We use the four-level model from Arwin Scholte’s Four Depths of Partnership [link to Part 2, as amended]. It fits on a napkin and it changes your quarter, so here it is in full.

The rule

A customer’s life with your product has three segments: sell, deploy, run. A partner covers a segment when it does work there that contributes to the customer’s outcome and bills for it, or saves you from carrying the cost.

Depth is how many of the three the partner covers, for that customer.

• L1. One segment. Usually sell: referrals, registered leads, pipeline. Sometimes run alone: a partner that inherited the account and keeps it working. One cell filled.

• L2. Two segments. Sell and deploy is the common shape: the partner brings the deal and is inside the implementation. Sell and run is the other, and it is more common than most books admit: the accountant who recommends a payments product and then runs it for the client every month, with nobody from the vendor involved after sign-up.

• L3. All three. The partner brought the customer, deployed them and is still there, billing, after go-live. This is the level the model is built around, because it is the first one where the partner is in the customer’s success rather than just their acquisition. Last letter’s CSI is how you see it: L3 partners leave fingerprints on the health of their customers.

• L4. All three, and then more. A service your product is sold inside, resell, a certified training practice, a global footprint. The extension does not count on its own; a global reseller that sells and deploys but does not run is L2 with an extension, not L4.

The segments are not ordered. Nothing says a partner must sell before it deploys or deploy before it runs. The three columns are the same for every partner; what differs is which cells are filled.

Depth is a count, not a badge

Here is where most partner-tier programmes go wrong: the tier is a field someone sets. Gold, silver, bronze, assigned in an admin screen, reviewed annually, negotiated at renewal. It describes the contract, not the relationship.

Depth, as we compute it, is read off what the partner actually does. Three questions per partner per customer, all of which you already have the data for if you measured the first three letters honestly:

1. Did the partner sell it? Is the opportunity’s referrer field set to them?

2. Did the partner deploy it? Were they on the plan, in the meetings, billing for the work?

3. Does the partner run it? Is there recurring billing after go-live: a managed service, retained time, support seats, a shared channel they staff?

Count the yeses. One, two, three. No committee, no negotiation. When the work changes, the count changes with it, in both directions.

Two consequences of counting instead of assigning. First, it cannot be gamed in a QBR. Second, every level tells you exactly which segment is empty, so the next step is obvious. An L1 with strong pipeline is one deploy engagement away from L2. An L2 that sells and deploys needs a reason to stay after go-live and something to bill for it. An L2 that sells and runs is already in the part of the lifecycle that matters; the question is whether deploy is a segment your product even has. The model is a coaching tool before it is a scoreboard.

Where health fits

Customer health is not a fourth segment and it is not what makes a partner L3. It is the evidence that a covered segment is doing something. A partner that bills for run across a book of customers whose health sits below your median is still L3 by count; the flag beside the count is the conversation. Next letter covers how that flag is computed honestly, because it is the place most tools quietly give up.

The partner that stops at L1 and is right to

Some partners cannot climb, and that is the model working. A referral partner with no services line has nothing to bill for at deploy or run; the count stops at one. Treating them as stuck is the category error. Their job is better routing, not deeper delivery.

So record what each partner is, rather than inferring it from silence: referral, implementation, advisory, managed services. A referral partner at L1 is at its ceiling. A managed-services partner at L1 is a problem. Same number, different meaning, and the only way the model can tell them apart is if you wrote it down.

Why one deeper partner beats ten new logos

Signing ten new partners gives you ten L1s: ten sources of referrals, zero post-sale coverage, and a partner-management workload that grew tenfold. Moving one existing partner from L1 to L3 gives you something none of those ten can: a partner inside the customer’s run segment, which is where retention and expansion live. It shows up in all three letters at once. CDE, because a partner that stays refers better-fitting deals. CSI, because they are working the lifecycle. ECS, because a partner covering run is covering more of the needs register.

Logo counting grows the top of your funnel. Depth grows everything below it.

Try it on your own book

You can run this in a spreadsheet this afternoon. One row per partner per customer, three columns (sell, deploy, run), a yes or no in each, and a fourth column for what the partner bills on a recurring basis. Be strict about the third column: billing, not fondness. Then a fifth column for what the partner is: referral, implementation, advisory, managed services.

