Definitions.
Same Word. Different Industry.Different Reality.Every industry uses the same operational vocabulary. Returning customer. Abandonment. Conversion. Active user. Risk.
None of them mean the same thing across industries. Most organizations have never formally defined what these words mean inside their own operations.
This is the problem Arché solves.
Most companies don’t have a data problem.
They have a meaning problem.
Two dashboards show different revenue numbers.
They are using different definitions of 'revenue'.
Marketing and Sales report different conversion rates.
They are measuring conversion at different boundaries.
A returning customer metric looks healthy at 40%.
Nobody asked: returning within what timeframe? Across which categories? By what measure?
Cart abandonment is at 68%. The industry average is 70%.
The system counts a price-comparison browser the same as a buyer who failed at checkout.
A predictive model is 94% accurate.
It was trained on data whose definitions changed three times in two years. It is accurate against a reality that no longer exists.
Compliance passed the audit.
The report was reconstructed from systems that were never designed to produce it. It reflects what the organization believes happened.
None of these are technical failures. The systems work. The integrations hold. The models run. The problem is underneath — at the layer where the definitions live. That layer was never designed. It just happened.
Returning Customer.
“A customer who has purchased more than once.”
A household goods retailer and a grocery chain both report 40% returning customers. One measures across years, the other across weeks. The metric is identical. The reality is not.
Bought a second appliance within 5 years. Happens once or twice in a decade.
Ordered again this week. Happens 50 times a year.
Returned for a follow-up within the care protocol window. Could mean good care — or misdiagnosis.
Placed a repeat PO for the same part. Could be a multi-year supply contract — or no qualified alternative supplier yet.
Renewed the subscription. Could mean daily usage — or a forgotten billing cycle.
Placed a second order within the reorder cycle. A quarterly chemical buyer is a different signal than a quarterly office supply buyer.
2+ transactions over lifetime, with first-to-second purchase gap under 5 years. Category expansion (e.g. washing machine → oven) signals deeper loyalty than same-category repeat.
Without this distinction, a customer who buys one appliance every 8 years is classified the same as someone building their entire household through your store. Marketing spend treats them identically.
A patient with 2+ visits within the care protocol window. A chronic patient returning for scheduled follow-ups is fundamentally different from an acute patient returning with a new complaint.
Clinics that count all revisits as 'returning patients' cannot distinguish between care continuity and care failure. A high return rate could mean excellent chronic management — or misdiagnosis.
A buyer that places repeat orders within the standard production lead-time window. Must distinguish between: contract-obligated reorders (annual supply agreements with fixed volumes), spot reorders (one-off needs driven by demand spikes or supply disruptions), specification-loyal reorders (same part, same spec — genuine product preference), and price-driven reorders (lowest bidder won again, not loyalty). A customer that reorders the same part at the same price is not the same signal as one that reorders despite a 5% price increase.
Sales teams treating all repeat POs as customer loyalty cannot distinguish between a procurement department locked into a multi-year supply agreement and one that reorders because no alternative supplier is qualified yet. The first is contractual inertia. The second is a switching cost — and the moment a competitor qualifies, the orders stop.
An account that renews after the first contract period AND maintains usage above the activation threshold. Renewal without usage is inertia, not loyalty.
SaaS companies celebrating 90% retention while 30% of renewed accounts are inactive are measuring billing persistence, not product value. The churn is invisible until it is not.
An account that places 2+ orders within the standard reorder cycle for their industry segment. A quarterly buyer in chemicals is a different signal than a quarterly buyer in office supplies.
Account managers treating all repeat purchasers equally cannot distinguish between a locked-in procurement relationship and a customer actively comparing alternatives between each order.
Abandonment.
“A user who started a process and did not complete it.”
An e-commerce site and a hospital both report 35% abandonment. One loses a cart. The other loses a patient. The word is the same. The operational response should be completely different.
Put a washing machine in the cart, left. Probably comparing prices — normal behavior for a 2–6 week decision cycle.
Built a full weekly cart and left at checkout. Something went wrong — delivery slot, coupon, payment. Immediate intervention needed.
