A pricing plan can be right about the product, right about the customer, and right about the competition, and still fall apart on one line nobody wrote down: that customers would accept the increase. Federal Reserve district reports through 2026 kept describing the opposite. In February 2026, Kansas City contacts said many firms were reluctant to raise prices further because customers had become more price sensitive and demand conditions remained soft. By July, the St. Louis report described hospitality businesses rolling back price increases they had already made.
That is not a forecasting failure. It is a failure to notice which assumption the whole plan was resting on. And it is the core of what it takes to build a strategy that survives market changes: not better prediction, but a clear-eyed inventory of what your plan quietly depends on, and a decision made in advance about what you will do when one of those things stops being true.
Every strategy has an “as long as” clause
Write your current strategy as one sentence. Something like: “We grow by selling implementation services to mid-sized logistics companies through outbound email and partner referrals.”
Now finish the sentence honestly, over and over:
- …as long as our two largest accounts renew.
- …as long as we can hire senior people at roughly current rates.
- …as long as cost per qualified lead stays under our current number.
- …as long as insurance and energy costs stay near where we budgeted them.
- …as long as our payment provider, hosting, and core vendors keep working the way they do today.
That list is the actual strategy. The sentence above it is just the headline. In February 2026, the Atlanta Fed reported that insurance remained a key cost driver across sectors, with some firms moving to self-insure rather than absorb rising premiums. Not one business owner in that group had “insurance premiums stay flat” written into their plan. They were all betting on it anyway.
The discipline here is unglamorous. Keep the list somewhere you will actually see it, phrase each item so it could be proven false, and stop treating it as a planning exercise you finish. A strategy you cannot break into falsifiable assumptions is not a strategy. It is optimism with a deadline.
Count your ones
Change alone does not break businesses. Change meeting concentration breaks businesses. A 30% cost increase in an input you buy from four suppliers is a bad quarter. The same increase from your only supplier, on your only product line, is an existential event.
So before you plan anything, count the places where you have exactly one of something.
One customer, one supplier, one channel
Customer concentration is the one people track, usually badly. My rule of thumb: once a single client passes roughly a quarter of revenue, you are no longer running a business with a customer. You are running a department inside their business, with worse job security. That does not mean you refuse the work. It means the strategy has to include a funded plan for replacing that revenue, not a vague intention to diversify later.
Channel concentration is more dangerous because it hides better. If nearly all new customers arrive through one paid platform, one marketplace, or one referral partner, your acquisition economics are set by someone whose interests are not aligned with yours. The cost of that channel will rise over time. It always does.
The dependency almost nobody puts in the plan
Here is the gap in most resilience advice. It talks about suppliers and customers as if it were still 2010, and ignores the fact that a modern small business runs on other companies’ software, rules, and price lists.
Look at what actually happens on those platforms. Stripe’s changelog shows a version family released on June 30, 2025 that added support for crypto payments through the Payment Methods API, settling as fiat in the Stripe balance, and a later version, 2025-09-30.clover, that introduced both breaking changes and new features. Two different kinds of event in the same year: one that hands you a new option, one that hands you engineering work. Both belong in a strategy conversation.
App distribution carries a standing upkeep bill. Apple’s App Review Guidelines, last updated June 8, 2026, require that apps use only public APIs, run on the currently shipping operating system, and phase out deprecated features, frameworks, and technologies. Material changes such as changing your business model mean restarting pre-order sales. Apple’s App Store Improvements page notes that developers of apps not updated within the last three years and below a minimal download threshold may receive notice of possible removal. Approval is not a finish line. It is a subscription with compliance terms.
AI vendors move faster still. OpenAI’s release notes record GPT-Live 1 becoming generally available on September 10, 2026, with voice sessions priced at $0.05 per minute, billed per second, and backend model and tool usage billed separately. If your product margin depends on that line item, your margin sits on someone else’s roadmap and your unit economics move with average call length. That is a strategic fact, not an engineering detail.
