SMS credit audit

Contents

1 · What SMS actually costs

Three figures that must not be confused. Billed spend is the invoice: a prepaid monthly tier, paid whether or not the credits are used. Allocated cost is units consumed × the rate actually paid ($0.00645 per unit: $21,610 for a $33,525 rate-card allowance, where one unit is one cent at rate card), used to cost a flow or a campaign. Avoidable cash is what the bill can actually fall by, which only happens when the tier is changed at a renewal or when an overage is avoided; cutting usage inside a cycle saves nothing in cash.

Billing

Usage and return

Credits used = Σ (count × $usage_amount) over every Relayed Text Message event in the cycle, bucketed on the observed plan-period flip (between 10:28 and 10:52 UTC on the renewal day, read off the events' $plan_period tag). Revenue and orders are the Klaviyo flow and campaign values reports with Placed Order as the conversion metric, which use the account's attribution settings; see section 5. Delivered and failed are the Received and Failed to Deliver event counts for the same cycle.

2 · Where the credits go

Campaigns consume the bulk of every cycle; flows are small and steady. Within that, the cost drivers are format and length: an MMS is three credits per segment in the US, a plain SMS one, and a body with an emoji splits into 70-character segments instead of 160.

Country is the recipient's phone region on the event. Credits on failed deliveries are the Failed to Deliver events priced at the same (country, format, segments) rate as the ledger; Klaviyo bills them.

What a message costs, by country, format and length (Aug 31–Sep 30)

How often each person is texted

Unique profiles from the Relayed event by month. "Both" = campaign uniques + flow uniques − all uniques, an estimate of the overlap between the two programmes. Inbound replies in the window: 14,516 STOP-type, 11,529 unrecognised, 2,195 subscribe. Campaign sends skipped (consent, frequency cap, duplicates) by cycle: Jun 30–Jul 31 147,660, Jul 31–Aug 31 108,728, Aug 31–Sep 30 278,758, Sep 30–Oct 14 (to date) 105,358; skipped sends cost nothing.

3 · Flows and flow messages

Three cycles combined (Jul 31–Aug 31 to today). Credits and sends are the ledger; delivered, orders and revenue are the flow values report (Placed Order, account attribution window). Rows shaded red spent credits and produced no attributed order. "Break-even CM" is the contribution margin at which the attributed revenue exactly covers the credits: cost ÷ revenue.

Every SMS message

4 · Campaigns

Every SMS campaign sent since 30 Jun, keyed by send time (names repeat). Credits per recipient is the clearest single number: an image (MMS) or an emoji-laden body shows up here as 3 or more credits per person where a plain text costs 1 or 2. Campaigns younger than five days are still accruing attributed revenue.

5 · Profitability

Attributed revenue is not profit, and attributed is not incremental. The numbers above are what Klaviyo credits to a text under the account's attribution window; they overlap with email (an order can only be credited to one message, but the same buyer is usually on both lists) and they include orders that would have happened anyway. No SMS holdout exists in the account, so the causal return of SMS is unverified. What can be stated is the contribution the attributed revenue would leave after variable costs, and the margin at which it stops covering the credits.

Accounting basis

• Revenue. Klaviyo's Placed Order $value, checked against the Shopify orders it came from: it equals the Shopify order total after discounts, including tax and shipping. The sampled orders:

• Attribution window. Klaviyo's defaults are 5 days for email opens and clicks and, for accounts of this age, 5 days for SMS clicks with a shorter window on SMS opens and deliveries; last touch, one message per order. The account's actual setting is [DATA MISSING: Settings → Attribution screenshot; not exposed by the API]. The 24 Aug billing-page figure and the values reports here both use whatever that setting is.
• Refunds. Store-wide returns as a share of gross sales, from Shopify by day: 1.18% on average over the two complete cycles. Applied as a haircut to attributed revenue because refunds are not traceable to an SMS-attributed order without an order-level join.
• COGS. Gross margin from 90,045 Shopify orders × variant unit cost (window 2026-07-01 to 2026-10-01): Jul 75.0%, Aug 74.7%, Sep 72.4%. The revenue basis in that scan includes tax and shipping, so the product margin is slightly understated.
• Payment fees. Shopify Payments processing fees sampled from the payouts ledger (250 charges in the current cycle): 3.14% of the charged amount; the per-transaction rate reads 2.5% plus $0.30. Other gateways are not sampled.
• Fulfilment, pick-pack and outbound shipping cost. [DATA MISSING: 3PL and carrier invoices; not in Shopify or Klaviyo]. Shown below as a 10- and 20-point allowance, labelled as assumptions.
• Discounts are already inside the Klaviyo value (it is the discounted order), so they are not deducted again. Credits are deducted once, at the allocated unit rate.

