Media bounds are set bucket by bucket - traditional, digital, e-com - and inside each bucket you start from the highest-spend channel and work down. In almost every market that top spender is TV (it takes 50-60% of most media budgets), and once TV is set it becomes the yardstick for every other traditional channel. Then repeat the whole logic for digital (pick its top spender, anchor the rest against it) and finally e-com. One structural rule to carry: e-com channels inside an offline model are always constrained hardest - offline channels earn better offline multipliers, and e-com media gets its fair read in the separate e-com model.

Read the bucket level first. A category leader whose spend share leans heavily traditional, with a modest total media ASSR, is a brand maintaining top-of-mind - expect the gravity of media contribution to sit with traditional, and expect base-driven contribution of 80-85% or more (module 1.6's top-down check).

The benchmark rule: a channel's ASSR is the floor its total (ST+LT) volume contribution should at least cross. The client logic is disarmingly simple - "if I put 2.5% of my revenue into this channel, I want at least 2.5% back", which is an ROI benchmark of ~1. Worked pattern (all numbers synthetic):

• Revenue 240M, TV spend 6M → TV ASSR = 2.5%.
• TV is a mature, long-term-skewed channel, so its ST contribution targets 70-80% of ASSR → 1.75-2.0% of volume.
• Pre-model ROI check: ROI = contribution / ASSR → 1.75 / 2.5 ≈ 0.7 ST ROI - the long-term side will carry it past 1.

ROI and contribution are co-dependent - set either and fine-tune the other. The team's practice is to set the ROI and check the contribution lands near ASSR; edge cases exist where this breaks, so always check both.

LTCF: the short-to-long bridge

The Long-Term Conversion Factor is defined off a channel's ST/LT split of total contribution: LTCF = LT share / ST share. The channel splits from KT need careful handling because they differ enormously:

Market maturity skews everything: in a developed market expect ~0.6-0.8 back short-term per unit invested and structured ROIs (do not promise 4-5); developing markets (India, Brazil) can plausibly show ROIs of 4-5 and higher LTCFs. The theory sheet adds the funnel rule: upper-funnel channels (TV, digital video, OOH, audio) earn LTCF > 1, sometimes 2-3× on business judgement; lower-funnel channels (search, retargeting, influencers) stay below 1.

The systematic-vs-burst signal (the radio trap)

Always inspect the weekly execution chart, never just yearly totals. The trap, rebuilt with synthetic numbers: radio spend is one-tenth of TV's while its CPM is one-quarter of TV's - naive arithmetic says radio buys impressions far cheaper, so it should earn a much higher ROI. But the weekly chart shows radio spent only in the first modelled year and went dark since: no carry-through in the analysis period, so it must not be allocated a high ROI. The general form: if a channel's spend collapsed much more than its CPM, its ROI inflates mechanically; systematic investment earns benchmark trust, bursts do not. Related pitfall: a tiny channel showing high ROI sits at the start of its saturation curve - never recommend "spend more" from that alone.

Competitor spends enter with a negative sign, and their bound is expressed as a fraction of your own direct contribution on that channel - never set independently, because the aggregated competitor series makes its raw correlation spurious. The KT ladder, with conditions:

These fractions are industry benchmarks from MMM studies - the standard starting point, with anomalies treated case by case. The underlying logic: competitor media's effectiveness rises with the volume share they capture.

A halo is the spillover of media tagged to one product onto another product's sales - a new variant's YouTube campaign lifting the core brand. The rule: size the halo at 5-10% (max ~10%) of that media's contribution in its own donor model, with the hard rule that a halo can never be greater than the direct media in the receiving model. So a variant campaign contributing ~9% in the variant's own model earns roughly a 0.5-1% halo contribution in the core model.

Why halo matters commercially: a fresh variant carries a huge media ASSR and its in-model ROI often lands below 1 - you are targeting tomorrow's customer. Attributing the spillover onto the core back to the variant's media is what justifies the launch investment honestly.

Two boundary cases:

• Master-brand media is NOT a halo. Brand-level media acts as a media funnel for every variant - the brand association sits with the whole brand. Halo runs only from media tagged to one product onto another product (module 1.1's mappability rules).
• India's format halo runs richer. With no master brand in India (media is tagged to formats), an established format's halo onto the newest format runs 2/3 to 3/4 of its own media contribution, the reverse direction ~1/4 - a general range of 25-75%, applying to new-variant and format cases only.

Before the media section is called done, three cross-channel checks must all tell one story:

• ROI ranking inversely tracks CPM ranking - as channels get costlier per impression, ROI should fall.
• Effectiveness ranking aligns with CPM ranking - costlier channels are more targeted and closer to conversion, so each impression should work harder.
• The research leg: every ROI decision is rule + secondary research - what media do this market's consumers actually consume? A market where OOH genuinely tops perception deserves a higher allocation than the pure spend/CPM arithmetic suggests.

And the data minimums that decide whether a channel or variant gets its own read at all: weekly model - at least 1 year + 1 quarter of sales AND media data; monthly model - at least 2 years. Below that, the investment is tested as a direct feature on the total, or via a nested/subtraction read - never as its own model.