Business questions, answered
73 short answers from 15 guides. Each set links to the full guide for the worked examples and the reasoning behind the answer.
AI for Sales, Marketing, Insurance and Restaurants: Prompts and Pitfalls
Does CAN-SPAM apply to B2B sales emails?
Yes. The CAN-SPAM Act applies to commercial email, and the FTC notes it makes no exception for business-to-business email. Its requirements include accurate header information, non-deceptive subject lines, identifying the message as an ad where it is one, a valid physical postal address, and a working opt-out honoured within 10 business days. An AI model will not add these unless told to.
Is it legal to use AI to write reviews for my business?
No, not in the US. The FTC's rule on consumer reviews and testimonials, 16 CFR Part 465, took effect on 21 October 2024 and bans fake reviews and testimonials, including ones attributed to people who do not exist, which covers AI-generated reviews. Businesses can use AI to reply to genuine reviews, never to write their own.
Can restaurants use AI for allergen information?
A model can format an allergen notice, but the allergen content of each dish must come from recipes and supplier specifications, checked by a person every time a recipe or supplier changes. In the US, sesame became the ninth major food allergen on 1 January 2023 under the FASTER Act, and in the UK 14 allergens must be declared.
What is the NAIC model bulletin on AI for insurers?
The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on 4 December 2023. It is binding only in states that adopt it, and it expects insurers to maintain a written program governing their AI systems. Agents and brokers may also have a carrier or agency AI policy built on it.
How do I reply to a negative restaurant review with AI?
Give the AI the review and what you know, such as the cause and any fix, and ask for a short public reply for the owner to edit. Tell it to thank the reviewer, acknowledge the problem without excuses, mention the fix once, avoid offering refunds or vouchers in public, and invite them to get in touch instead.
Barcode Types and Check Digits: EAN, UPC, Code 128 and More
What is a GTIN?
A GTIN, or Global Trade Item Number, is the number carried by GS1 barcodes such as EAN-13, UPC-A, EAN-8 and ITF-14. A company licenses a GS1 Company Prefix and assigns item numbers under it, so numbers are unique worldwide. Every variant sold separately, such as a different size, colour or pack count, gets its own GTIN.
How do barcode check digits catch errors?
A GS1 check digit catches every single-digit error, because the weights 3 and 1 are both coprime with 10, so one wrong digit changes the weighted sum by a non-multiple of 10. Swapping two neighbouring digits changes the sum by 2 x their difference, which is caught unless the digits differ by 5.
What is an ITF-14 barcode used for?
ITF-14 is used on outer cartons and cases. It is Interleaved 2 of 5 carrying a 14-digit GTIN, usually with a thick bearer bar around it so it scans when printed directly on corrugated cardboard. Interleaved 2 of 5 encodes digits in pairs, one in the bars and one in the spaces.
What is a barcode quiet zone?
A quiet zone is the blank space a barcode needs on both sides so the scanner can find where it starts. For EAN-13 the GS1 specification asks for 11 module widths on the left and 7 on the right, and Code 128 needs at least 10 modules on each side. Packaging artwork, text or label edges must not intrude.
What colours can a barcode be printed in?
Black, dark blue or dark green bars on white or light yellow are safe for barcodes. Most laser scanners use red light, so red bars disappear and red backgrounds look white. Avoid printing barcodes on transparent film without a white underlay, and keep high contrast between bars and spaces.
Depreciation Methods Compared: Straight-Line, DDB, SYD and MACRS
What is units of production depreciation?
Units of production depreciation charges each year by the output actually used rather than by time. Divide the depreciable base by the total units the asset should produce, then multiply by the year's units. A 50,000 machine with 5,000 salvage and a 100,000-unit life depreciates 0.45 per unit, so a 30,000-unit year costs 13,500 and an idle year costs nothing.
What is the depreciable base of an asset?
The depreciable base is cost minus salvage value, the amount spread over the asset's useful life. Cost is the purchase price plus what it took to get the asset working, such as delivery and installation, and salvage value is what you expect to get for it at the end. A 10,000 oven expected to fetch 1,000 has a depreciable base of 9,000.
