The SaaS Distribution Course #19: How Deliverect Turned POS Partners Into an 80,000-Location Channel

TL;DR

Deliverect did not begin by asking thousands of restaurants to trust an unfinished integration product. Founder Zhong Xu says the team first handled orders manually for roughly 50 to 100 restaurants, charged about $50 per month, and spoke with 200 to 300 restaurants in the first three months. Only after the pain repeated did the team automate it.

Direct restaurant sales proved demand, but they could not provide the speed Xu wanted. The scalable move was to work with point-of-sale companies whose customers already needed online orders to flow into the till. Deliverect gave partners a missing capability, protected their customer relationships, attributed partner-sourced accounts back to them, helped their sales representatives explain the product, and used restaurant references to open more doors.

The reusable system is manual delivery → paid proof → customer reference → partner fit → partner attribution → rep activation → geographic replication. Deliverect now reports more than 80,000 locations served. That milestone is company-reported and unaudited; no reviewed source isolates the share produced by partners, or discloses channel CAC, payback, retention, or partner-sourced ARR.

What you will build

You will build a partner distribution ladder: a set of gates that turns direct customer learning into a channel another company is motivated to carry. Its outputs are a paid manual pilot, a proof asset, a partner-fit scorecard, an attribution rule, a rep-activation kit, and a geographic expansion gate. Each stage must create evidence needed by the next one.

Use this system if

• your product completes a capability another vendor's customers already request;
• the partner keeps the core account relationship while your product increases its value;
• you can prove the workflow manually before asking a partner to risk its reputation;
• the partner has a repeatable sales or implementation motion, not only a logo page;
• both companies can agree who owns the lead, support, renewal, and expansion.

Do not use it if

• customers have not paid for the underlying job yet;
• the integration creates a substitute for the partner's core product;
• you cannot trace a partner introduction through revenue and retention;
• every deployment still requires unrelated custom work;
• the proposed partner has audience reach but no reason to activate its sellers.

Verified case snapshot

Stage	Evidence	Limit
Manual proof	Roughly 50–100 restaurants were served manually before automation	Founder-reported range; records are not public
Learning volume	About 200–300 restaurant conversations in the first three months	Founder recollection, not a CRM export
Early price	About $50 per month	Founder-reported historical price
Channel	POS vendors brought access to restaurant customers and their sales teams	No partner-level pipeline or ARR disclosed
Early scale	More than $10 million ARR in approximately two and a half years	Host-stated and founder-confirmed; unaudited
Later outcome	80,000+ locations on Deliverect's current website	Company-reported snapshot; not channel attribution

Independent reporting helps establish the direction of travel. In January 2022, TechCrunch reported 20,000 locations across 40 markets, more than double the prior year's location count, alongside a $150 million Series D at a valuation above $1.4 billion. That does not prove which channel caused the growth. It does show that the company-reported location milestone is part of a multi-year expansion rather than a single launch-week spike.

The model: earn the channel in six linked stages

1. Perform the job before productizing it

A partner cannot distribute a hypothesis. Deliverect's first useful move was operational: take online orders arriving from delivery platforms and make sure restaurants could process them, even when a person had to move the information manually. The team was testing the job, not a polished interface.

Start with one narrow workflow. Write the input, the transformation, the output, the failure modes, and the person who feels the cost. Charge for the result. Free interviews establish interest; paid manual delivery establishes that the result is worth budget and exposes the ugly exceptions your eventual integration must handle.

Set a pass condition before the pilot: at least ten completed workflows, three paying accounts, and the same painful step appearing in 70% of deliveries. Stop if buyers admire the concept but will not pay, if every account needs a different result, or if the manual service cannot produce a reliable outcome.

2. Automate the repeated path, not every request

Xu contrasts Deliverect with an earlier POS product that took about nine months to build. The previous experience taught him how easily a team can invest before the demand signal is clear. With Deliverect, the manual flow came first; a usable product followed in roughly two to three months, while his broader estimate for the first product phase is three to six months.

Turn the manual log into a frequency table. Automate the highest-volume input, the repeated transformation, and the most common exception. Keep the long tail manual. A partner needs a dependable path for a recognisable customer segment, not a promise that the platform handles every possible configuration.

Pass when the standard workflow completes without founder intervention in at least eight of ten cases and setup time is falling. Stop feature work when a request has no repeated paid evidence. The output of this stage is the first thing a customer can reference without apologising for it.

3. Exchange concessions for proof and introductions

Deliverect did not treat an early discount as a gift. Xu describes asking a discounted customer to serve as a reference and introduce four or five other restaurants. That creates a causal bridge between product proof and distribution: the operating result produces trust, and the customer transfers that trust to the next account.

