I want you to create a print-ready 16:9 slide deck. Here's what I need:
- About: Baltra Foods (plant-based dairy, Germany) is considering entering the UK market and needs a go/no-go recommendation with an entry strategy. Facts: Baltra EU revenue 2025: €48M, growing 17% p.a. UK plant-based dairy market: £310M, 11% CAGR. UK competitor shares: Oatson 34%, Milkfield 22%, Verdana 9%, rest fragmented. Entry capex estimate: £6.5M. Target: 4% UK share within 24 months. Retail pipeline: 2,400 doors in year one via two grocery chains. Three entry modes under discussion: partner with a UK co-packer, DTC-first launch, or full retail launch.
- Audience: Board investment committee
- Tone: Crisp, executive
- Language: English
- Must include: An explicit recommendation up front; market sizing; competitor comparison; the three entry modes compared against criteria; financials with payback; top risks with mitigations; next steps with dates.
Do any research or thinking this needs, then build it. Use as many slides as the content needs.
HOW TO BUILD IT — this is plain authoring, not a coding job. Write ONE self-contained HTML file by editing the example below, and return it. Do NOT reach for a slide library (PptxGenJS or similar), an artifact/deck builder, a headless browser, or any script to generate, render, export or preview it — unPaper converts the HTML to PowerPoint and renders the preview itself when I paste it back.
Work in three passes.
PASS 0 — the brief. If a brief is given above, or we have been developing the topic in this conversation, start from that. If neither exists, settle the brief yourself before anything else: name the topic, the audience and the ONE message the deck must land, then do whatever thinking or research that needs. Only if there is genuinely no topic anywhere, ask me for one instead of building.
PASS 1 — the storyline. Draft the deck as a list of slide titles so it has a spine. For an argument or recommendation make each title a full-sentence finding (an "action title") that, read top to bottom, delivers the whole case — answer first, next steps last; for a narrative or teaching deck a shorter title is fine. One idea per slide; split a slide that carries two.
PASS 2 — the slides. For each title ask: what is the most effective way to SHOW this idea? The example locks the STYLE — palette, type, spacing, chrome — never the structure. Don't copy its slide order or density; reuse one layout across parallel slides or skip it entirely, but never use a grid or block just to tick it off — a deck that uses each layout exactly once reads as a tour, not an argument. Don't converge on the obvious arrangement either: two runs of the same brief should not produce the same deck. Then build to your ability; every rung yields a complete, convertible deck:
• Floor (any model): put your content into the example's own layouts. Even titles, short paragraphs and lists make a valid deck.
• Standard: pick the element that fits each idea — statement (one bold line) · agenda · two-column split (claim beside evidence) · comparison table · KPI stat row · chart · 2×2 matrix · process flow · ring (ONE share as an arc around its own number — the hero figure; a meter when several figures are compared, a ring when one carries the slide) · timeline · roadmap (a plan over NAMED periods — one row per workstream, each item's grid-column set inline to say which quarters it covers) · progress meter · track (span bars on one CONTINUOUS scale — elapsed time, a phase; set left AND width inline) · paths (two journeys compared on ONE shared scale — old vs redesigned, plan vs actual; never give each its own axis) · cascade (a waterfall walking one total down to another — price to margin, plan to actual; set bottom AND height inline) · ribbon (ONE total split into shares that add to 100% — a bridge is a cascade, a split is a ribbon; set each segment's width inline) · stack (tiers on a foundation — list what each tier is MADE OF as .slcp parts, not just a label) · pull quote · loop (a cycle that RETURNS to its first step, unlike a flow, which ends) · flag (an exception pinned to a step, carrying a number — standalone, never nested) · plot (points on two real axes — the 2x2 matrix is a grid, this is a coordinate plot; set left AND bottom inline) · graph (systems joined by edges, one heavier than the rest — also the way to show a hierarchy) · callout · cards · divider · closing — and vary it across the deck. Give most slides one visual anchor in the template's idiom (a small inline-SVG icon, an oversized numeral, an arrow, a rule or tinted panel): a deck of nothing but text in boxes reads unfinished.
• Full (if you can reason): treat the template as a design language and compose beyond the catalogue — a bespoke diagram of THIS content's actual structure, an unexpected grid, a full-bleed moment, a free canvas (keep the slide's header/footer frame, compose the open middle yourself). Give the deck at least ONE signature slide whose composition appears nowhere in the example. Freedom of arrangement, not of technology: build from convertible primitives — text, bordered/filled/gradient boxes, inline SVG, a real
, a data-unpaper-chart — and even an invented layout lands as native, editable PowerPoint objects; exotic CSS flattens to a picture.
THE CONTRACT — the file is machine-converted to PowerPoint, so each rule is load-bearing:
1. Copy the example's
Make the argument land
A consulting deck is an argument, not decoration: lead with the answer, prove it beside the claim, close with the decision. This deck is built that way — and shows you how.
02 Part one
Part one
The case, in four moves
Answer first, then the evidence, the options, and the decision.
02 / 27
03 Horizontal logic
Lead every slide with the answer, not the label
A topic label
"Q3 revenue"
Tells the reader nothing. They have to dig for the point.
An action title
"Revenue grew 18%, carried by enterprise"
Read the titles alone, top to bottom — the whole story is there.
03 / 27
04 In one slide
Executive summary
Buy the platform and partner on delivery — faster, cheaper and lower-risk
01
Building in-house takes nine months and carries the most risk for the least control gained.
02
A platform ships in three months and covers 80% of the requirement out of the box.
03
A delivery partner closes the gap with no permanent headcount commitment.
