How Long Should You Fund a Side Project That Barely Makes Money? FreshBooks' Agency-Funded Runway
Short answer: Fund a weak-earning side project only while the subsidy is capped, the customer evidence is improving and the main business can absorb the cost without missing payroll or tax. Set a 90-day review, a maximum cash-and-time subsidy and a stop condition before the next cycle begins. Continue when paid retention, usage or efficient acquisition improves, not merely because total users rise.
Early revenue can be too small to support a product while still revealing a viable future business. It can also be a flattering distraction financed by profitable work elsewhere.
The answer is not “give it more time”. Time is not the scarce resource. Cash, founder attention and foregone agency contribution are.
What FreshBooks' slow start can and cannot prove
FreshBooks began after Mike McDerment built an invoicing system to replace a spreadsheet and developed it alongside his design agency. In a direct founder interview, McDerment describes a period with ten customers paying US$9.95 each, roughly US$100 in monthly revenue, while service work paid the bills. He also told Mixergy that the service operation supported the product and that the team worked from his parents' basement for years.
Published recollections conflict on the exact early calendar, so it would be false precision to declare a particular month the decisive one. A later interview about product reinvention supports the wider mechanism: customer contact and repeated product changes mattered more than an instant launch result. A Forbes account puts FreshBooks near 1,000 paying customers by mid-2006.
This case does not establish that persistence causes success. Survivorship bias is obvious. It shows how a profitable service business can buy product learning, provided the subsidy remains visible.
My position is stricter than the usual “keep consulting until software takes off” advice. If you cannot state what the next 90 days must prove, stop subsidising. Unpriced patience is not strategy.
Use the Side-Project Subsidy Audit
The Side-Project Subsidy Audit separates four things founders often combine: cash spent, labour diverted, evidence gained and risk imposed on the funding business.
| Audit line | What to record monthly | Continue signal | Stop signal | |---|---|---|---| | Cash subsidy | Hosting, contractors, refunds and tools | Within a fixed ceiling | Ceiling breached twice | | Time subsidy | Founder and staff hours at replacement cost | Hours per active customer fall | Support rises faster than revenue | | Customer evidence | Paid activation, use and retention | Cohorts improve | Sign-ups rise but use does not | | Agency damage | Missed work, margin and client delays | Core commitments remain protected | Payroll, tax or clients are exposed | | Opportunity cost | Contribution from work declined | Product evidence exceeds alternative value | Better work is repeatedly displaced |
Complete it using money and hours, not labels such as “promising”. The guide to [numbers every owner should track](/business-finance/numbers-every-business-owner-should-track/) provides the operating measures around cash and contribution.
Price the whole subsidy
Start with cash leaving the account. Add the replacement cost of founder and employee hours. Then add contribution lost when product work displaces billable work.
Suppose an agency developer spends 25 unbilled hours on the product. Their salary allocation may be £650, but the relevant opportunity cost might be £2,000 of contribution the agency would have earned from available client work. Do not count both if one already includes the other. Choose the higher economically relevant amount and document the method.
Also distinguish idle capacity from diverted capacity. Using six otherwise idle hours has a different cost from delaying a signed client project. “The staff are already paid” is only defensible when the time genuinely has no better near-term use.
Demand evidence that compounds
Revenue is one signal, not the only one. For a subscription product, compare cohorts rather than cumulative accounts. Record how many paid, completed the first valuable task, remained after 30 and 90 days, and required support.
An increase from ten to 18 paying customers means little if half leave next month. Ten customers who use the product weekly, refer two peers and require progressively less help can justify another cycle even when they do not cover salaries.
Choose one primary learning measure for the cycle. It might be 90-day paid retention, invoices sent per active account or the percentage of trials completing setup. Do not replace it mid-cycle because another number looks better.
Protect the business providing the runway
Ring-fence payroll, tax, committed supplier payments and at least one month's core operating expenses before funding a product. Then set the maximum transfer in cash and hours.
Client quality matters too. If product work makes the agency late, its future referrals and pricing power can decline before the accounts show a crisis. Create a service rule, such as no product work before contracted milestones are delivered and no more than 15% of team capacity without explicit review.
Tax treatment, employment rules and how costs can be allocated between entities vary by country and structure. Ask a qualified local accountant before transferring money or staff time between companies, and document the arrangement.
Worked example: LedgerNest Invoicing
LedgerNest Invoicing is funded by a three-person web agency. It has 32 paying accounts at £14 a month, so monthly recurring revenue is 32 x £14 = £448. These figures are illustrative.
| Monthly product item | Calculation | Amount | |---|---:|---:| | Hosting and services | fixed | £210 | | Founder time | 24 hours x £35 | £840 | | Developer time diverted from available client work | 18 hours x £28 | £504 | | Payment fees and refunds | fixed | £34 | | Total product cost | | £1,588 |
The monthly economic subsidy is £1,588 minus £448 = £1,140. That is not the same as cash burn. Hosting, services, payment fees and refunds use £244 of incremental cash, while founder time and developer capacity diverted from available client work account for £1,344 of the cost. Product receipts cover the incremental cash items, but the product remains economically loss-making because those hours could serve paying agency work. If the developer hours were genuinely idle with no near-term alternative use, they should not be valued this way.
