends on which adjuster you get – and inconsistency is exactly what a well-governed system fixes.

2. Keep the model on drafting and retrieval while a human owns every adverse decision, because a claims model operates on an adversarial input distribution and will be audited by regulators and litigated by plaintiffs.

3. Publish the evaluation set for anything that touches a denial. Nobody outside the company knows what ALLIE is measured against, and voluntary disclosure of the test is the cheapest credibility available.

4. Point the assistant outward. Build the version that answers the customer’s real question – what should I do about this risk – rather than the version that optimizes bind rate, because within a few years the second will be competing against a general-purpose assistant the customer already trusts.

5. Watch for inherited settlement behavior. Claims models trained on historical settlements inherit historical settlement patterns, and that is both a fairness exposure and a litigation exposure. (Independent audit: 6 months, then annually.)

6. Use AI on the content backlog. The compliance throughput problem is the binding constraint on the entire content strategy, and drafting, redlining and pre-screening against a cleared language library is a genuinely safe application. (Pilot: 90 days.)

7. Automate the local page layer. Thousands of genuinely differentiated agency pages are impossible by hand and straightforward with generation plus human review at the agency level. (Build: 6 months.)

8. Instrument for retrieval, not ranking. Structured coverage data, published pricing logic and a clean programmatic interface make the company legible to the systems that will do the shopping.

9. Price off vehicle telemetry rather than off a phone application, using the Arity data asset, with consent architecture designed for the enforcement environment that already exists. (Multi-year.)

10. Deploy AI in service triage where the frontline attrition problem is worst, because the marketing promise is delivered by the contact center and the contact center is the least funded part of the delivery chain.

11. Frame it internally as variance, never as headcount. The frontline cooperation the program needs evaporates the moment the framing becomes replacement, and adjusters are the source of the training data.

12. Build the agent copilot before the customer chatbot. An exclusive or independent agent who can quote, check appetite and answer a coverage question in seconds is a distribution advantage; a chatbot on the homepage is a cost saving.

13. Use AI to read the referrer logs and the claim files qualitatively at a scale no human team can, then hand humans the anomalies. (Pilot: 90 days.)

14. Publish an AI use policy in plain language before a regulator asks for it, covering what the company does with the model, where a human decides, and what a customer can appeal. (90 days.)

9.  Brief Hiring Ideas (if not already hired under different titles)

Twelve points. Stated as roles rather than as people, and sequenced. Time in parentheses is time to fill, not time to impact.

1. Head of AI Search Visibility, reporting into marketing, owning citation share, structured data strategy and machine-readable product documentation. Hire now, in the fourth quarter of 2026. (90 to 120 days to fill.)

2. VP of Incrementality and Measurement, independent of media buying, owning permanent holdouts and the marginal cost curve. Hire by the first quarter of 2027. (120 days.)

3. Director of Agent-Level Marketing, who builds the franchise playbook the field has been asking for. Hire in the first half of 2027. (90 days.)

4. Director of Claims Proof and documentary content, who turns the claims organization into publishable evidence with legal and privacy designed in from day one. (120 days.)

5. Head of Embedded Distribution to attack the dealership, lender and manufacturer channel before captive alternatives close it. (120 to 180 days.)

6. Machine Verifiability Lead sitting between product, legal and engineering, whose single job is to make the company the cheapest carrier in the market for an agent to verify. (120 days.)

7. Community manager for forums, empowered to be honest rather than defensive, whose first-year metric is credibility rather than sentiment. (60 days.)

8. Independent claims-quality analytics lead who does not report to claims, because a quality number owned by the measured organization is unfalsifiable. (120 days.)

9. A resilience underwriting team – underwriter, mitigation engineer, actuary – with its own profit and loss and a five-year mandate to write the hardened home. (6 to 9 months to assemble.)

10. State marketing managers in the top fifteen states with real budget authority, sitting beside the people who already own the rate filing. (6 months.)

11. A qualitative data analyst whose only job is reading raw referrer paths and claim narratives as evidence rather than as dashboard inputs. (60 days.)

12. Technical content engineers who can ship a page without an engineering ticket, because publishing velocity is currently gated by the wrong function. (90 days.)

10.  Open Questions Worth Exploring

Fourteen points. These are the questions the focus groups kept returning to and could not answer from public information. Several of them are answerable internally in a week.

1. Which business is the core – the exclusive agency franchise, the independent agency carrier, or the direct writer? Not answering is itself a decision, and marketing is currently asked to serve three different physics at once.

2. What is the target agency count? Six thousand, four thousand, or a number nobody has written down? The field is inferring it from a production statistic.

3. How often is one household quoted through two channels at different prices, and does anyone own that number?

4. What is ALLIE measured against, and would the company publish that evaluation set for anything touching a denial?

5. What percentage of estimates are revised upward after reinspection? One number settles the entire claims argument in either direction and no carrier will print it.

6. What would the plan look like at a 95 combined ratio instead of 86.6? Every recommendation here assumes marketing is funded out of margin.

7. Is branded search incremental? Nobody has tested it, it is testable in a quarter, and it is probably the largest single recoverable line in the budget.

8. What is a lost customer worth, fully loaded with reacquisition and word of mouth, and who sees that number monthly?

9. Could the company underwrite mitigation – pay for the hardening, then insure the hardened home, and keep the difference? A focused competitor is doing it at a 49 percent operating margin in the geography the nationals left.

10. Should protection plans lead the acquisition funnel rather than sit in a segment disclosure? It is the cheapest customer relationship in the company and it is not used as one.

11. What happens to the advertising when a majority of shopping is agent-mediated and there is no placement to buy?

12. Is the direct channel a competitor to the exclusive force, and if the honest answer is partly yes, what is the written rule that resolves it?

13. When is succession announced? Answering early is free and answering late is not, and the uncertainty freezes exactly the long-horizon marketing investments this document recommends.

14. Would the company publish a claims standard first, before any competitor, knowing the first number might be unflattering and that publishing it anyway is the entire point?

Disclosures

NO POSITION / NO COMPENSATION / NO MNPI. The author holds no long, short or derivative position in any security discussed, has received no compensation from any person or entity for the preparation, publication or distribution of this document, and has used no material non-public information. This is a marketing and operations discussion built entirely from publicly available sources. It contains no valuation, no price target, no earnings estimate, no recommendation to buy, sell or hold any security, and no mergers-and-acquisitions advice. The author is not affiliated with, endorsed by, or speaking for The Allstate Corporation or any competitor named here. For entertainment purposes only.

Figures cited are as publicly reported by the company or by reputable trade and financial press and are subject to revision, restatement and the company’s own definitions of non-GAAP measures. Time estimates are the author’s own judgment. Readers should consult the company’s filings directly.