04 / 27
05 Evidence
Put the evidence right beside the claim
A claim earns belief when the proof sits next to it, not three slides later.
·The text states the so-what
·The chart carries the number
·One reads the other
Source: illustrative.
05 / 27
06 Evidence
Pair two lenses when one chart cannot carry the point
Different chart types, one message each — the title states the combined so-what.
Revenue share by segment
Enterprise · 54%Mid-market · 31%SMB · 15%
Growth by segment, % YoY
Source: illustrative.
06 / 27
07 Compare
Make every comparison a real table
Option
Cost
Time
Risk
Build in-house
High
9 mo
High
Buy a platform
Medium
3 mo
Low
Partner
Low
6 mo
Medium
Source: illustrative estimates.
07 / 27
08 The numbers
Anchor the story in three numbers
18%
Revenue growth, year on year.
▲ 4 pts vs FY23
3.2x
Enterprise ARR since FY22.
▲ 1.1x vs plan
−40%
Cost to serve, after automation.
▼ 12 pts vs FY23
Source: illustrative.
08 / 27
09 The quarter
Review the quarter on one screen, one tile per question
Revenue, €m
NPS trend
Cost mix
People · 62%Tools · 23%Other · 15%
−40%
Cost to serve since FY22 — the number the board asked about.
09 / 27
10 Drivers
Group your reasons so they don’t overlap
Demand
More buyers, larger contracts.
Delivery
Automation cut the cost to serve.
Pricing
A tier the market accepted.
Three drivers, no overlap — that is what MECE means.
10 / 27
11 The plan
Hold four parallel moves on one screen
Workstream 1
Stabilise the core
Fix churn in the two at-risk accounts before anything else.
Workstream 2
Price the new tier
Launch the mid-market tier the pilot validated.
Workstream 3
Automate delivery
Roll the onboarding automation to every region.
Workstream 4
Staff the gap
Two hires in customer engineering by Q3.
Four cards, one glance — use a 2×2 of cards when the moves are parallel, a flow when they are sequential.
11 / 27
12 How it works
Run it as four steps, not one big bang
Step 1
Diagnose
Step 2
Pilot
Step 3
Scale
Step 4
Embed
12 / 27
13 So what
The point
A chart is not an argument. Say the so-what out loud — what should the reader now believe, and do, because of this number?
13 / 27
14 Open canvas
Free canvas — keep the header and footer, then compose any layout you like in this space. Text, boxes, inline SVG, a table or a chart all still convert to native, editable objects.
14 / 27
15 Progress
Where the rollout stands
Teams migrated to the templateon track86%
Twelve of fourteen teams have moved; the last two land next sprint.
Decks converting without a manual fixwatching64%
Charts are clean; the remaining gap is hand-drawn diagrams.
Legacy files still on the old masterblocked31%
Waiting on the brand refresh before we touch them.
15 / 27
16 Capabilities
Part three
What the format buys you
One file, start to finish
The rules, the styles and the content travel together — nothing to attach, nothing to fetch.
Prints exactly as drawn
A fixed 1280×720 page and a real @page rule, so the PDF is the slide and not an approximation of it.
Converts to editable PowerPoint
Shapes stay shapes and text stays text — you can move a box in the meeting, not just look at it.
16 / 27
17 Where the time goes
A refund takes nine days, and seven of them are waiting
Two queues hold the whole process. The dashed line is the five-day promise.
Submitted and triagedSupport
0.5d
Waiting for finance reviewQueue
3.5d
Approved against the ledgerFinance
1d
Waiting for the payment runQueue
3d
Money leaves the accountTreasury
1d
Day 0Day 3Day 5Day 7Day 9
17 / 27
18 Activation
Of 1,000 trials, 163 become paying teams
Two steps — the first invite and week two — cost more than the other five together.
1,000Started a trial
−46Never opened it
−131No data imported
−86Stalled in setup
−239Never invited a colleague
−217Lapsed in week two
−94Budget refused
−25Churned in month two
163Paying
18 / 27
19 Architecture
Four tiers, one shared foundation
01
What customers touch
Everything a person opens, and nothing that stores.
Web appMobilePartner widget
02
The API edge
One gateway: auth, rate limits, and every audit record.
AuthRate limitsAudit log
03
Domain services
Nine services, each owning its own store.
BillingDocumentsIdentity
04
The data foundation
One warehouse every tier reads from, and nothing writes around.
WarehouseEvent bus
19 / 27
20 Positioning
Everyone else trades speed against control
The two dimensions buyers actually ask about, and the corner nobody occupies.
More you can change ↑
unPaper
Legacy suite
Agency template
Deck generator
Chat export
Faster to a finished deck →
20 / 27
21 Data platform
Four systems, and everything crosses one link
Three of these edges fail loudly. The replication into the warehouse fails quietly.
01 · Ingest
Event stream
38 sources, change data capture
02 · Warehouse
Analytics store
The only place all four agree
03 · Models
Semantic layer
412 metric definitions
04 · Activation
Dashboards and sync
What the business acts on
replication · 12 min lag
nightly build
governed metrics only
write-back
single point of failure
21 / 27
22 The engine
The loop compounds, and one step leaks
Every turn should add users. Activation is where the turn loses them.
More users
Every new team adds sources
Better data
Volume sharpens the defaults
Better product
Fewer steps to a finished deck
Each turn should cost less than the one before it
sharpens
attracts
feeds
activation drop −38%lag 11 days
22 / 27
23 The quarter
Three numbers decide the quarter
68%of trials reach a finished deckwas 51% in Q1
41%of revenue sits in ten accountsconcentration risk