Can land be depreciated?
No, land is never depreciated, because it does not wear out. Depreciation matches the cost of an asset used for several years to the years that benefit from it, so it applies to assets such as buildings, furniture, machines, vehicles and computers that give up their usefulness over time.
What is the MACRS half-year convention?
Under the MACRS half-year convention, every asset is treated as placed in service in the middle of the tax year, so a 5-year asset is deducted over six tax years: 20, 32, 19.2, 11.52, 11.52 and 5.76 percent in IRS Publication 946, Table A-1. If more than 40% of the year's purchases arrive in the last quarter, the mid-quarter convention applies instead.
Is book value the same as what an asset is worth?
No. Book value is cost minus accumulated depreciation, and depreciation is an accounting allocation, not a valuation, so book value is not what the asset would sell for. After two years, the same 10,000 oven with 1,000 salvage shows a book value of 6,400 under straight-line and 3,600 under double-declining balance, and neither figure is wrong.
FMEA and RPN Explained: Severity, Occurrence, Detection
What is FMEA?
Failure Mode and Effects Analysis (FMEA) is a worksheet method that lists every way a design or process can fail, rates how bad, how likely and how hard to catch each failure is, and works down the list from the top. It started as the US military procedure MIL-P-1629 in 1949, was applied by NASA in Apollo and was taken up by the automotive industry in the 1970s.
What do severity, occurrence and detection mean in FMEA?
In FMEA each failure mode gets three ratings from 1 to 10. Severity is how bad the effect is if it happens, occurrence is how likely the cause is, and detection is how likely current controls are to catch it before the customer. Detection runs backwards: a test that catches every case is a 1, and a failure that would almost certainly escape is a 10.
Why can two failure modes with the same RPN have different risk?
RPN multiplies ordinal ratings, so different combinations give the same number. A wrong dose label rated 9, 2 and 5 and an off-centre box label rated 3, 6 and 5 both score an RPN of 90, yet the first could hurt a patient and the second is cosmetic. Most practitioners therefore act on any severity 9 or 10 failure regardless of its RPN.
What is Action Priority in FMEA?
Action Priority (AP) is the method in the 2019 AIAG-VDA FMEA Handbook that replaced RPN. Instead of multiplying, AP uses lookup tables that combine severity, occurrence and detection, giving severity the most weight, and rates each failure mode High, Medium or Low priority. A high-severity failure can come out as High priority even when occurrence and detection are good.
Can corrective actions lower the severity rating in an FMEA?
Usually not. Severity describes the effect of a failure, so only a design change that alters the effect can lower it. FMEA actions normally work on occurrence, by preventing the cause, or on detection, by catching the failure earlier. In a checkout example, optimistic locking cut an inventory row's RPN from 135 to 18 while its severity stayed at 9.
Force Field Analysis: Planning Change by Removing Resistance
What is force field analysis?
Force field analysis is a change-planning method that puts a proposed change in the middle, lists the driving forces pushing for it on one side and the restraining forces holding it back on the other, and rates each one. It comes from Kurt Lewin's field theory, which holds that any stable situation is held in place by a balance of forces.
What is Kurt Lewin's three-stage change model?
Kurt Lewin's three-stage model of change is unfreeze, change, refreeze. It comes from the same work as force field analysis, which is the unfreeze step made concrete: you cannot loosen a situation until you can see the driving and restraining forces that hold it in place.
How do I score forces in a force field analysis?
Give each force a strength from 1 to 5 based on how much it would move the outcome: 1 is minor, 2 noticeable, 3 significant, 4 major enough to delay the change by months, and 5 decisive. Score by effect on the decision, not by how often people complain, because the loudest objection in the room is often a 2.
Does a positive net force mean a change will succeed?
No. The net force in a force field analysis is not a probability of success, and the scores are structured opinion rather than measurement. One force rated 5 can outweigh five forces rated 1, and a column with more forces is not necessarily stronger. Look at the highest-rated forces on each side, and still remove the strongest restraint before a rollout.
What is an example of a force field analysis?
