How would you improve the average order value of Blinkit?

Before I touch any lever, three things change the whole approach, so I'll ask rather than assume.

I want to be explicit that AOV is a means, not the end, and anchor on contribution rather than vanity GMV.

I'll lay out the whole space first, then prune. Arithmetically, AOV = (number of items in the basket) × (average value per item). That gives an exhaustive tree of where AOV can come from:

1. More items per basket. Breadth: add a category the shopper isn't buying today (groceries, then personal care, household, snacks). Depth: more units of what they already buy (2 instead of 1, a larger pack).
2. Higher value per item. Premiumisation: trade up within a category to branded or premium SKUs, or push higher-margin private label. Mix shift: tilt the basket toward higher-ticket categories (electronics, appliances, large packs).
3. New high-value occasions. Capture order types that barely happen on Blinkit today: the planned weekly stock-up, party and festive orders, the high-ticket one-off. This is net-new basket, not reshaping an existing one.

Lever 1 is the safest because it adds value without touching the convenience promise. Lever 2 is real but rank-and-merchandising-led. Lever 3 is the biggest prize and the riskiest, because it can collide with the franchise. I'll let the data say which is live.

Average AOV hides the real picture, so I read it by occasion and by customer type.

Putting the tree against the segments, the live levers are:

• Checkout attach (more items, top-up orders). "Frequently bought together," "people also added," and last-aisle impulse SKUs (chocolate, batteries, gum) surfaced at checkout, personalised from the customer's own order history. Pure software, zero impact on speed.
• Per-segment free-delivery thresholds (more items). "Add ₹70 more to unlock free delivery," with the threshold set dynamically just above each segment's natural basket, not a flat high number. Tracked on net AOV so it isn't a disguised discount.
• Multi-buy and pack-size merchandising (depth + value per item). Surface larger packs and "2 for ₹X" combos on staples a customer reorders.
• One-tap reorder and replenishment (new occasion). "Your usual" rebuilds a full planned basket in one tap; "running low on milk?" nudges bring the weekly stock-up onto Blinkit.
• High-ticket and premium merchandising (value per item, new occasion). Rank premium and private-label SKUs higher for cohorts that have shown willingness to pay, and merchandise electronics and large packs to planned-shop occasions, not to top-ups.

Judged on impact against effort:

Do first (high impact, low effort, fully reversible):

1. Personalised checkout attach. Software-only, lifts items-per-basket directly, no friction. The single safest lever, so it ships first.
2. Dynamic per-segment free-delivery threshold. A classic, strong AOV lever, as long as it's set per segment and watched on net margin, not gross GMV.

Do next (medium effort): 3. One-tap reorder and replenishment reminders, to rebuild the planned weekly shop as a new occasion. 4. Premium and private-label ranking for willing cohorts, which lifts value-per-item and margin together.

Later (structural): 5. High-ticket category merchandising to planned-shop cohorts only.

What I'd explicitly deprioritise: a hard minimum order value. It's the bluntest AOV lever and it attacks frequency and the convenience moat head-on, the exact guardrail I set. Equally, sitewide discounting to inflate carts: that buys gross AOV at negative contribution, the opposite of the objective.

Sequence: ship attach and the dynamic threshold first (software, fast, reversible), then one-tap reorder, then high-ticket merchandising, with premium ranking in parallel since it's also software. Every change goes out as a cohort experiment against a hold-out.

Then measure in three layers, so I get an early read without fooling myself:

• Leading indicators (move within days): checkout attach rate, items per basket, share of orders crossing the free-delivery threshold, cross-category rate.
• Lagging headline (the outcome that matters): AOV gross and net of discount, and the real north star, contribution margin per order and per customer per month.
• Guardrails (must stay flat or better): orders per customer per month, 30- and 90-day retention, delivery-time SLA (the convenience promise), and return/cancellation rate on high-ticket items.

If items-per-basket moves but contribution per customer doesn't, the gain was discount-funded or it cost me frequency, and I'd re-cut rather than scale.

Before I commit, I'd pressure-test the top fix rather than present it as free:

Attach and thresholds suppress small, frequent orders, so frequency falls. This is the main risk and it threatens the moat directly. De-risk: hold-out cohorts, frequency as the hard kill-metric, and per-segment thresholds rather than a flat floor.

Threshold nudges are really discounts, so net AOV and margin don't move. De-risk: judge on net AOV and contribution, never gross GMV.

High-ticket pushes raise returns and tie up rider capacity. De-risk: cap by cohort, watch return rate and contribution per order.

The planned-shop push just shifts a basket the customer would've placed anyway, moving GMV around at higher delivery cost. De-risk: an incrementality test that counts only net-new GMV.

So: the biggest, safest win is frictionless personalised attach plus per-segment free-delivery thresholds to lift items-per-basket on the orders we already get, paired with one-tap reorder to bring the planned weekly shop onto Blinkit as a new occasion. I'd measure it on contribution margin per customer, not raw AOV, and hold order frequency as the kill-switch, so I never trade the ten-minute convenience moat for a vanity number.