Short answer: switching MMPs without losing history takes three to four weeks for most teams: export the data you need, run both SDKs in parallel for one to two weeks, compare installs and revenue by channel, move ad-network postbacks over in stages, then cut over. Historical data does not transfer between MMPs, so you keep it as exports. If you are moving to Linkrunner, a mobile measurement partner that unifies attribution and deep linking, the SDK integration takes the same day for most teams, fully live within a week, and billing is per attributed install (₹1.20 down to ₹0.80) after 25k free installs, with no annual lock-in.
You're spending ₹2.5 lakh a month on attribution, the renewal quote is up 40%, and engineering is stretched. Switching feels like changing tyres on a moving car: live campaigns, postbacks that train Meta and Google, and two years of cohort history that finance still asks about.
The fear is usually bigger than the work. The engineering part of a migration is small. The real job is data continuity: making sure you can still make confident budget decisions before, during and after the switch. This playbook covers the timeline, what it costs, what to export, how to run the parallel phase and the thresholds that tell you it is safe to cut over.
How long does an MMP migration take?
An MMP migration is the process of replacing one mobile measurement partner with another while keeping attribution, postbacks and reporting continuous. For a team with engineering support and decision authority, it runs three to four weeks end to end. Procurement, multi-team sign-off or custom integrations add another two to three weeks.
| Phase | When | Owner | Exit criterion |
|---|---|---|---|
| Export and event mapping | Week 1 | Growth lead and analyst | Exports stored, mapping sheet signed off |
| New SDK alongside old | Week 1 | Mobile engineer | Both dashboards receiving installs and events |
| Parallel run and comparison | Weeks 2 to 3 | Analyst | Paid installs within 10% by channel, revenue within 15% |
| Postback switch | Weeks 2 to 3 | UA manager | Conversions visible in Meta, Google and TikTok from the new source |
| Cutover and old SDK removal | Week 4 | Engineer and UA manager | Day-1 checks pass, old postbacks removed |
The engineering effort is roughly 8 to 12 hours spread over those weeks: the SDK swap, event mapping checks, about an hour of postback setup per ad network, and two to three hours of QA.
What does it cost to switch MMPs?
There is rarely a switching fee. The cost sits in four places:
| Cost line | What drives it | How to keep it small |
|---|---|---|
| Parallel-run overlap | Paying both platforms for two to three weeks | Time the switch to your renewal date |
| Old contract | Notice periods and any remaining annual commitment | Read termination and export clauses before you start |
| Engineering and QA | SDK swap, event mapping, postbacks, testing | Keep event names identical unless you are deliberately cleaning them up |
| Ad-network learning | A short relearning period after postbacks move | Avoid festive weeks and big launches |
Worked example: 100k installs a month
- Legacy MMPs typically cost ₹3 to ₹8 per install at this scale, so roughly ₹3 lakh to ₹8 lakh a month, or ₹36 lakh to ₹96 lakh a year.
- On Linkrunner's published INR card, 100k installs a month falls in the ₹1.00 band, so about ₹1 lakh a month, or ₹12 lakh a year, before the one-time 25k free installs.
- A three-week parallel run attributes roughly 75k installs on the new platform, which is still in the ₹1.00 band: about ₹75k of overlap.
- Against an annual gap of ₹24 lakh to ₹84 lakh, the overlap pays for itself within about two weeks of the switch.
For published prices across vendors at 100k and 1M installs, see what an MMP actually costs in 2026.
What historical data should you export before switching?
Historical attribution does not move between MMPs. You keep it as exports, so be deliberate about what you take.
Export (at least 30 days, ideally your full payback window):
- Attributed installs by source, campaign, ad set and creative
- Revenue events by install cohort and source
- ROAS by channel at D0, D7 and D30
- Retention cohorts by acquisition source
- Fraud and rejection data
Nice to have: click and impression logs, test results, custom funnels. Skip: device-level logs beyond 90 days and deprecated event schemas.
Export the same calendar periods from last year if you report year on year. Store everything in BigQuery, S3 or, at small volumes, a shared drive, with clear date stamps. If your current MMP rate-limits or charges for exports, prioritise the 30 to 90 day window that matches your payback period, and check the export clauses in your MMP contract before giving notice.