Make the exchange explicit. If you reduce price, ask for one bounded asset after success: a reference call, an anonymised proof sheet, or named introductions. Never ask before the result exists. Record the baseline, the operational change, the time to value, and the customer's own language. The proof should answer the objection a future partner will hear from its clients.

Pass when two customers will take a reference call and at least one introduction becomes a qualified opportunity. Stop discounting if references never happen or referred buyers do not share the same workflow. A discount without transferable proof is only lost revenue.

4. Select a partner whose product becomes better with yours

The POS company was not merely a directory with restaurant traffic. Its customers needed delivery-platform orders to enter the system they already used to run the restaurant. Deliverect filled that product gap while leaving the POS relationship intact. This is stronger than audience overlap because the products are operationally complementary.

Score candidates on five questions: Do the same customers experience the pain? Does your product increase the value or retention of the partner's core product? Can the partner identify the trigger in normal sales conversations? Is implementation repeatable? Can both sides support the customer without fighting for control? Require four yes answers before building.

Xu had more than a decade of restaurant and POS experience by this point. That network and vocabulary were a genuine unfair advantage. Founders without it should not pretend the channel will move as quickly. Use manual delivery and references to manufacture credibility, then begin with one regional partner whose customers closely match the proven segment.

5. Protect attribution and activate individual reps

Signing a partnership agreement is not distribution. Xu says Deliverect consistently gave the POS partner credit when a customer came through that relationship, even if the restaurant later contacted Deliverect directly. That reduces channel conflict. He also describes sitting with partner salespeople, teaching them the offer, and using small incentives such as gift cards to create the first behaviour.

Define attribution before launch: source, acceptance window, account ownership, renewal credit, support owner, and dispute process. Then build a one-page rep kit with the customer trigger, three discovery questions, qualification rules, one proof point, and the exact introduction path. Train five named people, not an abstract organisation.

Instrument the funnel as trained reps → introductions → accepted opportunities → activated accounts → retained revenue. Xu's example of signing ten partners that might each bring 100 customers per month is useful channel math, not a disclosed historical result. Replace assumptions with measured conversion rates after the first 30 days.

Pass when at least three trained reps create qualified introductions and one customer activates. Stop or redesign when the partner logo exists but no named rep changes behaviour, when disputes outnumber accepted leads, or when support cost destroys the economics.

6. Replicate only after the whole chain travels

Deliverect later opened ten offices in one quarter, according to Xu. COVID accelerated demand as restaurants urgently needed digital ordering, so the speed should not become a generic planning benchmark. The useful principle is narrower: geographic expansion followed a product and partner pattern that could travel.

Create a country gate with six fields: one anchor partner, twenty target accounts, local implementation readiness, support coverage, compliant contracting, and one owner. Expand only if the original market's partner funnel has stable activation and retention. An office does not create a channel; it supports one that already works.

Current ecosystem evidence is consistent with continued partner importance. Deliverect lists a broad integration-partner network, and DoorDash named Deliverect in its 2026 preferred integration program. These current pages establish active partnerships, not the economics of the early channel.

What failed, and what the success hides

The first warning is overbuilding. Xu's earlier company spent about nine months creating a POS product before putting it into the market. Deliverect's manual start reduced that risk, but it did not eliminate technical complexity. Integrations touch order accuracy, menus, payments, uptime, and frontline restaurant operations. A successful demo can still become an expensive support burden.

The second warning is channel conflict. If the vendor claims accounts a partner believes it sourced, the partner stops introducing customers. Attribution is not administrative cleanup; it is part of the product offered to the channel. The third is dependency. A route built on other companies' customers, APIs, and sales teams can change when partner priorities or platform rules change.

The fourth is timing. COVID pulled years of restaurant digitisation forward. It helped amplify a working system, but another founder cannot schedule a comparable shock. Finally, Xu now worries that integration infrastructure can become commoditised as AI improves. A channel advantage is not permanent. Deliverect's response is to move toward an intelligence layer that helps restaurants improve revenue and operations, with human approval and guardrails.

Your 30-day build

1. Days 1–5: pick the workflow. Interview ten operators, map one expensive handoff, and define the completed result. Reject broad platform language.
2. Days 6–12: deliver it manually. Recruit three paying design customers, execute ten workflows, log every exception, and record baseline and outcome.
3. Days 13–16: package proof. Automate the dominant path. Ask successful customers for two reference calls and four relevant introductions in exchange for any concession already agreed.
4. Days 17–21: rank partners. Build a list of twenty complementary vendors. Score shared pain, product lift, trigger visibility, implementation, and conflict risk.
5. Days 22–25: define the operating contract. Agree attribution, ownership, support, handoff, and reporting before announcing anything.
6. Days 26–30: activate five reps. Run a live enablement session, give them a one-page kit, and measure introductions through activation and retained revenue.