A rollout of an expense app scored driving forces of 5, 4, 4 and 3, totalling 16, against restraining forces of 4, 4, 3 and 2, totalling 13, for a net force of +3. Fixing the sales commission process that still needed the old spreadsheet moves the net to +7, and also fixing mobile receipt capture moves it to +11.
How to Calculate Your Freelance Hourly Rate (and Day Rate)
What is the formula for a freelance hourly rate?
A freelance hourly rate is the revenue needed divided by billable hours. Revenue needed is (target take-home / (1 - tax rate) + business expenses) / (1 - extra margin), and billable hours are working days x hours per day x billable share. The day rate is then the hourly rate x hours per day.
What is a good utilisation rate for a freelancer?
Utilisation, or billable share, is the fraction of working hours a client pays for. Freelancers building a client base in their first year often bill 40 to 50%, established ones with a few steady clients 60 to 70%, and those on long full-time agency contracts 75 to 85%. Over 90% is rare and leaves no time for sales.
What is the self-employment tax rate in the US?
Self-employed people in the United States pay self-employment tax of 15.3%, made up of 12.4% for Social Security up to the annual wage base and 2.9% for Medicare, on 92.35% of net earnings. Federal and possibly state income tax come on top, and half of the self-employment tax is deductible when working out income tax.
How do I gross up take-home pay for tax as a freelancer?
Divide the take-home target by one minus the effective tax rate, rather than multiplying by one plus the rate. Keeping 60,000 at a 25% effective rate needs 60,000 / (1 - 0.25) = 80,000 of profit before tax, because 25% of 80,000 is 20,000, which leaves exactly 60,000. Business expenses are then added on top.
How do I price a fixed-price freelance project?
For a fixed-price project, estimate the hours, multiply by the hourly rate, and add contingency for scope changes, commonly 15% to 25%. Tracking the actual hours afterwards improves future estimates. The hourly rate itself is a floor rather than a ceiling: it is the minimum that sustains the business, and clients may pay more for valuable results.
How to Use a 5x5 Risk Matrix (and Where Its Scores Mislead)
What are the problems with risk matrices?
Risk matrix ratings are ordinal, so multiplying them gives a score that looks precise but is not. A rare catastrophe at probability 2 and impact 5 scores the same 10 as a routine slip at probability 5 and impact 2, scores cannot be summed across a register, and ties hide differences. Louis Anthony Cox Jr. set these problems out in his 2008 paper "What's Wrong with Risk Matrices?".
Why are some scores missing from a 5x5 risk matrix?
Because both ratings are whole numbers from 1 to 5, a 5x5 risk matrix can produce only fourteen scores: 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, 16, 20 and 25. There is no 7, 11 or 18, which is why the bands jump from 16 at the top of High straight to 20 at the bottom of Critical.
How do I define likelihood and impact scales for a risk matrix?
Write one line for each level of each scale before anyone scores, tied to the project. For a 2 million ERP rollout, probability 1 might be under 5% and 5 over 80%, while impact 1 is under 10,000 or under a day of delay and impact 5 is over 500,000, a missed go-live or a regulatory breach. Written scales stop people scoring the same risk two levels apart.
What do the colours in a 5x5 risk matrix mean?
In the Risk Assessment Matrix, scores of 1 to 4 are Low (green), 5 to 9 are Medium (yellow), 10 to 16 are High (orange) and 20 to 25 are Critical (red). These bands are a common convention rather than a requirement of ISO 31000, so organisations with their own thresholds should use those.
How do I write a good risk statement for a risk register?
Name a specific cause or event rather than a vague category. "Resource issues" is not a risk, while "key consultant leaves mid-project" is. A specific statement lets the team argue about evidence, such as whether the consultant has been approached by a competitor, instead of arguing about what a score of 3 means.
How to Write an Invoice: Fields, Tax, Discounts and Terms
What should be included on an invoice?
An invoice should show the word "Invoice", a unique invoice number, the issue and due dates, the seller's name, address and contact, any tax registration number, the customer's name and address, a description, quantity and unit price for each line, the subtotal, discount, tax and total, payment details, and the customer's purchase order or reference.