The agency has £18,000 cash after ring-fencing £12,000 for payroll, tax and committed bills. It authorises no more than £6,000 of economic subsidy, equal to £6,000 divided by £1,140 = 5.26 months at the present rate. This is a capacity-allocation limit, not a claim that the bank account will fall by £1,140 every month.
LedgerNest sets a 90-day cycle capped at £3,420 of economic subsidy, while separately tracking the £732 of direct cash service cost. Its pass conditions are 50 paying accounts, at least 80% 90-day retention and founder support below 30 minutes per active account monthly. At 90 days it has 47 accounts, 87% retention and 18 minutes of support. It misses the account target but improves the harder quality measures.
The company funds one more cycle, but caps acquisition tests at £600 and requires £900 MRR. If retention falls below 75% or the agency delays a client milestone, the cycle ends. That is persistence with boundaries, not faith.
Set stop, continue and spin-out conditions
Write all three before the review date.
Continue when the chosen customer measure improves and the next constraint is identifiable. Stop when two consecutive cycles miss the same evidence threshold, the subsidy ceiling is breached or the funding business is damaged. Spin the project into dedicated capacity only when its contribution can cover the new fixed commitment with a buffer.
A working guide is to require forecast gross profit to cover at least 1.25 times the salary and fixed costs being moved. That margin absorbs forecast error. Check the underlying assumptions using [whether the business can make money](/business-finance/can-this-business-make-money/), not revenue alone.
Make the next 90-day decision
This week, reconstruct the last three months of product cash, time and foregone contribution. Ring-fence the funding business's non-negotiable obligations. Choose one customer-quality measure and record its current value.
By day 7, write the maximum subsidy, review date and three conditions. Review the score every Friday but do not move the thresholds. On day 90, choose stop, another capped cycle or dedicated investment. If the records cannot show what the subsidy bought, pause the project until they can.
Related guides
Frequently asked questions
Is £100 of monthly revenue meaningful evidence?
Yes, if it comes from unrelated customers who use the product and keep paying. It proves more than a large free-user count because money introduces a real trade-off. However, £100 does not demonstrate a viable business, price ceiling or acquisition system. Inspect who paid, what they completed, how they arrived and whether they remained. Ten friends paying after personal persuasion are different from ten strangers found through a repeatable channel. Treat the revenue as permission to run another bounded test, not as permission to fund the project indefinitely.
How much of my agency's time can the product use?
Set the limit from spare contribution and delivery risk, not a generic percentage. If the agency has signed work using 90% of practical capacity, even 10% may threaten deadlines. If demand is seasonal, documented idle hours may be used more safely. Start by protecting client milestones, payroll administration and sales activity. Then cap named people's product hours for the next month. Price those hours at replacement or opportunity cost. Review client delivery weekly, because the first sign of overuse is often slower work rather than a cash shortage.
Should I charge early customers even if the product is rough?
Usually yes. A modest price tests whether the problem matters enough to fund and creates clearer expectations about value. Offer a narrow outcome, disclose limitations and provide a straightforward cancellation route. Do not charge for capabilities that do not work or describe a manual service as automated. For consumer subscriptions, cancellation, renewal, data and refund rules vary by jurisdiction. Check the requirements where customers live and obtain qualified legal advice for your specific offer. A paid pilot can be time-limited without pretending it is a finished product.
What if the side project helps win agency clients?
Count that benefit only when you can trace it. Record leads that mention the product, proposals influenced and contribution from work won. Do not assign all agency growth to the side project because it appears in sales conversations. The product may function as marketing even if it never becomes independent, which can justify a different budget and objective. Decide which business it is serving. If its role is lead generation, compare its cost with other acquisition routes. If its role is a future standalone company, retain separate customer and unit-economics tests.
When should I hire specifically for the product?
Hire when recurring gross profit or committed financing can cover the full employment cost with a buffer and the work is stable enough to define. Salary is only part of the cost: include employer charges, equipment, management and recruitment. As a working guide, require 1.25 times coverage for at least three months, unless you have intentionally allocated investment capital. Employment obligations vary by country, and moving an agency employee into another entity may require consultation or new terms. Obtain qualified local employment and tax advice before making the change.
Can I keep funding it if customer numbers are flat?
Only when another preselected quality measure is improving enough to change the economics. Flat customer numbers can hide lower churn, higher usage, less support or a better price, all of which may justify a short additional cycle. They can also hide stagnation. Calculate net customer movement and cohort retention, then identify the bottleneck the next experiment will address. If numbers are flat for two cycles and you cannot name new evidence, stop the subsidy. Continuing because substantial time has already been spent is a sunk-cost decision, not a forecast.
Comments (0)