None of this argues for avoiding platforms. It argues for pricing the dependency: who can change your cost structure, your distribution, or your feature set without asking you, and what would you do in the first week if they did?
Decide what is load-bearing and what is furniture
A strategy that flexes on everything is not a strategy, and one that flexes on nothing is a museum. The useful question is which parts are structural.
Load-bearing, in my view, is a short list: who you serve, the problem you solve better than the alternatives, the standard of work you will not drop, and the economics you refuse to break, such as a minimum gross margin or a cash floor. Everything else is furniture. Channels, pricing structure, packaging, tech stack, team shape, geography. Furniture should move whenever the room changes.
The common failure is doing this exactly backwards. Businesses defend a legacy tool or a favourite service line for years, then change target customer twice in eighteen months because two quarters were soft. That combination produces churn without adaptation. It is one of the quieter reasons small businesses lose focus as they grow.
Occasionally the load-bearing wall really does need to move. Microsoft’s pivot toward cloud under Satya Nadella starting in 2014 is the standard example, and the thing worth noticing is what it was not. It was not a better version of the old core. It was a deliberate change in what the company was exposed to, because market structure had shifted. That kind of move should be rare, argued for explicitly, and never made in reaction to a single bad quarter.
Replace the annual review with triggers
Harvard Business Review’s 2024 argument about operating in volatility is that firms need prediction, adaptability, and resilience working together rather than picking one. I think that is right, but it needs a mechanism, and the mechanism is this: prediction’s job is not to produce confidence. Its job is to produce indicators you can watch.
Every assumption on your list gets three things attached to it. An observable indicator. A threshold. A response you decide now, while you are calm.
The pre-commitment matters more than the numbers. In the moment, every threshold breach has a reasonable explanation. Seasonality. A one-off. A bad month. Writing the response down in advance is how you stop yourself negotiating with the evidence. If you want the longer version of that reasoning, it overlaps heavily with making better business decisions with incomplete information.
Pick indicators that move before revenue does
Revenue is a lagging confession. By the time it drops, the decision window has been open for months. Better triggers sit upstream: quote-to-close rate, average discount granted, cost per qualified lead, renewal conversations that go quiet, supplier lead times stretching, a vendor deprecation notice landing in an engineering inbox.
The H-E-B example from that HBR piece is instructive precisely because it is boring. The chain began taking supplier-sensing actions in mid-January 2020, before COVID disruption had fully hit the United States. The advantage came from watching the supply side while everyone else was watching sales.
Use the free instrumentation
You do not need to buy market intelligence to see conditions shift. The Census Bureau’s Business Trends and Outlook Survey draws on a sample of about 1.2 million businesses, split into six panels of roughly 200,000 each, covering all U.S. employer businesses excluding farms from September 11, 2023 onward. It publishes high-frequency, near-real-time data on performance, revenue, employment, hours worked, demand, and prices. The Federal Reserve’s Beige Book adds the qualitative half: the August 2026 summary described a positive overall outlook but mixed sentiment, with uncertainty tied to higher energy prices, policy, and international conflict, and noted in the Chicago summary that business spending declined slightly while prices rose moderately and financial conditions tightened slightly.
That is a better early-warning kit than most annual industry reports, and it updates while the year is still happening.
A cadence that does not become bureaucracy
Monthly, check the indicators. Thirty minutes. Quarterly, reread the assumption list and ask which items now feel shakier than they did, then reallocate budget accordingly. Annually, ask one hard positioning question: is the problem we solve still the problem this market most wants solved? And separately from all of it, any trigger breach earns an unscheduled session within two weeks, regardless of what the calendar says. Strategy that never becomes a set of dated priorities stays decorative, so each of those sessions should end by feeding your 90-day execution plan.