Shopify net sales against attributed SMS revenue

Payment fee sample

The payouts ledger returned charges only for the current cycle under the date filter used; the earlier cycles are assumed to carry the same rate.

Contribution after SMS credits, by scenario

Break-even contribution margin on the whole programme: Jul 31–Aug 31 12.47%, Aug 31–Sep 30 6.50%, Sep 30–Oct 14 (to date) 18.84%. Even at 25% incrementality the programme as a whole covers its credits at every margin shown; the question is never whether SMS pays for itself in aggregate but which sends do, which is what the per-message and per-campaign break-even columns answer. Flow messages with zero orders have no break-even; they are pure cost.

6 · Savings opportunities

Credits per cycle are the three-cycle total scaled to a 30.4-day cycle. Cash is the credit value at the current unit rate and is realised only when the tier is stepped down or an overage avoided. "Calculable" means the credits are read directly from the ledger; "modelled" means they depend on an assumption stated in the validation column.

Timing, re-entry, suppression and post-purchase sends

• Smart sending is off on 187 of 187 live flow messages and on none of the campaigns (0 of 72 used it). With it off, a person in two flows and a campaign on the same day gets all three texts; the frequency table in section 2 shows the result.
• Quiet hours are off on 12 flow messages. Quiet-hour holds cost nothing; turning them off does not save credits, it only changes when the text lands.
• After a purchase, every abandonment and reorder flow carries a Placed Order filter at flow level (section 3 lists each flow's filters), so a buyer is not texted again by the same flow. Nothing stops a campaign from texting someone who bought that morning; the campaign exclusions used are hygiene lists and segment carve-outs, not recent buyers.
• Transactional flows (order shipped, delivered) are sent as marketing SMS: the transactional flag is false on every live message, so they reach only SMS-consented profiles and consume the same credits as marketing.

7 · Action plan and scenarios

Base = credits used over the three cycles since 31 Jul scaled to a 30.4-day cycle, launch included; the two complete cycles alone average 2,611,093. The per-cycle columns in section 6 show how much of each saving depends on a launch being in the cycle. Revenue at risk counts only the sends that stop (welcome #2/#3 and the same-day repeats); format and encoding changes keep every send.

Implement now

1. Turn off the 27 flow messages with zero attributed orders in three cycles (listed in red in section 3). Credits: 26,672 a cycle. Nothing attributed is lost.
2. Remove emoji and special characters from the plain-SMS bodies that split into 70-character segments (section 6 lists them). Same words, fewer segments.
3. Suppress numbers with repeated "Device disconnected" or "Device unreachable" failures from campaign audiences; those sends are billed and cannot deliver.
4. Decide the overage preference before the next full-list send: the cycle is at 90% with 17 days left and flows alone will use about 206,169 of the remaining 330,939 units, so one more full-list MMS (261,399 units) tips it into a one-off add-on or an automatic upgrade.

Test first

1. Plain GSM-7 copy versus the current MMS on the next full-list send: same audience, same hour, half get the image-free, emoji-free version. Judge on orders per delivered and revenue per delivered, never clicks. Under the current plan an MMS is a flat 3 credits up to 1,600 characters, so the saving only exists if the plain version fits two segments (306 characters) or, better, one (160 including the prefix and link).
2. Welcome message #2 and #3: 50/50 send versus skip at the day-1 node.
3. Second and third same-day campaign sends: hold out 10% of the audience from the later sends on the next launch day.
4. Flow messages sent as MMS (welcome #1 to #3, checkout #1): image and emoji off on the single highest-credit message first; the body must fit two GSM-7 segments or it saves nothing.

Retain

• Welcome message #1, abandoned checkout and abandoned cart first messages, and the first reorder text in each product flow: highest revenue per credit in the estate and the smallest share of credits.
• The current tier until a cycle without a launch shows where usage settles; a downgrade at 31 Oct would be undone by the next launch and Klaviyo does not refund mid-cycle changes.