Patient stopped showing up for treatment. Could be cured, referred elsewhere, or lost to the system entirely.
Confirmed PO cancelled before production. Customer demand shift? Spec change? Or your lead time was too long and they found someone faster?
Signed up, never activated. Could be wrong audience, bad onboarding, or a competitor won.
Qualified lead went silent after the proposal. Budget cut? Stakeholder changed? Or just vacation?
Must be segmented: research cart (< 30min session, information-gathering), decision cart (2+ sessions, returned to review), price-sensitive drop-off (reached checkout, stopped at payment), and comparison cart (multiple items in same category). Each requires a different response window and intervention type.
Treating all cart abandonment as a single remarketing trigger means sending discount codes to researchers (premature), urgency messages to comparers (annoying), and nothing to price-sensitive drop-offs (missed revenue).
Treatment abandonment: a patient who discontinues a prescribed care path without clinical resolution. Must distinguish from: completed care, referral to another provider, patient-initiated pause, and insurance-driven interruption.
A clinic reporting 'treatment abandonment' without distinguishing voluntary from systemic causes cannot determine whether the problem is clinical, administrative, or financial — and applies the wrong intervention to each.
Must be segmented by production stage. Order abandonment: a confirmed PO that is cancelled before production starts — distinguish between customer-initiated (demand change, budget cut), specification-driven (engineering change made the order obsolete), and capacity-driven (lead time exceeded customer's window). Production abandonment: a batch or run halted mid-process — distinguish between quality failure (out-of-spec material), equipment failure (unplanned downtime), and priority override (a higher-value order took the line). Each has different cost recovery paths and customer communication protocols.
A plant reporting 'order cancellation rate' as a single metric cannot determine whether the problem is commercial (sales overpromising lead times), engineering (specs changing too late), or operational (capacity planning failures). A 12% cancellation rate driven by lead time overruns requires different action than 12% driven by customer demand shifts. The first is fixable internally. The second is a market signal.
Trial abandonment: a user who signed up but did not reach the activation milestone within the trial period. Must distinguish between: never logged in (acquisition quality), logged in once (onboarding failure), used partially (feature gap or confusion), and used fully but did not convert (pricing or value mismatch).
Product teams reporting 'trial conversion rate' as a single metric cannot distinguish between a marketing problem (wrong users), an onboarding problem (right users, wrong experience), and a pricing problem (right users, right experience, wrong economics).
Pipeline abandonment: a qualified opportunity that exits the sales process without a decision. Must distinguish between: budget reallocation, stakeholder change, competitive loss, timing mismatch (need is real but not now), and vendor fatigue (too many rounds, lost momentum).
Sales leadership treating all closed-lost opportunities as competitive losses inflates competitive intelligence and misses the structural issues — procurement delays, internal politics, and process fatigue — that actually kill deals.
Conversion.
“A user who completed a desired action.”
Marketing reports 12% conversion. Product reports 8%. Finance reports 5%. All three are correct. They are measuring different things with the same word.
Visited → researched → added to cart → paid. 4 different conversions. Site-wide rate blends all four.
Landed → bought. Often a single-session event. Conversion is binary and fast.
Referred → booked → attended → started treatment → completed. The gap between 'attended' and 'started treatment' is where 40% of patients disappear.
Quote → order → scheduled → produced → shipped → accepted. Five different conversions. A high quote-to-order rate means nothing if 20% fail incoming inspection.
Signed up → activated → paid → expanded. A user who signs up but never activates is a cost, not a conversion.
Lead → qualified → proposal → negotiation → contract. A 6–18 month cycle where 'conversion' at each gate has different drivers.
Must specify: visit-to-cart, cart-to-checkout, checkout-to-payment, payment-to-fulfillment. Each is a different conversion with different drivers. 'Conversion rate' without specifying which boundary is meaningless.
A 4% site-wide conversion rate that blends all stages masks a 60% cart-to-checkout rate and a 7% visit-to-cart rate. The bottleneck is discovery, not checkout — but optimization spend goes to the payment page.
