Deliverect · Strategic Product Review
Senior Product Manager · Skills Showcase

Campaign
Manager

One place a merchant runs all marketing. One number it answers for: margin protected.

Presented byAlejandro Valero Navarro
AudienceMaría Palacios · Jelte Vrijhoef
Duration45 min, Q&A included
01The argument

Three decisions, up front.

Decision 01

Margin, not campaigns

Blanket discounting is an unmanaged cost line. Targeting it is the only defensible product here.

Decision 02

One object, no reorg

Four teams keep their capability. I own the merchant surface and the shared Promotion object.

Decision 03

Autonomy is earned

The agent climbs a trust ladder per merchant, against published gates. It can also fall back down.

The metric on the wall: margin protected per merchant, per month, the only number that can't be gamed by shipping a worse product.

02How I'm spending 45 minutes

Nine sections. Here's what I cut.

Deep, 25 min
  • The problem worth solving
  • Internal fragmentation
  • AI autonomy and trust
  • The risk that kills it
  • Pricing and the metric
Light, 7 min
  • Roadmap, 0–18 months
  • Competitive position
  • Naming and IA
  • Store readiness
Out of scope
  • Localisation by region
  • Wider multi-agent strategy
  • Paid social and search
  • Creative tooling

The rule: left column = expensive to reverse. Right column = decidable later, with better information.

01

The problem
worth solving

Six pains in the brief. Only one of them is a product.
03Problem definition

Four pains. Four different products.

If the real problem is……we're buildingWho else can build itVerdict
SpeedA campaign workflow toolAny POS suite, any loyalty vendorNecessary. Not defensible.
CostA self-serve agency replacementMarketing specialists, alreadyA wedge, not a moat.
MeasurementCross-channel reportingHard for all, but it's a dashboardProof, not product.
MarginA decision engine: which store, which offer, when, and when to stopOnly a platform between the operator and every channelThis is the product.

Not dropped, sequenced. Speed gets them in. Measurement proves the claim. Margin is why they pay.

04The thesis, in one picture

A 20% estate-wide promotion, 500 stores.

Discounted, didn't need it Genuinely underperforming
360
stores discounted that would have sold anyway
The cost
€3m
margin eroded per year, one campaign type, one brand (illustrative)

Not a marketing problem. An unowned cost line, and the one thing Deliverect's position lets us fix.

05Vision and goals

Every euro of discount goes to a store that needed it.

01 · The claim

Protect margin

Targeted, condition-based activation. Measured in margin protected, not campaigns launched.

02 · The enabler

Earn AI trust

Human-approved, guardrailed, auditable. Trust caps every other ambition here.

03 · The outcome

Deepen the platform

Marketing becomes the second reason they can't leave. ARPU follows stickiness.

Create → Communicate → Measure → Improve. My addition: and stop, automatically, when it stops paying.

02

Resolving the fragmentation

Four teams, two promotion creators, one merchant who doesn't care about our org chart.
06Diagnosis

Nobody owns the noun.

Team A

Coupons

Codes, redemption logic

Team B

Campaigns

Marketplace discounts

Team C

Loyalty

Points, tiers, rewards

Team D

CRM

Segments, lifecycle

What the merchant gets

  • Two places to create a promotion, different words for the same thing
  • A marketplace tool living inside Direct Suite
  • The same front-end, built four times

What I'm not proposing

  • A reorg, costs two quarters of goodwill I need elsewhere
  • A rewrite, the promotions API and the POC are assets
  • A committee, shared ownership is no ownership
07Operating model

One front door. Many kitchens.

I own
  • The merchant surface
  • The Promotion object
  • Activation logic and guardrails
  • Measurement
They keep
  • Capability, roadmap, headcount
  • Domain depth I don't have

The primitive, five fields

mechanic% · fixed · BOGO · code · member reward
targetlocations resolved by condition, never “all”
conditionvs. baseline, daypart, weather, stock, event
channelmarketplace · direct · push · email · SMS
guardrailmargin floor · max discount · budget · blast radius

Guardrail is the wedge: no team has it, every team needs it, and it only works shared.