8 · Experiment plan

Causal profitability is unverified for every send in this report. One holdout settles it for the whole programme; two smaller splits settle the two biggest modelled items.

All tests judge SMS on orders and revenue per delivered. Click rate is not a usable SMS metric in this account: roughly 95% of SMS clicks arrive within seconds of delivery from carrier and handset link scanners.

9 · Data gaps

• [DATA MISSING: 30 Sep 2026 invoice and any September–October top-up or overage invoices] — the billing page estimate was $25,770; the invoice record on file was read on 4 Sep.
• [DATA MISSING: line items for the 2 Jun, 9 Jul, 15 Jul, 17 Jul and 30 Jul 'Subscription' invoices] — needed to allocate a unit cost to the 4,289,949 credits used in Jun 30–Jul 31.
• [DATA MISSING: usage:read scope on the API key, so the Billing Usage API can be read first-hand] — a one-minute change in Settings → API keys; until then the $33,525 allowance and $30,170.84 used are second-hand, and the overage preference is unread.
• [DATA MISSING: Settings → Attribution: the account's SMS and email attribution windows].
• [DATA MISSING: fulfilment, pick-pack and outbound shipping cost per order] — the contribution table carries them as 10- and 20-point allowances.
• [DATA MISSING: revenue by recipient country] — Klaviyo attributes to a message, not to a country, so the Canadian credits cannot be matched to Canadian revenue.
• [DATA MISSING: any holdout or incrementality test] — none exists in the account; see section 8.
• [DATA MISSING: refunds on SMS-attributed orders specifically] — the store-wide refund rate is applied instead.

10 · Method, sources and reconciliation

• Usage ledger. POST /api/metric-aggregates on metric RPLmfe Relayed Text Message, measurement count, grouped by $usage_amount (the credits Klaviyo charged for that send), $message, Message Type, To Phone Region, Message Format and Segment Count; filters $flow per flow and $message per campaign; windows are the cycle bounds above in UTC. The endpoint returns at most 500 groups per call and does not paginate, so every pull was scoped small enough to stay under it; none hit the cap.
• Delivered, failed, unsubscribed, skipped, inbound. Metrics WKY8He, UkxcA8, W8X66B, TThDNb, RV6gt2, same windows.
• Orders and revenue. POST /api/flow-values-reports per cycle and POST /api/campaign-values-reports over the whole window (campaigns bucketed by send time), conversion metric UNe8Uh Placed Order; statistics recipients, delivered, failed, bounced, conversions, conversion_uniques, conversion_value, unsubscribe_uniques. Rows come back per message and are summed per flow here. Revenue per delivered is conversion_value ÷ delivered.
• Bodies, images, delays, filters, smart sending. GET /api/flows/{id}?additional-fields[flow]=definition and GET /api/flow-actions/{id}/flow-messages; GET /api/campaigns?include=campaign-messages for campaign bodies, media and render options. Segment counts in the tables are the observed counts on the Relayed events, not estimates from the bodies; the body analysis is used only to say whether the extra segments come from emoji.
• Billing. The invoice record read off Klaviyo Settings → Billing on 4 Sep 2026, the billing-page reading of 24 Aug 2026, and the Billing Usage API (GET /api/billing-usage, revision 2026-07-15.pre, scope usage:read) read on 14 Oct 2026 through a connector that holds the scope; the API reports usage at the account-family level in rate-card dollars.
• Shopify. ShopifyQL FROM sales by day for gross, discounts, returns and net sales; Shopify Payments balance transactions for fees; per-variant unit cost for COGS; three orders fetched by id to check what Klaviyo's $value contains.
• Seed sends. GlockApps and internal test sends are excluded from the campaign tables by name pattern; no SMS campaign in the window matched it.

Reconciliation of the per-message ledger to the cycle totals

Any gap is sends whose $flow or $message is blank on the event (deleted flows keep sending under a blank id) and is reported, not redistributed. The ET-day calendar in section 2 differs from the exact-cycle totals by the credits sent between midnight ET and the renewal flip on each renewal day: 06-30 48, 07-31 132, 08-31 851, 09-30 939.

Generated 14 Oct 2026 14:06 ET from live pulls; nothing on this page is typed by hand.