A software founder can answer every question about the product and still have no clear answer to a simpler one: how will the next ten suitable customers find it? The application works. The price seems reasonable. Early users have been helpful. Yet each new customer arrives through a different favor, introduction or accident. That is evidence of possibility, but a difficult basis for a business budget.
Distribution becomes a system when the path from discovery to sustained customer value can be described, observed and repeated within acceptable costs. For a validated app, the job is to choose a reachable segment, make a credible offer in a channel that segment uses, remove obstacles between interest and value, and measure the contribution from customers who remain. The repeatable unit is the whole path, including support and renewal.
This chapter develops that operating system. The earlier distribution test asks whether customers can be reached before significant development begins. Here the question changes: once a worthwhile problem and some willingness to pay have been established, how should the founder turn occasional sales into a maintained acquisition process? Both stages matter. Their evidence, spending decisions and failure modes differ.
Begin with a customer you can recognize
“Small businesses” is too broad to guide a distribution decision. It includes buyers with different problems, budgets, software habits and decision rights. Even “online merchants” can conceal a damaging mismatch. A shop with forty products and one supplier may find a problem irritating; a shop with several thousand products and frequent supplier changes may treat the same problem as a weekly operating expense. Their willingness to investigate a tool will differ.
Specify the starting segment in observable terms. What work does it do? What event makes the problem urgent? Who experiences the failure? Who can buy a remedy? What information can that person lawfully share during evaluation? Where does the person already seek help? These answers do more than improve an advertisement. They determine which prospects belong in the denominator when conversion is calculated.
Imagine a hypothetical application that flags supplier price changes before they damage a merchant’s margin. The starting segment might be merchants who regularly import supplier catalogs, maintain their own selling prices, and already review mismatches manually. A merchant whose supplier handles all pricing belongs to a different segment. Counting both as equally qualified prospects would make the first segment appear less interested than it is and could lead the founder to change the product unnecessarily.
The Small Business Administration’s market-research guidance asks founders to consider demand, market size, location, saturation and alternative prices. It also distinguishes broad secondary information from direct research with the intended audience. That is a useful foundation for segmentation, although the guidance does not establish demand for this hypothetical application. The segment still needs evidence from its own behavior. SBA market research guidance.
A useful segment definition has an exclusion rule. Write down whom the first offer will not serve, then explain why. Excluding a buyer who needs an unbuilt integration can protect both the buyer and the experiment. It lets the founder learn whether the currently deliverable result has a market without quietly mixing delivery promises into the sales pitch.
Map the path the buyer actually follows
A channel is a place where contact occurs. Distribution includes everything that must happen after contact. A marketplace listing, article, partner referral or search advertisement can introduce a product, but introduction alone cannot overcome an incomprehensible offer, an unsafe data request or an onboarding flow that requires half a day of preparation.
For the hypothetical margin application, the path could be: a merchant encounters a relevant explanation; recognizes a familiar catalog problem; checks compatibility; submits a small authorized sample; sees a credible flagged discrepancy; understands what correcting it would involve; buys a plan; connects the regular workflow; and receives useful results again the following month. Each transition has an owner, an obstacle and a measurable event.
Draw that path before buying reach. Then ask which transition is presently uncertain. If people read the offer but misunderstand which catalogs are supported, more traffic may simply produce more support questions. If they understand the offer but cannot safely prepare data, the missing element might be a sample-file guide or a narrower integration. If they receive a good first result and do not return, investigate the recurrence of the job before sending reminder emails.
These distinctions keep acquisition from becoming a catch-all explanation for every weak metric. Low sales can arise from poor reach, weak qualification, distrust, friction, missing value, unsuitable pricing or an offer arriving at the wrong time. The observation “few people bought” does not tell the founder which mechanism failed.
Start with an event dictionary that a second person could interpret. “Visitor” might mean a unique browser during a calendar month. “Qualified inquiry” might mean a person meeting the stated segment criteria who requests an evaluation. “Activated” might mean the first valid supplier mismatch reviewed by the merchant. “Paid” should specify whether it counts an initial charge, a completed trial or a retained subscription. Ambiguous definitions make the same dashboard tell different stories to different people.
