Jägermeister In-App Campaign — Engineering Case Study | James Levine

Problem

Jägermeister wanted to run a hospitality activation through ChargedUp's order-at-table app: their branding on the splash and menu, a curated drinks bundle promoted on launch, and analytics back to their marketing team.

The risk for engineering was obvious: bespoke campaigns are how SaaS codebases die. Either we forked the app (and inherited two long-term maintenance trees), or we hard-coded a one-off branding mode (and made it harder to do the next sponsor).

Neither was acceptable. We needed a campaign engine, not a campaign.

Process

I designed the work around three principles:

Campaigns as data, not code. A new Campaign entity in DynamoDB holds theme tokens, copy, hero artwork URLs, drink-bundle SKUs and a venue allow-list. The React Native app loads the active campaign at boot from /v/<venue-id>.

One feature-flag toggle per venue. A campaign goes live by flipping a single flag. Off by default, no risk of leaking branding into non-participating venues.

Built-in analytics from day zero. Every interaction inside the campaign overlay (impression, view bundle, add to basket, complete order) emits an event with campaign_id, dropped into the existing analytics pipeline so the Jägermeister team got dashboards on day one.

I worked directly with Jägermeister's marketing leads on creative review — vetting the Lottie splash animation, the colour palette, and the way bundle offers expressed themselves on a phone screen. Two iterations of design review, then in-venue testing across five sites before full rollout.

Outcome

Jägermeister activated in 20+ venues across the campaign window. Order activation rate (scan → completed order) lifted by ~30% versus the non-campaign baseline at participating venues. Critically, zero engineering work was needed to onboard the next sponsor — the next campaign reused the same engine with a different Campaign row.

Campaign data model

type Campaign = {
  id: string;                 // CAMPAIGN#jager-2020
  venueIds: string[];         // allow-list
  startsAt: string;           // ISO
  endsAt: string;
  theme: {
    primaryColor: string;
    accentColor: string;
    fontUrl?: string;
    splashAnimation?: string; // Lottie JSON URL
    heroImage: string;
  };
  copy: {
    splashHeadline: string;
    bundleHeadline: string;
    ctaLabel: string;
  };
  bundle: {
    sku: string;
    items: Array<{ menuItemId: string; modifiers?: string[] }>;
    priceOverride?: number;
  };
};

Boot-time hydration

The PWA loaded campaign data on first venue scan and cached it in localStorage keyed by (venueId, campaignId). The <CampaignProvider> wrapped the entire app, exposing a single hook:

const { campaign, theme, bundle } = useCampaign();

Every UI component branched on campaign?.id — when null, the app rendered the default ChargedUp experience.

Why feature-flagged, not config-flagged

We considered storing the active-campaign ID directly on the venue document. Feature flags won because:

Instant rollback. Killing a campaign mid-night was a flag toggle, not a deploy.
Gradual rollout. We turned campaigns on across 5 → 20 venues over three days, watching analytics for issues before each tier.
A/B compatibility. We could send a fraction of customers in a participating venue to the non-campaign experience to measure lift.

Analytics pipeline

Existing infra: app emits events to API Gateway → Kinesis Firehose → S3 (Parquet) → Athena. We added two derived tables: campaign_impression and campaign_conversion, joined on (session_id, campaign_id). The Jägermeister team got a Quicksight dashboard reading those tables — no engineering involvement past wire-up.

Trade-offs

Themes as data, not stylesheets. We accepted constrained theming (colour tokens + a hero image, not arbitrary CSS) to keep the design system intact. The team disliked it briefly; the next two sponsor campaigns shipped in days because of it.
Bundle pricing. We stored an optional override on the campaign rather than splicing into the menu schema. Cleaner, but it meant the bar-side print receipt needed a small change to render the bundle SKU.