What does 2/10 Net 30 mean?
2/10 Net 30 means the customer gets 2% off if they pay within 10 days, and otherwise owes the full amount within 30 days of the invoice date. The early-payment discount is expensive for the seller: giving 2% to be paid 20 days earlier works out to an annualised cost of about 37%, so it suits cases where cash flow really matters.
What must a UK VAT invoice include?
HMRC's guidance asks a UK VAT invoice to show a unique number that follows on from the last, the business name and address, the VAT registration number, the invoice date, the time of supply if different, the customer's name and address, a description, quantity and price excluding VAT for each item, the VAT rate and the total VAT charged.
How do I correct an invoice that has already been sent?
Issue a credit note against the wrong invoice and then send a new invoice with a new number, rather than editing and resending the same number. Invoice numbers must be unique, and in VAT systems they must follow a sequence, so avoiding gaps and reuse keeps both your records and the customer's consistent.
Should tax be rounded per line or on the invoice total?
Tax can be rounded on each line or once on the total for each rate, and the two can differ by a cent or two on long invoices. Both are accepted in most systems; HMRC, for example, allows VAT to be calculated per line or on the invoice total as long as the method is used consistently.
Margin, Markup and Break-Even: The Arithmetic of Pricing
What is the break-even formula?
Break-even units equal fixed costs divided by contribution per unit, where contribution is price minus variable cost. Break-even revenue equals fixed costs divided by the contribution margin ratio. With fixed costs of 80,000, a price of 250 and a variable cost of 150, break-even is 800 units or 200,000 of revenue. Round fractional answers up.
How many units do I need to sell for a target profit?
Add the target profit to fixed costs and divide by contribution per unit. With fixed costs of 80,000, a contribution of 100 per unit and a 20,000 profit target, that is 1,000 units. For an after-tax target, gross it up first: keeping 15,000 at a 25% tax rate needs 20,000 of pre-tax profit.
What is margin of safety in break-even analysis?
Margin of safety is how far expected sales can fall before a business stops making money, calculated as (expected sales minus break-even sales) divided by expected sales. Expecting 1,000 units against a break-even of 800 gives a 20% margin of safety. Lenders read it as a blunt but honest measure of how much bad luck a plan can absorb.
How do tiered commission plans work?
Marginal, or graduated, tiers pay each slice of sales at its own rate, while whole-amount, or retroactive, tiers pay the highest rate reached on everything. With tiers of 5% to 10,000, 7% to 20,000 and 10% above, sales of 25,000 pay 1,700 on a marginal plan but 2,500 on a whole-amount plan, which creates cliffs at thresholds.
What is a draw against commission?
A draw against commission is an advance paid every period. Under a recoverable draw, any shortfall against earned commission carries forward and is repaid from later commission, so total pay equals total commission earned. A non-recoverable draw forgives each shortfall and works as a guaranteed minimum. What happens to a balance when someone leaves depends on the plan and local wage law.
Performance Reviews and Development Plans Engineers Can Act On
What is a performance review rating scale?
A performance review rating scale is usually five points. In the Performance Review Templates, 5 is Exceeds Expectations, 4 Meets Expectations, 3 Partially Meets, 2 Below Expectations and 1 Does Not Meet. Company scales may label the numbers differently, but in every case the written comments should justify the number given.
What are the most common biases in performance reviews?
The most common performance review biases are recency bias, where the last six weeks crowd out the rest of the year, halo and horns, where one quality colours every section, vague adjectives with no example, and central tendency, where everyone gets a 3 or 4 to avoid hard conversations. A running list of dated wins and misses counters recency bias.
How do you use the STAR method in a performance review?
To use the STAR method in a performance review, write each comment as situation, action and result. For example: checkout latency regressions were triggering weekly pages, Dana profiled the service and fixed an N+1 query, and p95 latency fell from 840 ms to 310 ms. Three sentences with a number and a date range hold up in calibration.
What is an individual development plan?