Choose a channel for a reason you can test
Channels differ in access, intent, cost, credibility and control. A reader searching for a specific integration error is closer to an immediate problem than a person browsing broad software news. An agency referral can begin with trust but carry a revenue share and a service obligation. A marketplace can place the app near a relevant buyer while imposing approval, billing, support and listing requirements.
The choice should begin with a hypothesis about customer behavior. For example: merchants confronting supplier import errors consult platform-specific help pages; therefore a useful worked explanation of that problem may attract suitable evaluations. Or: agencies repeatedly encounter the problem while maintaining client catalogs; therefore a supported referral arrangement may reach merchants at a timely moment. Neither sentence establishes that the channel works. Each explains what observation would make the test informative.
Shopify’s current App Store documentation illustrates both the attraction and the boundary of marketplace distribution. Public listing information supports discovery through browsing, recommendations and Sidekick. Distribution requires submission to the approval process and meeting applicable requirements. The documentation describes a channel and its rules; it cannot guarantee that a particular app will receive qualified installs or profitable customers. Shopify App Store documentation.
Compare channels using the buyer’s job rather than the founder’s comfort. Writing articles may feel familiar, but some urgent operational problems are resolved through agencies or platform support communities. Conversely, cold outreach may be available while creating little trust for a product that asks for access to sensitive business information. Familiarity is a useful execution advantage only when it overlaps with a plausible buyer path.
The founder should also distinguish borrowed reach from owned relationships. A marketplace listing is subject to platform decisions. A useful email relationship can be more direct, provided it is built and maintained with appropriate permission. Search visibility depends on external systems and competitors. Channel diversity can reduce dependence, but premature expansion can spread a small operator across several processes before any one of them works.
Make the offer small enough to believe
An early offer should promise a result the product can reliably deliver for the selected segment. “Protect your revenue with AI” is broad enough to hide several untested assumptions. “Review supplier price changes in this supported catalog format before your next price update” tells the merchant what work is addressed, when the output matters and where the boundary lies.
That narrower promise gives the distribution experiment a cleaner outcome. The buyer can determine whether the problem is relevant. The founder can check whether the delivered result matches the offer. Support can explain compatibility without inventing custom features during a sales conversation. If the product later expands, the promise can expand with verified capability.
Credibility often comes from showing the mechanism. A sample report can display the supplier price, the merchant’s current price, the calculation and the timestamp. The visitor learns what the system will inspect and what it will return. This is especially valuable when the result will influence a financial or operational decision. A polished interface without traceable reasoning can leave the buyer unsure whether the output deserves attention.
Use customer language carefully. Describing a recurring frustration in words customers use can improve recognition, but quotations require permission and accurate context. An invented testimonial is not an acceptable shortcut to trust. A hypothetical example should remain labeled as an example even when it resembles the intended customer’s daily work. The customer-language chapter examines how to preserve that distinction while improving copy.
The offer also needs an honest next step. If the application is a pilot, say what participation includes, what it costs, what support is available and what happens when the pilot ends. A button marked “Start saving” should not conceal a consulting intake that cannot produce the promised result without several days of manual work. Honest expectations reduce avoidable friction after the sale.
Use the first test to learn the expensive unknown
An experiment should answer a consequential question. If the uncertainty is whether the target merchant will entrust a sample catalog to the product, a broad campaign measuring clicks will miss it. If the uncertainty is whether agencies can introduce qualified buyers, the relevant observation is a completed, suitable introduction and its downstream outcome, not the number of agencies that reply politely.
Define the test before the results arrive. State the segment, channel, offer, spending limit, duration, success condition and stopping condition. Include a failure condition that would change the decision. A founder who cannot describe a disappointing outcome in advance can often explain away any disappointing outcome afterward.
