The short version: Feature lists make every platform look similar and comparison tables make the wrong things look decisive. A scorecard fixes both, provided you do one thing first: write down your weights before you look at any candidate. Ten criteria, weighted to your business, scored 0 to 3, and the winner is arithmetic rather than impression. The weighting matters more than the scoring, because a platform that is excellent at everything you do not need still loses. And there are two automatic disqualifiers worth applying before you score anything: it cannot express your collection schedule, and it cannot cap what you can actually make.
Because comparison tables are built by whoever is selling, and they weight everything equally.
A feature table gives the same visual space to "abandoned cart recovery" and "per-location cutoffs," when for a market vendor one is decorative and the other decides whether the platform works at all. Reading down two columns, you end up counting ticks rather than weighing what matters.
A scorecard forces two things a table does not:
That second point is the whole discipline. Requirements written after you have seen the options are a description of what you liked, not a specification.
The scorecard exists because feature tables reward whoever lists the most rows, and your business only needs a handful of them to work. Fill it in before opening our cottage food platform comparison, not after.
Drawn from what actually goes wrong for small food businesses, rather than from any vendor's feature list.
Ten is deliberate. Fewer misses something structural; more turns into a table again.
Assign each criterion a weight from 0 to 3 based on your business, before looking at any platform.
Two worked examples, to show how different the same ten criteria look:
A baker at three markets, 40 orders a week:
A holiday-only gift-box maker, three months a year, shipping nationally:
Same criteria, almost opposite answers. The first vendor should ignore any platform that cannot do per-location cutoffs. The second should ignore any platform that charges in months they do not trade.
Score each platform 0 to 3 on each criterion, then multiply by your weight.
Multiply, add, compare. A criterion weighted 3 and scored 0 contributes nothing, which is the point: a platform can score well overall and still be unusable.
Two rules that keep the scoring honest:
Score what you tested, not what the page claims. A feature name is not a capability. "Sales tax autopilot" tells you nothing about whether it files.
Score your hardest case. The product with variants, a lead time, and a per-day cap. Simple products work everywhere and separate nothing.
Two, and they apply before you score anything, because a weighted total can hide a fatal gap.
It cannot express your collection schedule. If you sell at more than one place or on more than one day, and the platform supports only a single global cutoff, then every week you are either turning down orders you could have taken or losing prep time. No amount of scoring elsewhere compensates.
It cannot cap what you can actually make. A platform that lets forty-seven people order forty items has not saved you time; it has created a morning of refunds. Test this directly: set a cap of two and try to order three.
If a candidate fails either, stop scoring it. That is not a low score, it is a different category of problem.
Four things that feel important in a demo and rarely decide anything.
Storefront design. Your product photographs do the work. A beautiful template with poor photographs converts worse than a plain one with good ones.
The number of features. Most are for businesses unlike yours. Cococart's eight modules describe a café; Shopify's app ecosystem describes a national brand.
A processing discount on a higher tier. Shopify Advanced needs about $810,000 in annual sales to justify its rate saving; Square Plus needs roughly $12,250 a month. Buy tiers for features, never for rates.
Subscription price alone. At $12,000 a year in sales, processing is roughly four times the subscription, so choosing on the headline optimises about 19% of the bill.
Our guides to profit margin benchmarks and tracking income and expenses cover where these costs sit against everything else, which is usually lower than the attention they get.
Three that get underweighted and cost real hours.
The pick list. You will use it every single week, in a hurry, often with wet hands. A platform that gives you a flat list of orders sorted by customer name means you re-sort it weekly, which puts a spreadsheet straight back into your process.
Manual order entry. If any orders arrive by text or phone and cannot be entered, you keep a second list. Two lists means two stock counts and a pick list that is always incomplete.
Data portability. Boring until the day it matters. Ask what exports, in what format, and whether email consent status is included, because without it you have addresses rather than a mailing list.
Six steps, and the order is what makes it work.
Step three is where most comparisons go wrong. Five shallow evaluations produce less information than two deep ones, and the platform that wins a shallow comparison is usually the one with the best marketing site.
Step four is the only part that generates real information. Load your actual catalog, set your real cutoffs, place a real order, and produce a real pick list. Everything else is reading.
About a week of elapsed time and roughly four hours of actual work, which is worth stating because people either rush it or let it run for months.