08Two creators, one survivor

Strangle it. Don't switch it off.

Month 0–3

New is default

  • Every new merchant lands in Campaign Manager
  • Old creator frozen: bugs only
  • Both write to the same object
Month 3–9

Migrate by cohort

  • SMB self-serve, one-click carry-over
  • Enterprise account-managed, never mid-contract
  • Blocking gaps are P0 for my team
Month 9–12

Read-only, then gone

  • Read-only at 85% migrated
  • Decommissioned at 97%
  • Kill criterion published day one

To the team that built it: your logic survives, your surface doesn't. In writing, before I ask them for anything.

09The boundary I'd defend

One question settles every edge case.

Does this offer need to know who the customer is?

No → Campaign Manager

Acquisition

Anonymous, store- and menu-level. Wins a transaction that wasn't going to happen, including on channels we don't own.

THE LINE
Yes → Loyalty

Retention

Known customer, history, entitlement. Changes the next transaction. Needs first-party identity.

Why it survives: anyone can apply it without me, it maps to a real dependency (identity resolution), and shared cases live on the object instead of in a negotiation.

10Leading without authority

Four levers. Only one is mine to pull.

01 · Mandate

A signed Product Goal

One sentence owned by the CPO: the merchant has one place to create an offer. I'd ask for it in week one.

02 · Contract

An API, not a meeting

Versioned promotion contract with a public changelog. Alignment becomes an artefact, not a calendar invite.

03 · Incentive

Make my way the cheap way

I absorb front-end, guardrails, support and measurement. Integrating should delete backlog items. If it doesn't, my model is wrong.

04 · Evidence

Ship one public proof

One merchant, one quarter, one number. The first proof buys the next three integrations.

03

AI, autonomy
and trust

The agent isn't the hard part. Getting an operator to switch it on is.
11Autonomy is earned, not configured

The trust ladder.

00
Shadow

Agent proposes, we log what would have happened. Nothing executes.

6 weeks
01
Recommend

“Store 114 is 22% below baseline, three offers from your library.” Human picks.

v1 ships here
02
Approve each

Agent proposes a specific activation. One click, expires if ignored.

Gate: ≥70% accept / 30 activations
03
Approve by policy

Pre-authorised class: “≤10% off, ≤20 stores, weekdays, within margin floor.”

Gate: 60 days, zero breach
04
Autonomous, bounded

Activates and deactivates inside a signed envelope. Daily digest, not a request.

Opt-in, revocable

There is no rung five. The agent selects from a human-authored library and never invents an offer, otherwise every incident becomes unattributable.

12Guardrails

Four hard limits. Human-set, unbreakable.

Limit 01

Margin floor

Per item, per store. Neither the agent nor a human in a hurry crosses it.

Limit 02

Max discount

Brand-level ceiling per mechanic.

Limit 03

Budget cap

Daily, per store and per estate. Hard stop.

Limit 04 · the missed one

Blast radius

Max share of the estate on an agent-triggered promo at once. Caps a bad rule.

Always human

  • Authoring the promotion library
  • Setting guardrails and the envelope
  • Brand voice and price architecture

Agent, once earned

  • Detecting underperformance vs. baseline
  • Selecting from the approved library
  • Activating and, the valuable half, deactivating
13The risk that ends the product

An unexpected discount, on a store that didn't need it.

Before, prevention

  • Six weeks of shadow mode
  • Four hard guardrails, blast radius included
  • Staged rollout: 5 stores → 50 → estate
  • Plain-language reason on every activation
  • Simulation against last month's data

After, when it happens anyway

  • One-click rollback across every channel, from any screen
  • We tell them first, with the exact cost
  • Full audit trail: rule, baseline, guardrails checked
  • Automatic demotion, the ladder runs both ways
  • Pre-written incident comms

Trust is destroyed by surprise, not by error. Operators forgive €4,000 they hear about within the hour. They don't forgive €400 they hear about from a store manager.