A hypothetical four-week test might seek a small set of qualified merchant evaluations through a tightly relevant article and agency referrals. The purpose could be to learn whether merchants can prepare supported data and identify a useful discrepancy without intensive help. The budget would include writing, partner conversations, diagnostic delivery and support time. Those numbers would be planning assumptions; they would not represent a proven industry threshold.
Keep the comparison clean enough to interpret. If the product, price, segment and channel all change together, a better result cannot readily be attributed to any one change. Small samples do not justify elaborate causal claims, but stable conditions still make observations more useful. Record what changed and why. If an unexpected problem requires a mid-test correction, preserve the earlier results rather than quietly blending them into the revised offer.
Include a counterexample in the review. Which suitable prospect declined despite recognizing the problem? Which buyer used the product successfully but did not find the result valuable? Which agency considered the referral unattractive because it conflicted with its existing service? These cases can reveal a boundary more efficiently than another enthusiastic response from someone already favorably disposed.
Calculate acquisition cost with the work included
Customer acquisition cost is often discussed as if it were simply advertisement spend divided by new customers. That calculation can be useful for a campaign, but it can hide the operator’s time and the labor required to move a prospect through evaluation. A founder who spends ten hours helping each buyer connect a catalog has created a delivery obligation even when the campaign itself cost little.
Define which costs belong in the measurement and keep that definition stable. A practical channel view might include campaign spending, content production, attributed partner commissions, prospect-specific sales effort and evaluation work. Broader company overhead can remain separate, provided it is not forgotten when assessing total profitability. The important distinction is visible scope rather than one supposedly perfect label.
Consider an explicitly hypothetical test. The founder spends $300 on a targeted campaign and $200 on attributable design and writing work. Five merchants become paying customers. The narrow campaign acquisition cost is $100 per paying customer. If sales and setup assistance take another fifteen hours valued at a planning rate of $30 per hour, the expanded cost is $950, or $190 per customer. Neither number is false; they answer different questions.
Now consider what those customers contribute. Suppose a customer pays $40 per month and requires $12 of variable delivery cost. Monthly contribution before acquisition and fixed expenses is $28. Recovering a $190 acquisition cost would require roughly 6.8 months at that contribution, assuming continued payment and unchanged delivery cost. That assumption is substantial. Early cancellations, refunds, usage spikes or increased support can lengthen or prevent recovery.
The contribution-margin chapter develops the cost categories in detail. For distribution, the implication is immediate: a channel can appear efficient at acquiring initial payments while consuming more value than its customers subsequently generate. Measure the customer path beyond the first charge before turning a trial budget into an ongoing commitment.
Avoid treating monthly recurring revenue as if it were cash available to fund acquisition. Stripe’s documented MRR definition uses monthly-normalized active and past-due subscriptions, excludes certain categories including metered products, and allows discount-related reporting settings. It is a useful provider-defined subscription metric, but it is neither cash collection nor profit. Stripe subscription analytics.
Read cohorts before reading averages
A cohort groups customers by a shared starting event, such as the month of first payment or the channel that acquired them. It helps the founder ask whether a distribution process produces customers who continue realizing value. A total subscriber count can rise while each new cohort leaves faster than the last.
For a small app, cohort analysis can begin in a simple table. Record first-payment date, segment, source, activation date, initial support effort, subsequent usage appropriate to the job, renewal, cancellation and refund. The purpose is to preserve the relationship between acquisition promises and later customer outcomes. It does not require a complex analytics platform or a dashboard full of percentages with unstable denominators.
Suppose, hypothetically, agency referrals yield fewer customers than a broad campaign but those merchants connect supported catalogs more successfully and require less setup help. The difference might arise because agencies qualify buyers and prepare data. It might also arise because the referred merchants are larger or have more urgent problems. The table reveals a pattern worth investigating; it does not prove the referral channel caused better retention.
Segment differences matter when comparing channels. If one channel attracts occasional users and another reaches operators with a weekly job, raw monthly activity will differ even if both products serve their intended purpose. Define useful recurrence according to the customer’s work. A seasonal tax tool should not be judged by the return pattern expected from a daily inventory application.