An individual development plan is two or three goals for the next review period, each with an action and a way to tell whether it happened. The goals should be tied to real work, include the manager's actions with dates, and be checked monthly in 1:1s. George Doran's SMART criteria from 1981 are the quickest test of each goal.
What makes a development goal SMART?
A SMART development goal replaces a vague aim with something concrete and dated. "Get better at system design" becomes: write the design doc for the ledger rewrite, have it reviewed by two senior engineers and present it at the architecture review by 15 March. George Doran set out the SMART criteria in 1981.
PEST and PESTLE Analysis: Writing Factors That Lead to Decisions
What does PEST stand for?
PEST stands for Political, Economic, Social and Technological. A PEST analysis is a structured scan of the outside forces that no single company controls, such as regulation, wage rates, household trends and new software. It looks at the world around a business rather than its internal issues, which belong in a SWOT analysis instead.
What is a STEEP or STEEPLE analysis?
STEEP covers Social, Technological, Economic, Environmental and Political factors and is common in futures and scenario work. STEEPLE adds Legal and Ethical, and is used in sustainability and ESG reporting. Both are the same exercise as PEST with more categories, so the choice matters less than using the same version each year so analyses can be compared.
How do I write a good factor for a PEST analysis?
A useful PEST factor names a specific change, ideally with a number or date, gives its direction, and states the effect on your business. "Inflation" is too generic to help. "Delivery driver pay is rising faster than our box price" already hints at a decision. If a factor suggests nothing you might do differently, it probably does not belong.
How many factors should a PEST analysis have?
Three to six well-written factors per column is plenty for most businesses. Twenty factors per column means nothing has been prioritised. A good PEST analysis fits on one screen, names real changes with dates, and leaves two or three things to act on, while most of the remaining factors simply inform planning.
Product Manager Interviews: Product Sense, Metrics and Estimation
What is the CIRCLES method in PM interviews?
The CIRCLES method, from Lewis Lin's book on product manager interviews, is a structure for product sense questions. It follows a similar order to clarifying the goal, choosing a user segment, listing that user's problems, prioritizing one, generating solutions, choosing one and defining success. Its value is narrowing before expanding, because a narrow segment makes every later step concrete.
How do I answer a PM strategy question like "should company X enter market Y"?
Work through why X would want to enter, how attractive Y is in size, growth and margins, who already wins there and why, what X's advantage would be, how X would enter by building, buying or partnering, and the main risks. Then give a clear recommendation and state the condition that would change your mind, because "it depends" without a recommendation loses the round.
Which metrics frameworks help in a product manager interview?
Google's HEART framework (Happiness, Engagement, Adoption, Retention, Task success) and the AARRR funnel (Acquisition, Activation, Retention, Referral, Revenue) are useful vocabularies for choosing what to measure in a product manager interview. Name one primary metric rather than listing a whole framework, and add only two or three guardrail metrics alongside it.
Do product managers need technical skills for interviews?
Product manager candidates do not need to code, but they should be able to discuss APIs, latency, data pipelines and what makes a feature expensive to build. Execution and technical rounds assess trade-offs with engineering, sequencing and risk. Engineers moving into product should spend extra preparation time on metrics and strategy, which engineers practise least.
Scenario Planning: Building Futures That Change What You Do
What is scenario planning?
Scenario planning is a strategy method in which, instead of betting on one forecast, you describe a few plausible futures, decide in advance how you would respond to each, and agree the signals that tell you which one is unfolding. Its modern roots are Herman Kahn's work at the RAND Corporation in the 1950s and Pierre Wack's team at Royal Dutch Shell in the early 1970s.
How do you build a 2x2 scenario matrix?
List the external drivers that matter for the decision, then rate each for importance and uncertainty. Drop drivers that are predictable or unimportant, pick the two that are most important and most uncertain, and cross them. Each of the four quadrants becomes a scenario with its own internally consistent story. If one quadrant is implausible, drop it and keep three.
Is best case, base case and worst case planning the same as scenario planning?
Not really. A best, base and worst case that differ only in one number, usually revenue growth, is a sensitivity analysis rather than a scenario set. The futures that hurt a business are usually different in kind, not just in size, which is why the 2x2 of critical uncertainties produces scenarios that need genuinely different responses.