14The pain with no owner

A marketing win that becomes an operational loss.

0
teams own what happens in the store when a campaign goes live

Store readiness, built into activation

  • Campaign scheduled → store notified with expected uplift and items
  • Store acknowledges, or flags low stock, which narrows or blocks activation
  • Post-campaign: snooze rate, rejections, rating movement

Metric: % of activated stores that acknowledged before launch. Below 80%, the campaign is a liability whatever the revenue says.

15Roadmap · 0–18 months

What I'd be judged on, per phase.

Phase 0 · 0–6 months

One object, one door

  • Promotion primitive shipped and versioned
  • Campaign Manager the single creation surface
  • Margin targeting live with 3 design partners
  • Agent in shadow mode

Judge me on: margin protected for 3 partners

Phase 1 · 6–12 months

The agent earns rung two

  • Recommend, then Approve-each, per merchant
  • Push, email, SMS in the same flow
  • Store readiness loop shipped
  • SMB migration complete; paid tier live

Judge me on: acceptance rate ≥70%

Phase 2 · 12–18 months

Orchestration, not creation

  • Policy autonomy for merchants who earned it
  • Old creator decommissioned
  • IA rework; loyalty mechanics on the object
  • More marketplace channels where commercials allow

Judge me on: share of spend orchestrated in-platform

16Business impact

They don't pay for campaigns. They pay for the discount they didn't give.

For the operator

Margin60–70% of blanket discount spend recoverable, on the earlier model
CostAgency retainer partially displaced in mid-market
SpeedLive in minutes, weather and event moments become reachable
OpsFewer stock-outs on promoted items; protects rating and ranking

For Deliverect

StickinessSecond lock after order flow, holds the marketing team too
ARPUPaid tier sold to a budget holder we don't touch today
PositionThe only layer spanning first-party and every marketplace

What I can't yet defend: clean incrementality where we don't control the experiment. Phase 0 uses matched-store comparison and says so.

17Packaging and pricing

The paywall sits at intelligence, not capability.

Included · €0

Create

  • Promotions on connected marketplaces
  • Manual targeting and scheduling
  • Single-channel reporting

Why free: cheap to serve, moves the habit inside the product.

Paid · per location / month

Orchestrate

  • Cross-channel calendar, unified ROI
  • Push, email, SMS in one flow
  • Condition-based targeting
  • Store readiness loop

Claim: one place, one number.

Premium · + platform fee

Protect

  • The agent, at every rung
  • Margin floors, item-level cost
  • Blast radius, policy autonomy
  • Simulation and incrementality

Claim: margin protected, in euros.

Messaging: credits at cost plus a transparent fee, hard cap, never a margin line. Profit per send makes our incentive volume and theirs efficiency.

18Product-led growth

Show them the money they're losing, before asking for any.

Hook · zero friction

The Margin Leak Report

  • Ungated, automated: “last month you discounted 412 stores. 289 were already above baseline.”
  • Built from data we already hold, no integration, no setup, no sales call
  • Nobody else can generate it: it needs cross-marketplace performance per store
  • Free promotion creation as the landing surface

Aha moment: a number with a euro sign on it

Habit · in-product triggers

The agent as the salesperson

  • At Recommend rung, free: “Store 114 is 22% below baseline. Three offers available.”
  • Acting on one is free. Acting on forty, automatically, is the paid tier
  • Every recommendation ignored is a quantified, visible cost
  • Self-serve upgrade, no call, for anything under ~50 locations

The paywall is scale, not capability

Expand · two loops

Seats, then networks

  • Inside the account: marketing invites ops, area managers, store managers, store readiness pulls in people who never had a Deliverect login
  • Across the network: a franchisor adopting it onboards its franchisees, each of whom becomes a paying account of their own
  • Priced per location, so expansion is automatic as the estate grows

The franchise loop is the real virality

The PLG metric I'd watch

Time from first login to first targeted promotion activated. Not signups, not MAU, the moment the operator does the thing the product exists for. Everything upstream is optimisation of that one number.

