Founder Digital Hygiene Is a Procurement Habit

The habit that moves the number is the buying habit

A subscription bills on its own schedule whether or not anyone opens the tool, so the decision that fixes a startup’s monthly software outflow was made at checkout, several months before any audit finds it. Advice about founder digital hygiene almost always describes what happens after that point: close the tabs, batch the notifications, review the subscriptions once a quarter. None of it is wrong, and none of it touches the mechanism that sets the number.

What makes that decision expensive is not carelessness. It is that vendor billing design encodes a commitment the buyer did not consciously agree to. The unit the bill is priced on decides how fast it grows. The basis under the headline figure decides whether you have bought a month or a year. The seat minimum decides the floor. And the exit terms, where they exist at all, decide what the mistake costs to correct.

Below are four habits, each with the vendor behaviour that makes it necessary, the count from our own catalogue, and the check that answers it. Every figure is computed from the 102 software profiles we maintain, by the script at scripts/compute-procurement-stats.mjs, and re-derived on 19 August 2026 after a round of pricing corrections. Every count is reported over the number of profiles where the field it depends on was actually readable, never over a flat 102. Three denominators appear: 102 profiles in total, 92 carrying at least one plan with a real price, and 90 publishing an entry price that names a figure. Billing basis is undefined for a product with no paid plan, which is why the first two differ.

Habit one: price the unit, not the plan

The vendor behaviour. A pricing page presents tiers, because tiers are what a vendor wants compared. The tier is not what you are buying. You are buying a rate against a unit, and the unit is usually one line of small text under the number. Per-seat pricing, also written as per user or per member, charges the stated rate for every person with an account, so the bill tracks headcount rather than how much anyone uses the tool. Usage pricing charges by consumption instead, and a growing number of products do both at once.

What our catalogue shows. Of the 92 products with a paid plan:

Billing basis	Count	Share of 92	Read from
Per seat (per user, seat, member, agent, editor, licence)	58	63.0%	Recorded field
Usage as the pricing model itself	8	8.7%	Recorded field
Per seat plus a separate metered line	7	7.6%	Recorded field
Per non-seat unit (per host, per channel)	2	2.2%	Recorded field
Volume-priced by contacts, subscribers or sends	2	2.2%	Read from prose
No billing unit recorded	25	27.2%	Residual

The overlap between those categories is the finding. On GitHub, Team is $4 per user/month while Actions minutes meter at $0.008 per minute and storage at $0.25 per GB, so headcount and build volume grow the bill independently. Intercom is Essential at $29 per seat/month billed annually plus from $0.99 per Fin resolution, which is why its own profile records “variable per-resolution cost makes budgeting harder” as a con. The GitHub and Intercom figures were verified on the vendors’ pricing pages on 9 August 2026.

Three things about that table are worth knowing before anyone quotes it. The rows do not sum to 95 and are not meant to, because a product can sit in two categories and two more - Omnisend and WooCommerce

• carry a metered line without a recorded seat unit and so appear in no row at all. The “Read from” column matters as much as the counts: five of the six categories come from fields a researcher filled in deliberately, while the volume-priced count of 5 is inferred from the wording of plan notes and is the one figure we would expect to move under a stricter reading. And the 24 products with no billing unit recorded are a residual rather than a claim about flat pricing - two are demonstrably in the wrong place, since Metabase charges $6 per user/month above its first five and Framer charges per published site, but in both cases the vendor’s wording sits in a plan note rather than the structured field the count reads. The per-seat count of 56 and the non-seat-unit count of 2 are floors, not ceilings.

The check. Write down the unit, then write down what the bill is at three times your current size on that unit. If a tool charges on two units, do it twice. A tool priced on seats and a tool priced on volume can diverge sharply between signing and the second renewal.

Habit two: price the floor, not the entry

The vendor behaviour. A per-seat price is quoted as a single-seat figure even where a single seat cannot be bought. The minimum lives in a footnote, and it is a multiplier on the number you just anchored to.