Referral-to-appointment, appointment-to-attendance, attendance-to-treatment-start, treatment-start-to-completion. Each boundary has different drivers: scheduling friction, insurance verification, clinical assessment, patient compliance.
A clinic measuring 'patient conversion' as referral-to-first-visit misses that 40% of patients who attend the first visit never start treatment. The real conversion failure is clinical, not administrative.
Quote-to-order, order-to-scheduled, scheduled-to-produced, produced-to-shipped, shipped-to-accepted. Each boundary has different drivers: quote conversion depends on pricing and lead time competitiveness; order-to-scheduled depends on capacity planning; produced-to-shipped depends on quality pass rate; shipped-to-accepted depends on whether the part meets the customer's incoming inspection. A 'conversion rate' that blends all stages is meaningless.
A manufacturer reporting 'quote-to-ship' conversion as a single metric cannot tell whether the bottleneck is commercial (losing on price at the quote stage), operational (capacity constraints delaying scheduling), quality (high rejection rates at final inspection), or logistics (shipping damage causing customer rejections). Each requires a different investment. The blended number hides all four.
Sign-up, activation (reaching the value moment), paid conversion, expansion (upgrading or adding seats). A user who signs up but never activates is not a conversion — it is a cost.
Reporting sign-up-to-paid conversion without tracking activation creates a funnel where marketing optimizes for sign-ups, product optimizes for activation, and sales optimizes for upgrades — all against different populations, none coordinated.
Lead-to-qualified, qualified-to-proposal, proposal-to-negotiation, negotiation-to-contract. Enterprise sales cycles spanning 6–18 months require conversion measured at each gate, not end-to-end.
A B2B company reporting 'lead-to-close' conversion as a single metric cannot tell whether the pipeline problem is at the top (lead quality), the middle (qualification process), or the bottom (pricing and negotiation). All three require different investments.
Active User.
“A user who has engaged with the product recently.”
The product team says there are 50,000 active users. The CEO reports this to the board. Nobody agreed on what 'active' means. Nobody asked.
Visited the site this month. But the purchase cycle is 3–5 years. A monthly visitor is a researcher, not an active buyer.
Ordered this week. If they skip two weeks, something is wrong.
Has an open care plan. A patient who has not visited in 18 months may be perfectly on schedule — or completely lost.
A production line with scheduled jobs. But '14 active lines' may include 3 running end-of-life parts and 2 with no orders in 8 months.
Logged in today. But a login is not usage. A password reset counts as a daily active user in most dashboards.
Has an open contract. But a 3-year auto-renewed contract with zero engagement is not 'active' — it is abandoned-but-billing.
Must distinguish: browsed (visited site), engaged (viewed product detail, used search/filter), transacted (completed purchase), and loyal (transacted 2+ times in the natural purchase cycle for the category).
Reporting monthly active users based on site visits inflates engagement metrics with bot traffic, price-checkers, and one-time visitors who will never buy. The number looks healthy. The business is not.
An active patient is one with an open care plan and a scheduled or recent interaction within the protocol window. A patient who has not visited in 18 months is not 'inactive' if their annual screening is not due for 6 more months.
Clinics that classify patients as inactive based on visit recency alone send recall campaigns to patients who are on schedule and miss patients who have genuinely disengaged. The outreach is noise; the silence is risk.
An active product line is one with current orders, scheduled production runs, and maintained tooling. Must distinguish from: dormant lines (tooling exists but no orders in 6+ months), legacy lines (still producing but for a single customer on life-of-program commitment), prototype lines (pre-production, not yet revenue-generating), and sunset lines (final production runs before discontinuation). An active machine is one with scheduled jobs — not just powered on. Active workforce includes qualified operators, not just headcount on payroll.
A plant reporting '14 active product lines' without distinguishing between full-rate production, legacy commitments, and dormant tooling overstates operational capacity. Three lines running for a single end-of-life customer and two with no orders in 8 months means the real active capacity is 9 lines. Quoting new work against 14 lines creates delivery commitments the plant cannot meet.
An active user is one who performed a core value action — not just logged in — within the measurement period. The core value action must be defined per product and tied to the outcome the user is paying for.
