Over the past several years I have seen more than a few enterprise software fundraising decks claim to have completed the shift from project work to product economics, and I have watched a fair number of them come apart the moment someone actually asks about revenue structure. Pull apart what gets called product subscription revenue, and it is often still a custom build, a custom implementation and a custom quote for every single customer, with a one-time license fee simply relabeled as an annual subscription — the economics have not changed at all. This is not a semantic issue. It is one of the easiest things to overlook in this industry, and one of the easiest stories for capital markets to buy, right up until a renewal rate misses expectations one year and the story falls apart for real.
This piece is about what that shift actually looks like when it is real, and how far it has gotten across China's enterprise software market. The direction is not in question — moving away from an economics built on custom implementation and one-time fees, toward one built on a standardized product and predictable recurring revenue, follows the logic of software as an industry, and is why capital markets pay a higher multiple for it. But a correct direction does not make the path easy. Most companies are stuck in the middle — neither having shed project cost structure nor built real product predictability — and that middle state is worth pulling apart carefully.
What actually differs between the two economic models
Ask what actually differs between project economics and subscription economics, and the first answer people usually give is billing — one-time versus annual. That reading is too shallow, shallow enough to misjudge which stage a company is genuinely at. The real difference sits in the cost structure. Under project economics, marginal delivery cost barely falls as the customer count rises — every new signing requires another round of implementation staff learning that customer's specific process, doing custom development, deploying on site. Under subscription economics, marginal delivery cost is supposed to keep falling as customers are added, because the product itself is standardized and onboarding a new customer leans on self-service configuration rather than an engineer rewriting code line by line for each one.
To judge which model a company is actually in, I do not look at its marketing language. I look at two plainer numbers: whether the implementation team's headcount scales roughly in proportion with the customer count, and whether the average time from signing to a new customer going live is getting shorter. If the customer base doubles and the implementation team roughly doubles with it, the company is still a project business in its cost structure, whatever the contract calls the fee.
Why mid-market vertical depth travels further than horizontal ambition
Most founders, early on, get drawn to a horizontal story — build one general-purpose module, in theory applicable to enterprise customers in any industry, with a market size that sounds unlimited. I have watched too many companies burn several years and several funding rounds down that path, only to arrive at an uncomfortable fact: a general-purpose module can sign customers well enough, but renewal willingness tends to run low, because the product never really embeds itself in that company's core process, switching cost stays low, and once a competitor undercuts on price, churn moves faster than anyone expected.
Mid-market vertical depth follows an entirely different path: pick one specific sub-sector first, refine the product until it genuinely understands that sector's particular process and compliance requirements, then gradually extend into adjacent sector segments rather than chasing every industry from day one. It sounds like a narrower market on paper. Once it is actually working, though, customer stickiness and renewal rates run noticeably higher than a horizontal product's, because the product is embedded in that sector's specific workflow and the switching cost is real rather than theoretical. The price of going deep vertically is that early growth looks slower, which is also why capital markets routinely underrate it — a growth curve that is not steep enough gets misread as a ceiling that is not high enough.
- Depth of understanding in industry-specific process: whether the product can handle the genuinely awkward exceptions in that sector, not just the standard eight-tenths of a workflow while the remaining two-tenths is patched manually forever.
- How deeply compliance and regulatory requirements are embedded: these requirements change often, and if compliance logic needs manual upkeep per customer, marginal cost never really falls. Only embedding that logic into the product's own architecture counts as genuine productization.
- The path for expanding into adjacent segments: going vertical does not mean serving one sub-sector forever. It means building a real moat in one sector first, then judging which adjacent segments can reuse the existing product capability. The pace runs slower than a horizontal instinct would suggest, but the reuse rate ends up markedly higher.
Retention: the one number that does not lie
Among the external metrics an enterprise software company reports, new bookings, customer count and the number of industries covered can all be dressed up to look good, and none of them is especially reliable. New bookings are heavily shaped by a sales team pushing volume in the short term; customer count can be padded with discounting or even free trials; industry coverage often just means one or two flagship customers were signed and the industry gets declared entered. The metric that is hardest to dress up, and that says the most about whether a product has genuinely embedded itself in a customer's business, is net revenue retention — whether revenue from the existing customer base, after accounting for churn, is expanding or shrinking.
Retention tells the truth because it reflects whether a customer, with no sales team re-persuading them, is willing to keep paying — or even willing to pay more to use additional modules. A company can report impressive new bookings and still be running net revenue retention that sits at a mediocre level year after year, or slipping. When that happens, it is nearly certain the product has not genuinely embedded itself in the customer's core process, and renewals are happening out of sunk cost or inertia rather than because the customer cannot do without the system. When assessing a company like this, I will largely set new bookings aside, look first at the three-year retention trend, and only then decide whether the rest of the numbers are worth a closer look.
- Product genuinely embedded in core process
- Renewal driven mainly by inertia and sunk cost
Product genuinely embedded in core process — Year 1: 104, Year 2: 108, Year 3: 112, Year 4: 115; Renewal driven mainly by inertia and sunk cost — Year 1: 98, Year 2: 94, Year 3: 89, Year 4: 85
Illustrative framework showing the explanatory power of retention, not the actual data of any specific company
Where this shift stands, and where it has stalled
Break this shift down by specific functional area, and the fastest-moving segments are the ones where the workflow itself is already relatively standardized with limited variation across companies — financial accounting, parts of human-resources management. The business logic in these areas differs little from one company to the next, productization is comparatively easier, and subscription economics has, as a result, largely taken hold here already. Leading companies' net revenue retention in these areas genuinely reflects rising product stickiness.
The segments where the shift has stalled most visibly are the ones where the underlying process is highly individualized and tied directly to a company's own competitive edge — production scheduling, supply-chain coordination, parts of sales-process management. The problem here is not entirely a matter of vendor capability. Customer-side process genuinely varies a great deal — one manufacturer's scheduling logic can look nothing like another's — and finding a balance between real standardization and still supporting that variation is hard on its own terms. We stay cautious on these segments, not because the long-run direction is in doubt, but because the claim of having already productized is especially prone to being untrue here, and needs a longer observation window to verify.
Yunjie Software: a sample walking this path
Yunjie Software is one specific window we watch this shift through — it provides operations management software to mid-market manufacturers and has chosen the vertical-depth path rather than chasing enterprise customers of every size and every industry. In assessing this company, more time goes to whether its implementation cycle within manufacturing sub-segments keeps shortening, and whether existing customers are voluntarily expanding into new modules, than to how fast the new-customer count is growing. That is the consistent method here: however well a new-customer story is told, it ultimately has to be verified by whether existing customers keep paying, and keep paying more.
This shift will not happen evenly
The direction of the shift from project to subscription is clear, but it will not happen across the whole market at the same pace, and it does not complete just because one sentence in a fundraising deck says so. How fast a given functional area moves depends on how standardized the underlying business process already is; how genuine a given company's shift is depends on whether it has honestly brought down marginal delivery cost and truly embedded the product in the customer's core process, rather than simply relabeling how it charges. The most reliable way to see where this shift actually stands is not to listen to how a company describes itself, but to look at two or three plain numbers — whether implementation cost is really falling with scale, whether retention is really improving. These numbers rarely lie, and they are where I start when judging where value in this market ultimately settles.