Most books come out heavily L1, with one or two L2s that sell and run which nobody had named, and the run column full of partners the customer would mention and your systems do not hold. Then pick one partner with an empty segment it could fill, and spend the quarter you would have spent recruiting on filling it.

Start the 5-Customer Audit

If you would rather have the count derived continuously from live pipeline, engagement and billing data, with the empty segment surfaced per relationship, that is what Entwine does. The free workspace monitors five relationships, no credit card:

Start with the needs register. The rest follows from it.

The three-metric letters (CDE, CSI, ECS) plus this one complete the foundation. From here: field notes on running an ecosystem with these numbers on the wall.

SUBTITLE: The Ecosystem Completeness Score (ECS): how much of what your customers need can your ecosystem deliver, and where the gaps are. Part 3 of the ecosystem-metrics series.

BYLINE: Sion Welchart, Director of Marketing, Entwine

---

The first two metrics in this series ask how well your ecosystem handles the customers you already have. This one asks something you may not want the answer to. Of everything your customers have asked you for, how much has anyone actually delivered?

Most scores in this category are built to reassure. This one is not, and I would not trust it if it were.

Start from the need, not the partner

Most ecosystem measurement starts at the partner list: who signed, who sourced, who renewed. ECS starts at the other end, with a register of what your customers have said they need.

A need here is concrete. A named customer asked for a named capability, with a priority and a target date. Not a market hypothesis. Not a segment.

Each need resolves over time, met or not, and the score is built from that register in two parts. Fulfilment rate is the backbone: of everything your customers have asked for, what share has actually been met? Alignment quality is the refinement: of the needs that were met, how well matched was the partner who met them? Meeting a need is good. Meeting it with the right partner, at the right depth, is better.

The two combine into a score from 0 to 10, weighted toward fulfilment by default and configurable per company. Same principle as CSI last week. The platform computes, you define what counts.

The gap list is the product

The number is the least useful thing ECS produces.

What matters is what falls out of computing it. Every unmet need becomes a row in a gap analysis: which capability, for which customer, at what priority, overdue or not. Sort that list by priority and date and you are holding something genuinely rare in partnership work. It is a partner recruitment shortlist, written by your own customers.

High-priority unmet needs are your next partner conversations, already qualified, with somebody waiting on the other side of each one.

Overdue needs are churn risk wearing a different name. A customer who asked for something months ago and is still waiting has started looking elsewhere, whether or not they have told you.

Five customers missing the same capability is not five problems. It is one build-or-partner decision that you can now put a date and a customer list against.

That is why we call it a completeness score rather than a coverage score. Coverage sounds like a market-sizing exercise. Completeness is personal: these specific customers, these specific asks, this much of it handled.

Almost nobody has a needs register, and that is the point

I should be straight about the prerequisite. Very few companies keep a register of customer needs. The asks are scattered across QBR notes, support threads, a product board and somebody’s inbox, which is precisely why nobody in the business can answer the question.

So do not start with the score. Start with ten accounts and one question in your next round of reviews: what have you asked us for that you are still waiting on? Write down the capability, how badly they want it, and when they wanted it by. Ten rows in a spreadsheet is a working register. It takes an afternoon, and in my experience the first overdue row on it is the one you wish you had found a quarter earlier.

ECS will not flatter an empty register either. No recorded needs does not produce a triumphant 10 out of 10 on the grounds that nobody asked for anything. An empty register means the discovery work has not been done, and the score’s breakdown says exactly that rather than rewarding the silence.

The three together

CDE asks whether your ecosystem gets customers from need to solution faster than you manage alone.

CSI asks whether those customers then succeed across their lifecycle.

ECS asks whether your ecosystem can deliver what they need next.

A partnership programme that can answer all three from its own records has stopped counting logos and started managing an ecosystem. One that can answer none of them is running on anecdotes, however good the slides look.

Where the numbers come from

Needs reach the register three ways: typed in by your team, imported from wherever you already track requests, or submitted by the customer directly through their own portal view. This register is the only thing in the whole series that asks you to collect something new, which is why I would start it by hand this quarter whatever tooling you end up using.

Run the five-customer audit

Same offer as the last two letters, and this time the gap list is the output. Five accounts, one question each, and you will have the beginnings of a register plus your first overdue row. The template is free and ungated:

Take the 5-Customer Audit

If you would rather see all three scores side by side for your own book, computed as the data arrives, the free Entwine workspace monitors five relationships and takes no card:

Start with the needs register. The rest follows from it.