The failure mode at one end is deciding from a demo in an afternoon, which produces a choice based on whichever storefront looked nicest. At the other end is a comparison that runs for three months while you keep taking orders in messages, which costs more in admin hours than any platform difference on the scorecard.
A week is enough. If you cannot decide after a week of real testing, the candidates are close enough that the choice does not matter much, and starting a trial on the one you found easier beats another fortnight of reading.
Our guides to getting repeat customers and calculating the real cost per item cover two things worth considerably more of your attention than the tenth hour of platform comparison.
Using the three-market baker's weights, on a fictional pair:
| Criterion | Weight | Platform A score | A total | Platform B score | B total |
|---|---|---|---|---|---|
| Collection scheduling | 3 | 3 | 9 | 1 | 3 |
| Quantity caps | 3 | 3 | 9 | 2 | 6 |
| Pick list | 3 | 2 | 6 | 1 | 3 |
| Manual order entry | 2 | 3 | 6 | 3 | 6 |
| All-in cost | 2 | 2 | 4 | 3 | 6 |
| Support | 2 | 2 | 4 | 3 | 6 |
| Fee structure fit | 1 | 3 | 3 | 3 | 3 |
| Sales tax | 1 | 3 | 3 | 1 | 1 |
| Data portability | 1 | 2 | 2 | 3 | 3 |
| Setup effort | 1 | 2 | 2 | 3 | 3 |
| Total | 48 | 40 |
Platform B is cheaper, better supported, faster to set up, and loses, because the three things weighted 3 are the three it does worst. That is the scorecard working: it stops a platform winning on the criteria that were easiest to evaluate.
Then the decision genuinely does not matter much, and you should stop optimising.
Within about 10%, pick the one you found easier to use during the trial. At typical small-vendor volume every credible platform lands within roughly $50 a year of the others once processing is counted, so a near-tie on the scorecard usually means a near-tie in reality.
What you should not do is add criteria until one wins. That is not analysis, it is justification, and it is how vendors end up switching again in eighteen months.
The SBA's guidance on managing your finances is a reasonable framework for tracking the cost side properly once you have chosen, and the FTC's privacy and data security guidance covers obligations that follow you to whichever platform wins.
If your weights come out heavy on collection scheduling, caps, and the pick list, Homegrown is $10 a month billed annually with 0% commission and 2.9% plus $0.30 processing published up front, and it handles pickup at each place you sell with its own schedule and cutoff, local delivery with a radius and a route, and sales tax calculated, filed, and remitted in all 50 states. The honest bounds, which are real scorecard zeros for some businesses: no point-of-sale, no national shipping, no drop countdowns, no app ecosystem, and it is not a website builder. If any of those carry a weight of 3 for you, score it accordingly. You can trial it with your hardest product and score from what you tested rather than from this page.
Because tables weight everything equally and are built by whoever is selling. A scorecard makes you set your own weights, which means your business decides the answer rather than a vendor's feature list.
Before looking at any platform. Weights written afterwards describe what you liked rather than what you need, and that is how vendors end up choosing on storefront design.
A platform that cannot give each collection point its own day and cutoff, and one that cannot cap quantities the way you bake. Both are structural, and a good total score elsewhere does not compensate.
Two, properly. Five shallow evaluations produce less information than two deep ones, and the winner of a shallow comparison is usually just the one with the best marketing site.
Storefront design, feature count, processing discounts on higher tiers, and subscription price alone. At $12,000 a year in sales, processing is about four times the subscription.
The pick list, manual order entry, and data portability. You use the first every week in a hurry, the second decides whether you keep a second list, and the third only matters on the day it matters a great deal.
Pick the one you found easier during the trial and stop. At typical volumes the credible options are within about $50 a year of each other, so a near-tie on paper is a near-tie in practice.
The scoring is the easy part. The weighting is the decision, and it has to be written down before you look at anything, or it will quietly reshape itself around whichever platform you started to prefer.
Apply the two disqualifiers first: can it express your collection schedule, and can it cap what you can actually make? A platform failing either is not a low score, it is a different problem, and no total elsewhere fixes it.
Then weight the boring things properly. The pick list, manual order entry, and data portability are the three that get underweighted and cost the most, while storefront design and feature count are the two that feel decisive in a demo and almost never are.