Small denominators should remain visible. “Retention improved from fifty to seventy-five percent” sounds impressive, but the interpretation changes if it describes two retained customers out of four followed by three out of four. Write the counts beside the rates. Let the evidence support the scale of the claim rather than letting the percentage supply unwarranted confidence.
Attribute enough to decide without pretending to know everything
Attribution asks which interaction receives credit for a sale. A customer might read an article, ask an agency, compare a marketplace listing, leave, return through search and then buy. Any single-touch attribution rule simplifies that path. Simplification can be acceptable when it is explicit and useful; it becomes misleading when the dashboard presents the assigned source as the full cause.
Use a consistent rule and add the customer’s own account where practical. Record first known source, last meaningful interaction before purchase and any partner referral. A short optional question about how the buyer heard of the product can fill gaps, although memory and response bias limit it. The goal is enough understanding to allocate the next test budget sensibly.
Privacy and practicality constrain instrumentation. Do not collect unnecessary personal details just to produce a more elaborate acquisition diagram. If a business must work with aggregated data or incomplete identifiers, report the uncertainty. An honest incomplete map is more useful than a precise-looking explanation built from assumptions that cannot be checked.
Look for decisions robust to attribution uncertainty. If a channel produces no qualified evaluations despite a bounded test, perfect attribution will not rescue it. If several customers explicitly identify one article as helpful and use its example during onboarding, the founder has evidence that the article contributes to the path even without claiming it caused every sale. Spend more cautiously when the conclusion depends on the most favorable attribution rule.
Turn a working channel into an operating routine
A channel becomes repeatable when someone can run it again with comparable inputs, quality and cost. Record the target audience, offer, preparation steps, source materials, publishing or referral process, follow-up boundary and measurement schedule. Preserve the version of the product and promise used during the successful test. Otherwise later results may be compared with an experiment whose conditions no one remembers.
Automation can help with repetitive preparation, scheduling and reporting. It should be introduced where the process is understood. Automating an uncertain offer can spread the uncertainty faster. Automating unqualified outreach can produce more replies and more support work without advancing useful demand. Human judgment remains important when interpreting why a prospect’s situation does or does not fit.
Create a small operating record for each channel. Include a responsible owner, time allowance, current cost assumptions, known risks and the next review date. Describe what would trigger a pause: rising support effort, a platform rule change, a decline in qualified activation or a mismatch between the advertised promise and deliverable product. A review date keeps yesterday’s successful tactic from becoming an indefinite expense by inertia.
The routine should include the customer-facing handoff. Who answers a technical question? Who owns a failed setup? What response can a partner promise? How does a trial become a paid subscription? Distribution that stops at purchase leaves these responsibilities to emerge as surprises. The agency-partnership chapter develops the accountability needed when another business introduces the customer.
Scale the constraint that is actually binding
Once a channel produces suitable customers with acceptable contribution, the founder may increase spending. The next constraint often appears elsewhere. A successful article can generate more evaluations than the operator can review. An agency partnership can accelerate introductions while making customization requests harder to refuse. A marketplace promotion can increase installs from buyers outside the original supported segment.
Scale in increments that preserve visibility. Compare new customers with the earlier cohort and watch the delivery burden. If support time rises sharply, investigate whether qualification weakened, onboarding failed or the product attracted a broader job. More sales do not automatically justify broader promises. Sometimes the best response to increased demand is a clearer eligibility boundary.
Capacity deserves an explicit calculation. If a founder can responsibly devote six hours per week to evaluations and each evaluation requires forty-five minutes, eight evaluations consume the allowance before any exceptional issue arises. This is an illustrative capacity calculation, not a claim about a typical app. It makes the operational consequence visible before acquisition spending outruns delivery.
Some constraints call for product work. A repeated data-format failure might warrant an importer. Others call for better distribution design. A repeated mismatch between prospect expectations and the product might warrant clearer copy. Treat every bottleneck as a question about mechanism; adding code to a qualification problem can create a feature burden without resolving the original confusion.