Treat market research as a subscription, not a launch task
The SBA’s guidance is that market research helps you find customers while competitive analysis helps you make your business unique, and that together they help you find a competitive advantage. It suggests market research should answer questions about demand, market size, economic indicators, location, market saturation, and pricing, and that competitive analysis should assess market share, strengths and weaknesses, window of opportunity, the importance of your target market to competitors, barriers to entry, and indirect competitors. Its business-plan guidance adds that market analysis should look for trends and themes, and study what successful competitors are doing and why it works.
All sensible, and almost universally done once, at founding, then filed.
Run it again every year and read it as a difference report rather than a document. What changed since last time? Two of those dimensions do most of the survival work. Window of opportunity tells you whether the opening you built on is still open. Indirect competitors is where displacement actually comes from, because the thing that replaces you usually does not look like you, does not describe itself the way you do, and does not show up in a search for your category. Direct rivals take share. Indirect ones remove the need. If you want a sharper version of that question, it runs through how you differentiate your business online.
Buy slack where failure is unrecoverable, and nowhere else
The efficiency-versus-resilience argument is usually posed as a philosophy question. It is a purchasing question. Redundancy is insurance, and insurance is worth buying only against losses you cannot absorb.
Apply a recovery-time test to each dependency. If this breaks on a Tuesday, how long until we are operating again, and can we survive that gap? A hosting provider you could migrate off in three days needs no backup plan. A single supplier of the component your main product cannot ship without, with a four-month replacement cycle, needs a second source even though the second source costs more per unit and you will resent paying it every month for years.
That is the honest trade. Selective redundancy on critical dependencies, ruthless leanness everywhere else. Blanket slack is just weak management with a nice name.
Cash is the exception, because it is the only buffer that works against risks you failed to anticipate. It is also the buffer that disappears exactly when you need it. When the Fed describes financial conditions tightening, that is the moment the credit line you were relying on gets smaller or more expensive. Optionality is cheapest to buy when you do not need it.
Three pieces of standard advice to drop
The failure statistics. You will see claims that 70%, 80%, or 90% of strategies fail. These numbers circulate widely, but they trace back to weak sourcing, shifting definitions, and recycled consulting lore. Do not build a planning process on folklore, and be suspicious of any article whose urgency rests on one.
Blanket AI positions, in either direction. Census reporting shows that between December 14, 2025 and May 3, 2026, overall AI use among U.S. businesses hovered between 17% and 20%, with 20% to 23% expecting to use it in the next six months. As of May 3, 2026, 37% of firms with at least 250 employees and 32% of firms with 100 to 249 employees reported using it, while use in Information was 39.7% and in Finance and Insurance was 33.9%, against a national rate of 19.8%. Over that same period, use rose among firms with at least 20 employees but did not change significantly among smaller ones. Earlier Census research, covering September 2023 to February 2024, put biweekly estimates at 3.7% rising to 5.4%, with common uses being marketing automation, virtual agents, and data and text analytics.
Resist the temptation to draw a straight line between those two periods. The survey question changed in November 2025, from AI use in producing goods or services to AI use in any business function, which means the before-and-after numbers are not cleanly comparable. That is worth sitting with for a second, because it is the same trap your own dashboard sets. Adoption varies enormously by size and sector, so the defensible position is selective: adopt where a specific function has clear economics, ignore the rest, and measure carefully enough that you know when your own definitions changed.
Annual planning as sufficient. With near-real-time official data on demand and prices available for free, and vendor policy changes landing mid-quarter, a once-a-year cycle is not a plan. It is a ritual.
What this looks like on one page
Your strategy sentence. Under it, the assumptions it depends on, phrased so each could be proven wrong. Beside each assumption, the indicator you will watch, the threshold that counts as a breach, and the response you have already agreed to. Then a short list of your ones: single customers, suppliers, channels, platforms, and key people. Then the two or three places you have deliberately paid for redundancy, and the reason.
That page takes an afternoon to write and twenty minutes a month to maintain. It will not tell you what the market is going to do. It will tell you, much earlier than your revenue would, which of your beliefs has stopped being true, and it will have already decided what you do about it before the pressure arrives to do nothing.