Where PLG stops, honestly

A 900-store brand does not self-serve a margin policy. PLG wins mid-market outright and gets us into one region of an enterprise account. The enterprise close is sales-led, and pretending otherwise is how PLG strategies quietly fail.

19Go-to-market

We already have distribution. What we don't have is the buyer.

The unfair advantage

100,000+ locations already connected

Zero-CAC distribution into an install base that already trusts us with order flow. Most marketing platforms spend years buying the access we already have. First campaign: in-product, to existing merchants, not paid acquisition.

The gap · the real GTM risk

We sell to ops. This is bought by marketing.

Different persona, different budget line, different objection set, and finance is the one who cares about the margin claim. Our AEs have never had this conversation. That's a bigger threat to this product than any competitor.

Phase 0 · Months 0–6 · Prove

3–5 design partners

  • Hand-picked mid-market and enterprise, in regions where marketplace integrations are live
  • Hand-held, non-scalable, deliberately: we're buying evidence, not revenue
  • Exit criterion: one published margin-protected number we'd stake the pricing on
Phase 1 · Months 6–12 · Land

Install base + self-serve

  • Margin Leak Report to every eligible existing merchant
  • Free tier on by default; self-serve upgrade under 50 locations
  • Sales-assist for mid-market, armed with the ROI calculator
  • Marketing buyer persona built, with its own enablement track
Phase 2 · Months 12–18 · Expand

Enterprise and networks

  • Enterprise land-and-expand: one region, then the estate
  • Franchisor-led rollouts as a named motion, with franchisee onboarding kits
  • POS and marketplace partner channels, where commercials allow
  • Rules of engagement agreed with Loyalty so we don't collide in accounts

What I'd build before selling anything: an ROI calculator that turns a merchant's own estate data into a euro figure. Same engine as the free report. It's the sales asset, the marketing asset and the product hook, and without it “we protect your margin” is just a claim in a deck.

20Measuring it

One number that can't be gamed.

Leading, is the model working?

  • Share of promotions created inside Campaign Manager
  • Agent acceptance rate
  • Targeted vs. blanket ratio
  • Store acknowledgement rate

Lagging, is the business working?

  • Margin protected, in euros
  • Free→paid conversion and marketing ARPU
  • Retention delta for Campaign Manager users
  • Merchants remaining on the old creator
The metric on the wall

Margin protected per merchant, per month

The only metric that gets worse when we build the wrong thing, and it's already on the operator's P&L.

21Competitive position

Two competitors, converging from opposite sides.

POS-bundled suitesLoyalty specialistsCampaign Manager
OwnsThe till and the menuThe known customerThe layer between operator and every marketplace
Can'tReach marketplace demandAct on marketplace performance per storeMatch their creative or CRM depth, so don't try
How we loseBy building a worse version of their product inside oursBy integrating where they're better and owning the decision layer

Naming and IA: “Marketing Manager” is the buyer's own word for the job. And the 1P/3P split is our org chart leaking into their navigation, channel is a property of a campaign, not a section of the product.

22Intellectual honesty

Three things I don't know.

Unknown 01

Will teams adopt the contract?

My model assumes integrating is cheaper than maintaining their surface. False if that surface is load-bearing elsewhere.

Two half-days per tech lead, before any spec.

Unknown 02 · the worrying one

Is margin data real?

Everything rests on item-level cost per store. Modelled or missing, and this becomes a targeting product.

Audit 10 merchants' data in week one.

Unknown 03

Will they pay for margin?

A finance story sold to a marketing buyer. That mismatch may need a different route to the sale.

8 mid-market operators, 3 enterprise CFOs.

Deliverect · Strategic Product Review
In one sentence

A margin product that happens to work through marketing.

Which is exactly why four surfaces have to become one. You can't protect margin from four places at once.

Thank youWhat would you push back on first?
01 / 27