DAU/MAU ratios built on login events count password resets, accidental opens, and settings checks as 'activity'. The engagement metric is high. The product could be unused.
An active account is one with ongoing commercial engagement: open orders, active contracts, or recent procurement activity within the industry-standard cycle. A dormant account with a multi-year contract is not active — it is locked in.
Account managers maintaining 'active account' counts that include auto-renewed contracts with zero engagement cannot distinguish between satisfied retention and abandoned-but-billing accounts. Revenue looks stable until the contract ends.
Risk.
“Something that could go wrong.”
The compliance team, the clinical team, and the finance team all report 'risk levels' to the board. They are measuring three different phenomena with one word. The board hears one thing.
Fraud risk, chargeback risk, inventory risk, reputational risk. Four different teams, four different dashboards, one word.
Clinical risk: patient could die. Compliance risk: fine from regulator. Financial risk: insurance denies reimbursement. Same 'risk level' — completely different urgency.
Quality escape: immediate customer containment. Safety incident: 8-hour OSHA clock. Single-source supply risk: 6–18 months to qualify an alternative. Three amber alerts, three completely different timelines.
Churn risk: one account might leave. Concentration risk: three accounts are 40% of revenue. The churn model shows green. The business is structurally fragile.
Credit risk: they might not pay. Relationship risk: their procurement director just left. The credit score is solid. The relationship is over.
Fraud risk (transaction-level), chargeback risk (customer-level), inventory risk (SKU-level), and reputational risk (review and social signal). Each has different detection windows, thresholds, and response protocols.
A 'risk score' that blends fraud probability with inventory exposure produces a single number that is actionable for neither the fraud team nor the merchandising team. Both receive a signal. Neither can act on it.
Clinical risk (patient safety), operational risk (process failure), compliance risk (regulatory violation), financial risk (reimbursement denial). A high-risk patient is not a high-risk process. They require different interventions from different teams.
Hospitals that report 'risk' without domain qualification cannot prioritize: a clinical risk requires a physician, a compliance risk requires legal, a financial risk requires billing. The alert goes to all. Nobody owns it.
Quality risk (out-of-spec parts reaching the customer), safety risk (workplace incident or near-miss), supply chain risk (critical material shortage or single-source dependency), equipment risk (unplanned downtime on bottleneck machines), and compliance risk (regulatory audit finding, environmental violation). Each has different response protocols: a quality escape requires immediate customer notification and containment; a safety incident requires OSHA reporting within 8 hours; a single-source supply risk requires qualification of an alternate — which can take 6–18 months.
A manufacturer managing 'risk' through OEE dashboards alone sees equipment effectiveness but misses that a critical raw material comes from a single supplier in a geopolitically unstable region. The OEE is 87%. The supply chain is one disruption away from shutting down the line. A plant-level risk score that blends quality, safety, and supply risk into one number is actionable by nobody — quality needs PPM data, safety needs incident reports, procurement needs supplier diversification plans.
Churn risk (account-level), security risk (infrastructure-level), compliance risk (data handling), and concentration risk (revenue dependency on a small number of accounts).
A SaaS company that measures 'risk' as churn probability alone ignores that its three largest accounts represent 40% of ARR. The churn model shows green. The business is structurally fragile.
Credit risk (payment default), supply chain risk (delivery failure), contract risk (terms exposure), and relationship risk (key contact departure or organizational change at the client).
A B2B company that manages 'risk' through credit scoring alone misses that its largest account just replaced the procurement director who championed the relationship. The credit is solid. The relationship is over.
Before any system runs.
Before any automation fires.
These questions must be answered. If they are not, the automation runs on assumptions. The system produces numbers. The numbers look right. The definitions behind them were never checked.
What counts as a returning customer in this business?
What counts as abandonment in this purchase cycle?
When does an inactive period signal churn — and when is it a natural cycle?
What does 'conversion' mean at each stage of the funnel?
Which risks does this system measure — and which does it miss?
If the definitions are not governed, the system does not understand your business.
It just moves data.
Definitional authority
determines strategic control.
If your organization is making decisions on words it never formally defined — the consequences are already compounding.