What are no-regret moves in scenario planning?
No-regret moves are actions that help in every scenario, so they can be approved and started immediately. Actions that only pay off in some futures become contingent moves, attached to a specific trigger signal instead of a date. A short list of no-regret moves plus a set of contingent moves with clear triggers is the main product of a good scenario plan.
How often should a scenario plan be reviewed?
Assign someone to check the scenario signals monthly and review the plan formally every quarter. Scenarios age, so update them when a signal fires or at least once a year. Signals work best when they are specific and checkable, such as "diesel above 1.80 per litre for two consecutive months" rather than "fuel prices rise".
The 5 Whys: Finding Root Causes Without Blaming People
What is the 5 Whys technique?
The 5 Whys is a root cause analysis method: you keep asking why the previous answer happened until you reach a cause you can change that would stop the problem coming back. It comes from Toyota, where Taiichi Ohno made it standard on the shop floor. Five is a rule of thumb, and each answer should be backed by evidence.
What is Taiichi Ohno's 5 Whys example?
Taiichi Ohno's 5 Whys example is a machine that stopped. A fuse blew from an overload, because the bearing was not lubricated, because the lubrication pump was not pumping enough, because the pump shaft was worn, because there was no strainer and metal scrap got in. Replacing the fuse restarts the machine, but fitting a strainer stops it happening again.
How do I write a good problem statement for a 5 Whys analysis?
A good 5 Whys problem statement is a fact with a time, a place and a number, not a feeling or a solution. "Dispatch is a mess" cannot be analysed and "We need more staff" is a conclusion, but "14 of 120 orders missed the 3 pm same-day dispatch cutoff on 2 October" can be traced to a cause.
What are the limitations of the 5 Whys?
The 5 Whys follows one chain, so problems with several contributing causes can lose all but one. Results depend on who is asking, since two teams can reach different roots on the same incident by following the branch they know best. The method also does not rank risks: it explains one problem after it happened rather than scoring many possible failures.
What is the difference between 5 Whys and FMEA?
The 5 Whys works back from one problem that has already happened to find its root cause. Failure Mode and Effects Analysis (FMEA) looks ahead at many possible failures and scores each for severity, occurrence and detection to decide which to address first. A 5 Whys finding can feed back into an FMEA worksheet as a new or re-rated row.
Which Compliance Rules Apply to Your Software? A Map for Engineers
What is the difference between SOC 2 and ISO 27001?
SOC 2 is an attestation: a report from an independent CPA firm against the AICPA Trust Services Criteria. ISO 27001 is a certification of an information security management system by an accredited body. Neither is required by law, but enterprise customers may require either one before signing, and both rest on controls such as access control, audit logs and encryption.
What are the maximum fines under GDPR?
The upper tier of GDPR fines is EUR 20 million or 4 percent of worldwide annual turnover, whichever is higher. GDPR has applied since 25 May 2018 and covers any company offering goods or services to, or monitoring, people in the EU, wherever its servers are. Many other regimes, such as HIPAA and COPPA, adjust their penalty figures for inflation every year.
How quickly must a data breach be reported under GDPR?
Article 33 of GDPR requires notifying the supervisory authority of a personal data breach within 72 hours of becoming aware of it. A written incident response plan, an on-call path and notification templates should be built to meet the shortest clock you are subject to, which under GDPR is 72 hours.
When did PCI DSS 4.0 become mandatory?
PCI DSS version 4.0 became the only active version when version 3.2.1 retired in March 2024, and its future-dated requirements became mandatory at the end of March 2025. PCI DSS is an industry standard rather than a law: the card brands set it through the PCI Security Standards Council, and acquiring banks enforce it through merchant agreements.
How can a startup reduce its PCI DSS scope?
Keep card numbers off your own servers. Using a payment processor's hosted fields or a redirect means card data never touches the startup's systems, which shrinks the PCI DSS assessment to a short self-assessment questionnaire instead of a full audit. Watch for accidents such as logging middleware that records request bodies, which can put the whole stack in scope.