This concludes the three-part series on the ecosystem metrics. Next: the four depths of partnership, and why moving one partner from transactional to strategic beats signing ten new logos.

This concludes the three-part series on the ecosystem metrics. Next up: the four depths of partnership, and why moving one partner from transactional to strategic beats signing ten new logos.

Speed was the easy one. This week is the claim that gets budget approved and almost never gets tested: that partners do not only bring you customers, they make those customers succeed.

I wanted a score that could prove that claim wrong. A metric that can only flatter you is decoration.

What CSI measures

CSI is a weighted average across five components of a customer’s health, each scored 0 to 10.

Product adoption. Is the customer actually using what they bought?

Engagement quality. Are the delivery engagements around this customer healthy, moving and completing?

Advocacy. Has the customer become a reference, a case study, a referrer? It is the hardest component to fake, because nobody advocates for a product that is failing them.

Business outcomes. The commercial trail: renewals, expansion, realised value.

Customer sentiment. What the customer says when you ask, folded in from survey responses and support experience.

The weights are yours to set. A product-led company loads adoption. A services-heavy one loads engagement quality. Our defaults are sensible, and they are defaults. Success is defined by you in configuration, not hardcoded by us. That is a recurring theme in this series: the platform computes, you define what counts.

The question I would ask any vendor in this category, us included

What does your score do when a component has no data?

Say you run no surveys, so there is no sentiment signal at all. Most scoring systems substitute a neutral value, a 5 out of 10, and roll it in. The result looks complete and is partly fabricated. The score in your quarterly review now contains a digit nobody measured.

CSI does not do that. A component with no data is excluded, and the weights renormalise over what is actually present. Run no surveys and your CSI is a real average of the four components you do measure, rather than four real numbers and one invented one.

There is also a floor. If too little of the weighted picture is present, the score is withheld entirely. Not zeroed, not defaulted, withheld, and it arrives with a list of exactly which inputs are missing. A blank that tells you why it is blank is more useful than a number that cannot tell you anything.

This is the design decision I would defend hardest. If a metric is going to sit in a QBR and move budget between partners, every digit in it has to trace back to something that happened.

Where partners come into it

CSI scores a customer. Its purpose is a partner question: which partners are attached to customers that succeed?

Line your partners up against the CSI of the customers they touch and the four depths become visible in your own data. A transactional partner appears at acquisition and disappears, so the health of that customer is whatever your own team makes of it. An implementation partner lifts early adoption and then hands over. A success-level partner leaves fingerprints across adoption, engagement, outcomes and eventually advocacy. Seeing that difference per partner, in data rather than asserted on a slide, is what lets you invest in depth instead of counting logos.

It is also what keeps last week’s metric honest. A high CDE with a low CSI means your ecosystem finds deals quickly and hands over customers who then struggle. You want both numbers side by side before you call any partnership strategic.

Reading your number

Score withheld? Read the list of missing inputs as a to-do list. The first gap is usually sentiment, and a simple recurring survey closes it inside a quarter.

High CDE, flat CSI? Stop widening the funnel and look hard at onboarding. The partners you need to recruit next do implementation and success work. More referrers will not help.

CSI rising wherever one specific partner is involved? That partner is doing lifecycle work nobody asked them for. Name it, measure it, build the relationship deeper. Most ecosystems never have that conversation, because nobody ever had the number in front of them.

Where the numbers come from

Same answer as last week. Adoption, outcomes and sentiment come out of systems you already run: your CRM, your support desk, your survey tool. They import into the workspace, or arrive by export if you would rather start there. The only input CSI may ask you to create is the survey itself, and only if you do not run one today.

Run the five-customer audit

As with CDE, you can approximate this in a spreadsheet: five columns, your own weights, and the discipline to leave a cell blank instead of guessing at it. The template is free and ungated:

Take the 5-Customer Audit

If you would rather have it computed continuously, per customer and per partner, with the missing-input list telling you what to instrument next, that is what Entwine does. The free workspace monitors five relationships and takes no card:

Run it by hand and I would like to hear what pattern you find. Reply to this email and it comes straight to me.