Implement it in Lead Scorer

Begin by storing the narrow customer, disqualifiers, workflow, permitted proof, and partner value proposition with icp-offer-context. Keep direct buyers and partner candidates as separate segments. A good restaurant customer proves pain; a good POS partner also has a repeatable way to recognise and introduce that customer.

Build the partner universe with daily-vertical-prospecting, one relevant ecosystem category at a time. Use signal-research-dossier to require at least two dated signals per candidate: a relevant integration strategy, customer segment, marketplace launch, partner program, expansion, or job opening. Skip companies with no verifiable evidence instead of filling the list with plausible logos.

Use icp-scoring-rubric for a weighted partner score. Give the largest weights to shared customer pain, product complementarity, sales-trigger visibility, and implementation repeatability. Penalise substitution risk, unclear ownership, weak support, or a marketplace that produces listings without seller activation. Calibrate the rubric against three partners you already understand before scoring the backlog.

Only after a company passes the rubric should lead-enrichment-pipeline identify the partnerships, product, integration, and sales leaders who can sponsor the motion. Contact data is not the first step. The useful record is the person plus the verified reason this partnership could improve their customers' workflow or their own product's retention.

Create an outreach draft with ai-authored-campaign, but keep every message queued for human review. The first note should lead with the repeated customer problem and one proof point, not “strategic partnership.” Ask for a short working session to compare workflow evidence. Use outreach-qa-audit to reject unsupported personalisation, multiple asks, or claims the sources do not permit.

Once replies arrive, use reply-triage to separate interest, timing, wrong-person routing, objections, and rejection. Do not automate acceptance. The account owner approves the next message, the partnership owner approves commercial terms, and support approves the handoff boundary. These gates protect the same trust that Deliverect's attribution practice protected.

Report both activity and economics: researched partners, qualified partners, named reps trained, introductions, accepted opportunities, activations, 30- and 90-day retention, support hours, sourced ARR, and disputes. Pass the pilot when one partner produces retained customers with tolerable support cost and the relationship remains complementary. Pause it when the evidence weakens, the partner cannot activate reps, or account ownership is contested.

Operator checklist

• One painful workflow has been delivered manually and paid for.
• The repeated path works without the founder in at least eight of ten cases.
• Two customers will provide references after receiving a verified result.
• The partner's core product becomes more useful, not less necessary.
• Lead and revenue attribution are written before launch.
• Five named partner reps can recognise the trigger and make the introduction.
• Introductions are measured through activation, retention, and support cost.
• Geographic expansion waits for a complete, repeatable channel chain.

Sources and evidence limits

The operating sequence comes primarily from Zhong Xu's March 2026 interview on The SaaS Podcast. Deliverect's company history and ecosystem pages provide current first-party context. TechCrunch independently reported earlier financing and location growth in 2021 and 2022.

The manual-customer count, conversation count, early price, development timing, ARR milestone, office expansion, attribution policy, rep activation tactics, and reference exchange are founder-reported and unaudited. The interview's “ten partners times 100 customers” is an example of channel leverage, not an actual disclosed result. Public sources do not provide early cohort retention, partner conversion rates, channel CAC, payback, gross margin, partner-sourced ARR, or the causal share of the 80,000-location milestone. Later first-party pages cite 95,000 and then 101,000-plus locations at later dates; this edition conservatively anchors the episode-time 80,000-location snapshot rather than blending different reporting dates.

Frequently asked questions

Should the product be built before approaching partners?

Build only enough for a reliable repeated workflow. The stronger sequence is paid manual proof, narrow automation, customer references, then one partner pilot. A broad integration platform without paid proof transfers product risk to the partner.

What is the first partner metric?

Count qualified introductions from named, trained reps. Marketplace views and signed agreements are upstream activity. Continue through activated accounts, retention, support cost, and sourced revenue before calling the channel healthy.

How should channel conflict be handled?

Write the source and ownership rules before launch, make attribution visible, and resolve edge cases quickly. If a partner introduced the account, preserve the agreed credit even when the customer later enters through a direct path.

When is it safe to enter a new geography?

When an anchor partner, target customer set, implementation path, support coverage, commercial rules, and local owner are all present. Expansion should replicate a measured system rather than compensate for a weak one.