What our catalogue shows. Six of the 58 per-seat products publish a seat minimum. That is a small count and worth stating as small, but where it applies the effect is not:

Product	Advertised	Minimum	Real floor
Box Business Starter	$5 per user/month billed annually	3 users	$15/month
Adobe Express Acrobat Express for Teams	$11.99 per licence/month	2 licences	$23.98/month
monday.com Basic	$9 per seat/month billed annually	3 seats	$27/month
Dropbox Standard	$15 per user/month billed annually	3 users	$45/month
SurveyMonkey Team Advantage	$30 per user/month billed annually	3 users	$90/month
Apollo.io Organization	$119 per user/month	3 users	$357/month

Box is the row that shows why the floor and the advertised price are different questions. Business Starter is the cheapest per-seat figure on Box’s page and cannot be bought at all below three users, while Personal Pro sells to one person for $10 - so the cheapest way to actually buy Box is the plan that looks twice as expensive. Until recently this site ranked Box on the $5.

Apollo.io’s own profile records “credit limits and seat minimums complicate the real cost” among its cons, and monday.com’s says plainly that it is not for teams of one or two. These figures were read from the vendors’ pricing pages, the SurveyMonkey and Adobe Express rows on 18 August 2026 and the rest on 9 August. Apollo.io is the one row whose profile records no annual or monthly split, so its $119 is stated as the vendor states it.

The check. Multiply the per-seat rate by the minimum before comparing anything. For a founder or a pair, a three-seat minimum is a threefold price increase disguised as a footnote, and it changes which tool wins the comparison. Our own profiles now record the minimum as a structured field and do that multiplication on the page, because we were getting it wrong too: three of the six rows above were positioned on their advertised per-seat price until we fixed it.

Habit three: “billed annually” is the price of an option

The vendor behaviour. “Billed annually” under a monthly-looking price means that figure is only charged at that rate if you commit to twelve months and pay for them upfront; choosing to pay month-to-month instead raises the rate, often sharply. A pricing page showing “$12 per user/month billed annually” is quoting $144 a year per user, not $12 a month with an exit each month. The discount for committing is advertised. The premium for not committing usually is not.

What our catalogue shows. The commitment is advertised far more often than the way out of it:

Of the 90 products publishing an entry price	Count
Record an annual billing basis somewhere in their published tiers	59
Of those 59, also state a month-to-month rate anywhere we could read it	20
Of those 59, state the annual commitment but no monthly alternative	39

The first row says “somewhere in their published tiers” rather than “on the entry price” deliberately. The count is of vendors whose published pricing assumes annual payment at some point, not of entry prices specifically: several publish an entry tier with no annual basis recorded and introduce one further up the ladder.

Where both rates are recoverable - 37 plans across 17 products - the premium for paying monthly is:

Statistic	Premium for month-to-month billing (37 plans, 17 products)
Median	25.3%
Mean	28.3%
Range	11.1% to 58.3%

ClickUp Business sits at the top of that range: $12 per user/month billed annually against $19 paying monthly, a 58.3% premium. Box’s Business Starter is $5 per user/month billed annually against $7 billed monthly, 40%. Typeform charges the premium across its whole ladder rather than on one tier - Basic $28 against $39, Plus $56 against $79, Growth Flow $266 against $379 - which is 28% to 43% for the same right to stop. At the other end, Shopify Advanced is $359/month billed annually against $399 billed monthly, 11.1%. The Typeform figures were read on 18 August 2026; the ClickUp, Box and Shopify figures on 9 August.

Our article on reading a SaaS pricing page measures the same mechanism from the other side, reporting a median annual discount of 17% across 43 plan pairs. The two figures are not the same number inverted, and it is worth saying why rather than letting them sit on the site looking contradictory. They use different samples, because the pricing-page article admits any pair we could compare while this measurement uses only pairs where the vendor states a true month-to-month rate. They also divide by different bases, since a discount is measured against the monthly price and a premium against the annual one. Neither is wrong. They answer different questions.