This is part 2 of a series on the three ecosystem metrics. Part 1 covered CDE, Customer Discovery Efficiency. Next: ECS, the Ecosystem Completeness Score, and the uncomfortable question of how much of what your customers need your ecosystem cannot yet meet.

This is part 2 of a series on the three ecosystem metrics. Part 1 covered CDE, Customer Discovery Efficiency. Next: ECS, the Ecosystem Completeness Score, and the uncomfortable question of how much of your customers’ needs your ecosystem cannot yet meet.

Customer Discovery Efficiency is that answer, and it is one division:

CDE = average days-to-close for direct deals, divided by average days-to-close for ecosystem-sourced deals

Closed-won only, on both sides. A CDE of 1.0 means your ecosystem road is no faster than the direct one. A CDE of 2.0 means a customer who arrives through your ecosystem gets from need to signed solution in half the time.

What the ratio is comparing

A customer with a problem takes one of two roads to you. On the direct road they find you, you qualify them, you run your process. On the ecosystem road, someone they already trust says “talk to these people, this is the shape of your problem.”

Everything else in this metric is a decision about what to let into those two averages. Those decisions matter more than the arithmetic does.

The decisions underneath it

Count only what happened. CDE reads your opportunity records: created dates, close dates, referrer fields. No sentiment weighting, no estimated influence, no model deciding that a partner probably helped. If the score comes back empty, your data is empty, and that is worth knowing before it is worth fixing.

Keep the window recent. A partner who introduced you to three excellent customers in 2022 and nothing since is not accelerating anything today. Lifetime totals flatter dormant relationships. CDE is computed on a rolling recent period so the number describes the ecosystem you have now.

Count customer referrals as ecosystem. We segment the ecosystem side by referrer type, so partner-referred, customer-referred and other-referred sit against purely direct deals. Your happiest customer is doing discovery work for you whether or not anyone gave them a partner tier. Most measurement frameworks in this category ignore that entirely.

Show the median beside the average. One enormous deal that closed in a week will write you an efficiency story you cannot repeat. We compute both. If your average says 2.1x and your median says 1.2x, you do not have a fast ecosystem. You have one good anecdote and a measurement habit worth fixing.

Why speed, of all things

Because discovery speed shows up before the attribution argument starts. Sourced-revenue percentages collapse into debates about who touched what and whose number it was. Days-to-close is harder to argue with. The deal either sat in your pipeline for ninety days or it did not.

And the saving is not only yours. Take an illustrative pair: ecosystem-sourced deals closing in sixty days against a hundred for direct. That is forty days of selling cost you did not spend, and forty days your customer did not spend living with an unsolved problem. That second number is where partnership measurement usually stops and customer-success measurement should start.

Reading your number honestly

First, what I cannot give you. There is no credible cross-industry benchmark for CDE, from me or from anyone else publishing in this space, because almost nobody computes it yet. If someone hands you an industry average for this metric today, they are guessing. What CDE is good for right now is your own baseline and the direction it moves over the next three quarters.

No number at all means your opportunities are not recording a referral source. Fix the habit before the metric. A quarter of honest data beats a year of backfilled guesses.

Around 1.0 means your ecosystem is not accelerating discovery to close. That is not a verdict on whether partners create net-new demand. Volume, win rate and source uniqueness answer that question, not this one. Treat it as a signal to look closer rather than a conclusion.

Comfortably above 1 earns you the next question, which is whether those faster-won customers also succeed and stay. Speed of discovery is the door. Lifecycle contribution is the house, and that is the Customer Success Index, next week.

Where the numbers come from

Worth stating plainly, because a letter about continuous computation should not leave you guessing about the setup. CDE runs on opportunity records you already keep. They come in from your CRM, or from a spreadsheet export if you would rather start there and connect properly later. Nothing here asks you to collect new data. It asks you to populate the referrer field.

Run the five-customer audit

You do not need us for this one. Take five accounts, two healthy, two struggling and one you are unsure about, then map which partners touched each of them and when. The template I use is free and ungated:

Take the 5-customer Audit

If you would rather have it computed continuously, per partner, medians included, updating as deals close, that is what we built Entwine for. The free workspace monitors five relationships and takes no card:

https://app.entwineapp.io/Landing/Index?intent=starter#signup

This is part 1 of a series on the three ecosystem metrics — CDE, CSI (Customer Success Index) and ECS (Ecosystem Completeness Score) — and the four depths of partnership they measure.