If your revenue arrives through partners, marketplaces, and resellers, closing the books is hard for one reason: the money comes in a dozen streams and has nowhere to land.
At a glance
- Multi-channel businesses don’t struggle to reconcile because their accounting is weak. They struggle because every channel reports the same sale differently and pays on its own schedule.
- In most of these businesses, the place where it all comes together is one person and a spreadsheet. That works until the month that person is out.
- Five rules turn scattered streams into one place you own, a data lake. Once it exists, accounting, forecasting, partner statements, and AI can all draw from it.
It’s the fifth business day of the month. Someone in finance has three numbers for the same period: what customers paid, what the sales channels say they sent, and what actually landed in the bank. None of them match.
Nobody made a mistake. One partner adds its fee on top of your price; another takes its cut out of it. One pays when the customer buys, one pays weekly after the work is delivered, one pays monthly, and one has never told you its schedule. A refund from last month shows up as a negative line this month. Every stream is telling the truth, each in its own language.
So the person reconciling translates. They download, paste, retype, and write the owner an email explaining the gap. Next month, they do it again.
This is one of the most common problems we see in growing businesses, and one of the least discussed. It rarely gets named, because from the inside it looks like a bookkeeping chore. It isn’t. It’s a missing piece of infrastructure, the one engineers call a data lake.
Your revenue has tributaries
Think of each way you get paid as a tributary. Marketplaces. Resellers and distributors. Retailers that take deductions before they pay. Franchise networks. Booking and ticketing platforms. Agents and partners paid on commission. Payment processors that hold your money for a few days before releasing it.
Each stream carries water from the same source, your customers, but each arrives with its own character:
| The stream tells you | What you actually need |
|---|---|
| The price the customer saw | What the partner actually paid you |
| A “payout,” “remittance,” “settlement,” or “net” | One definition of revenue across every channel |
| Its own payment schedule | When the money was earned vs. when it arrived |
| Cancellations and refunds, sometimes weeks later, sometimes never | Which month each change belongs to |
| Its own fee logic, on top of or out of your price | Your true margin by channel |
None of this is exotic. It’s what selling through other people looks like, in almost any industry. The trouble starts when there’s no single place for all of it to flow into.
The person reconciling is the lake
Most companies assume their accounting system is where everything comes together. It isn’t. Accounting software is built to record consequences: what you earned, what you owe, what you hold for someone else. It isn’t built to collect thousands of transactions from a dozen sources that disagree and work out which version is right.
So the work lands on a spreadsheet, and the spreadsheet is really just a bucket someone carries by hand every month. The actual lake is a person. It’s their memory of which partner reports cancellations late, which column holds the real number, and which gap is “normal.”
That arrangement has three costs. When the person is out, the month doesn’t close. When they leave, the knowledge leaves with them. And no reporting tool or AI system can read a lake that lives in someone’s head.
What a data lake actually is
A data lake is a central place where a business keeps the raw records from all of its sources, in their original form, so they can be cleaned, combined, and analyzed later. Every partner report, every processor statement, every refund file lands there as it was received, with nothing thrown away.
It’s often confused with a data warehouse, and the difference matters. A warehouse stores data that has already been cleaned and shaped for specific reports. It’s fast and orderly, but it only holds what someone decided in advance to keep. A lake holds everything first and imposes structure afterward, which is exactly what you need when a dozen sources describe the same sale a dozen different ways and you don’t yet know which version is right.
The term usually comes up in enterprise technology conversations, which makes it sound bigger and more expensive than it has to be. For most growing businesses, the useful version is modest: keep the original record from every stream, build one clean, agreed-upon layer on top of it, and connect that layer to the tools you already use. You don’t need a platform migration or a data team. You need a place the streams can land, and rules for what happens when they do.
The five rules of the lake
1. Map every tributary
Before writing any software, write down how each stream works: how it prices, how it takes its fee, what it calls the payout, when it pays, and how it reports changes. It’s unglamorous work, and it pays for itself right away. The map almost always surfaces problems nobody knew existed: a partner whose cancellations never reach your main system, a feed that quietly stopped arriving, a fee that’s been calculated the wrong way for a year.
2. Trust whoever pays you
Most businesses pick one system as “the source of truth.” For multi-channel revenue, that’s the wrong frame. The source of truth is set per stream, and for money that comes through a partner, it’s the partner’s record of what it paid you, not the price your own system recorded at the time of sale. Those two numbers are rarely the same, and the difference is where most unexplained gaps come from. Record what arrives, not what was advertised.
3. Let the accountant define the events, then prove them on real data
Engineers shouldn’t invent accounting rules, and accountants shouldn’t have to guess how data behaves. The best results come from letting each check the other.
A common first instinct is to run everything through a single holding account. A good accountant will push back: money a payment processor is holding and money you haven’t earned yet are two different things that clear on different schedules, so they belong in different places. Then the real data usually pushes back on the accountant’s model in turn. Partners that collect payment up front and partners that pay after delivery rarely fit the same pattern, and forcing them into one creates the very gaps you were trying to close.
Let the accountant own the rules, and let the data test them. The payoff is that the timing gap between earned and received stops being a mystery and becomes a stated balance on the books.
4. Keep the detail in the lake and send summaries downstream
Your accounting system doesn’t need every transaction. It needs accurate totals, organized the way your accountant reports, with the detail one click away. Summarizing turns thousands of entries a month into dozens, each with its supporting records attached and a unique reference so nothing can post twice. And closed months stay closed: a refund that arrives late becomes a correction in the current month, never an edit to a period someone has already signed off.
5. Read it back
Three habits keep a lake honest. First, silence isn’t zero: a partner that sends no data hasn’t canceled everything, so missing data has to be treated as missing, not as a number. Second, admit what can’t be pulled. Some figures, like card processing fees or a vendor’s own payouts, often live only on a statement that no system exposes. Give them a deliberate place to be entered instead of pretending to automate them. Third, check the destination’s own reports, not your own math. A system that reports “success” isn’t proof. The books agreeing is.
Once you have a lake, everything can draw from it
Accounting is only the first pipe. Once every stream lands in one place you own, the same data can feed forecasting, partner statements, commission payouts, dashboards, and eventually AI. This is where “AI readiness” actually lives. Not in choosing a model, but in giving it one trustworthy body of data to read.
When you don’t need a lake
To be fair, plenty of businesses don’t need one. If you sell through one or two channels, get paid at the point of sale, and your accounting software’s built-in integrations handle the rest, keep it simple. If your volume is low enough that a monthly spreadsheet costs less than a build, the spreadsheet is the right answer. And if your channels change every quarter, do the mapping first and build later.
The signs you’ve outgrown the bucket: closing the month takes more than a few days, only one person can explain the numbers, partner disputes keep coming up, or nobody can answer “what did we actually make from that channel last month?” without a week of work.
Five questions to ask your team on Monday
- How many streams pay us, and on what schedule?
- For each stream, where does the real payout number live, and what is it called?
- Who reconciles today, and what happens to month-end if they’re out for two weeks?
- Could our accountant write down our money events on one page?
- What can’t we pull automatically, and where does it get entered now?
Most businesses in this position ask which tool will finally fix their reporting, or which AI will make sense of their numbers. That’s the second question. The first is simpler: where do the streams land?