Event mapping
Build a three-column sheet: current event name, new event name, and a revenue flag. Most broken migrations trace back to one of these:
- Different names for the same action ("checkout_complete" against "purchase_completed")
- Missing revenue value or currency parameters
- Assuming event parameters carry over automatically (they don't)
- Undocumented events used in SKAN conversion value schemas
How to run the parallel tracking phase
Install the new SDK alongside the old one, initialise both at app start, and fire identical events to both. You are not switching yet; you are proving the new platform matches reality before your old contract ends.
For 7 to 14 days, compare daily:
- Attributed installs, total and by source (Meta, Google, TikTok, organic)
- Revenue event counts and values by source
- ROAS by channel at D0 and D7
- Click-to-install time distribution
Pull numbers from each platform's API or CSV export, not screenshots. A simple sheet with date, old MMP, new MMP and variance per channel is enough.
Variance thresholds
| Signal | Normal: proceed | Investigate | Pause the migration |
|---|---|---|---|
| Total attributed installs | Within 5% to 10% | Over 15% on a paid channel | Over 30% on any major channel |
| Revenue events | Within 8% to 12% | Over 20% | Revenue events missing entirely |
| Campaign coverage | All live campaigns present | Some campaigns missing | A whole network missing |
| Direction | Random day-to-day noise | New platform always 20% higher or lower | SKAN postbacks not arriving |
The usual causes are predictable. Fewer Meta installs in the new MMP usually means an attribution window mismatch (7-day click in one, 1-day in the other). Missing revenue usually means an event name or currency parameter mismatch. Organic differences are often fine, because organic definitions differ; focus on paid channels, which drive budget. Our attribution discrepancy diagnostic guide walks through each cause.
How to move ad-network postbacks without breaking campaigns
Postbacks are what train Meta, Google and TikTok, so move them in stages:
- Add the new platform's postbacks without removing the old ones. You need ad account IDs, event mapping and windows that match your settings.
- Run both for 5 to 7 days. Confirm conversions appear in each network and counts roughly match.
- Remove the old postbacks. Leaving both on double-counts conversions.
- Meta: connect the new source to the same dataset in Events Manager and watch delivery for 48 hours.
- Google: create the new conversion actions, keep one Primary action per event so Smart Bidding doesn't double-count, and expect a short learning period.
- TikTok: update the app event source in Events Manager and check the campaign's optimisation event hasn't reset.
Expect a two to three day learning period and CPA movement of 10% to 15% that settles within a week. Done this way, switching postbacks does not reset account learning.
Cutover checklist
First 6 hours after removing the old SDK:
- Installs arriving in real time in the new MMP
- Conversions visible in each ad network
- Revenue events firing with the right INR values
- No SDK errors in logs
After 24 hours: install volume within normal range of the previous day, ROAS calculated the same way, no missing campaigns, stakeholder reports without gaps.
After 30 days: paid installs by channel within 10% of the pre-migration baseline (allowing for seasonality), no high-value event lost, and the actual platform saving calculated.
Frequently asked questions
Will I lose historical data when I switch MMPs?
Not if you export it first. Historical attribution stays with the old platform, so export 30 to 90 days of installs, revenue and ROAS before your access ends and keep it in your own storage for year-on-year comparisons.
Can I run two MMPs at the same time?
Yes, and you should for one to two weeks. Both SDKs track the same events and you pay each platform for its own usage. Budget for about three weeks of overlap.
How long does the Linkrunner SDK integration take?
Same day for most teams, fully live within a week. SDKs cover iOS, Android, React Native, Flutter, Expo, Unity, Capacitor, Cordova and web.
When is the worst time to migrate?
During festive sales or a major launch. Pick a period with predictable traffic so anomalies stand out.
How do I compare year on year after switching mid-year?
Document window, organic and event definitions, keep the parallel-run data as a calibration bridge, and present comparisons as directional trends.
Making the switch
A migration succeeds when it is treated as a data continuity project, not a vendor swap: export first, validate in parallel, move postbacks in stages, then cut over. If you are still choosing where to move, our comparison of AppsFlyer alternatives for Indian app teams sets out the options.
If you want to see how attribution, deep linking and postbacks look in one dashboard before you plan a switch, request a demo from Linkrunner.