Know when a channel has stopped teaching you
A bounded experiment needs an ending. Continuing until something encouraging happens quietly converts a test into an open-ended commitment. Review the results at the stated date even when the outcome is disappointing. Distinguish an informative failure from a test that never reached its intended audience. Those outcomes imply different next steps.
An informative failure might show that suitable merchants recognize the problem but prefer their existing spreadsheet because the new product cannot accommodate an important exception. That evidence can support a narrowly scoped product revision or a decision to leave the segment alone. An uninformative test might show that an article attracted readers interested in starting an online store rather than merchants already managing supplier catalogs. The offer never reached enough appropriate buyers to examine willingness to pay.
Do not erase either result. Record the original hypothesis, execution, observations and limitation. A channel can fail for the current segment and remain plausible for another; the test establishes its scope. Conversely, a campaign that produces purchases during a brief promotion does not establish routine demand at the regular price. Preserve the promotional terms beside the cohort, especially when discounting changes the buyer’s expectations or the support burden.
Seasonality complicates the review. A catalog-management problem may become urgent before a major supplier update, while another product has demand tied to annual reporting. Comparing adjacent months without acknowledging the job’s calendar can misread a temporary surge as a repeatable acquisition improvement. Ask what external event changed alongside the campaign. Record the event even if the available sample cannot quantify its effect.
Stopping can also preserve useful assets. A failed channel may leave a clearer explanation, a compatibility guide or a consented set of questions that improves onboarding elsewhere. Those benefits should be recorded separately from sales. They do not make an uneconomic campaign profitable, but they can change what work needs to be repeated in the next experiment.
The review should end with a specific decision: repeat under comparable conditions, revise one stated element, narrow the audience, wait for a relevant buying event, or stop. Assign a budget and an owner to the next action. “Keep trying marketing” leaves the most important questions unresolved and makes it difficult to learn from the next month’s spending.
What Would We Do at Salars?
For Salars, the proposed software venture engine would begin with one narrowly defined customer problem and one credible path to that customer. Supplier Margin Guard and Merchant Revenue Guard are candidate product ideas in this series, not established businesses with measured acquisition or retention results. Their names cannot stand in for evidence that merchants want the proposed workflow.
A first distribution record could connect a useful Salars article, an explicitly labeled diagnostic pilot and a supported purchase path. The article would explain a real mechanism using transparent illustrative data. The pilot would request only the information needed to inspect the selected problem. The purchase offer would identify supported formats, delivery expectations and the limits of any automated recommendation.
The operating ledger would record qualified inquiries, consented evaluations, first useful results, paid conversions, refunds, retained use and attributable support time. It would also retain the cases that did not fit. Those cases would inform whether to narrow the segment, revise the offer, improve delivery or stop the test. No universal conversion target should be copied from an unrelated business and called proof.
Agency referrals could be tested after the result and support boundary are understandable. Marketplace distribution could follow if the chosen platform, data permissions and product requirements fit the app. Those choices would remain contingent on current rules and actual customer observations. The storefront should publish a promise that the maintained product can fulfill, rather than allowing distribution ambitions to dictate unsupported capability claims.
The decision to scale would require more than a favorable first-payment count. We would want a repeatable route to suitable buyers, evidence that those buyers receive recurring value, and a contribution calculation that includes the human work required to keep the system running. A modest channel meeting those conditions deserves more attention than a spectacular traffic chart that cannot explain what happens after the click.
The next ten customers should arrive through a path the operator understands well enough to improve. That understanding is the first durable product of a distribution system.
Explore the complete AI Software Factory series and the wider AI section.
Sources
- SBA, Plan your business: market segmentation, research methods and business-planning guidance.
- Shopify, About the App Store: listing discovery, approval and platform responsibilities.
- Stripe, Subscription analytics: provider-specific MRR definitions and reporting choices.
Official passages were checked on October 7, 2026. Merchant scenarios, budgets, conversion counts and capacity calculations are explicitly hypothetical. The proposed Salars products have no measured business results established in this article.
Loading comments…