The framing is the reason to compute the second one. A discount reads as money saved; a premium reads as the price of an option, which is what it is. You are being quoted a fee to retain the right to stop paying, and it is worth paying when the tool’s chance of not lasting the year is high enough to beat the premium. Where exactly that lands depends on when in the year you would leave, which is why this is a way of framing the decision rather than a formula.

The check. Find the month-to-month rate. If the pricing page does not show one, toggle the billing switch and read the number again, because on 39 of the 59 products that record an annual basis, no month-to-month rate is stated in the copy. Then ask whether you would still buy at that rate. If the answer is no, you have not decided to commit for a year - you have been priced into it.

Habit four: find the exit before you sign

The vendor behaviour. Signup is a two-click flow. Cancellation terms are in a subscription agreement that is linked, not shown.

What our catalogue shows. This is the starkest count in the piece. Of the 102 products we track, exactly one publishes contract terms specific enough to record: Adobe Creative Cloud, whose subscription terms state a 12-month term, a 14-day refund window, and a cancellation fee of 50% of the remaining balance thereafter. Seven products of the 102 have enough published detail for us to record switching costs at all.

One honest qualification, because the number invites over-reading. Contract terms are the one place our usual distinction does not hold: the field defaults to empty, so a profile where we looked and found nothing published looks identical to one where the question was never asked. The same is true of the switching field. So the reportable figure is that one vendor publishes terms we could record - and that is a statement about our research as much as about the vendors. We are not entitled to the stronger claim that 101 vendors withhold their terms, and we are not making it.

What the count does support is narrower and still worth having: across 102 products researched by people specifically looking for commercial terms, exit conditions surfaced once. Whatever share of that is vendor opacity and whatever share is our own coverage, a founder who expects to find the number on the page they are buying from will usually be disappointed.

The check. Before entering a card, find the notice period, the auto-renewal behaviour and whether any part of a prepayment is refundable. If the pricing page cannot answer all three, record that as the answer. An undisclosed exit is information about the vendor, and it is the single most useful thing this check produces.

Three mechanisms that move software spend and this catalogue cannot count

Three mechanisms that plainly do move a founder’s software spend are not measurable from this catalogue at any denominator, and saying so is more useful than estimating them.

• Whether a free trial requires a card. Not a field in our schema. We record a trial on 65 of 102 products, a trial length on 62, and 37 as not disclosed - but card requirement appears in only four profiles, and in two of those four it refers to card processing fees rather than to signing up. Two usable mentions is not a statistic.
• What happens when a trial ends. Whether it lapses or auto-converts is not recorded either.
• Credit expiry and rollover. Also not recorded, despite credits appearing in the pricing of several products above.

Each is a real mechanism and each is a gap in our data rather than an absence in the world. Treat the card question as a manual check rather than a statistic.

The limitation that bounds everything above. This collection is skewed towards vendors that publish pricing at all - only six of 102 are quote-only, and those six are enterprise suites: NetSuite, Qualtrics, Rippling, SAP Concur, ServiceNow and Workday. Three of the six give us some evidence of what sits behind a private quote. Qualtrics we record as running annual negotiated contracts with limited flexibility, Workday as pricing per worker per year on multi-year agreements, and NetSuite as negotiating firmly at renewal. Rippling records nothing of the kind, and none of the six gives us a minimum spend.

Three of six is thin, and it is the wrong direction of thin: these are the vendors whose terms are hardest to see, which is why there are only six of them. If commitment is more common where pricing is private, then every share reported here is an undercount and the annual-basis figure most of all. We cannot size that from data we do not have.

Every profile records the starting price with its billing basis and the date it was checked

Every software profile here carries the starting price with its billing basis, the plan ladder with the unit each tier is priced on, the seat minimum where one exists, and the date each was checked. Where a vendor does not publish something we record it as not disclosed rather than estimating it, which is why the counts above can be reported with their denominators instead of as round percentages. The method behind this article is a script over those profiles, and it can be re-run as the catalogue grows. How the verification works is set out in our methodology.